SYSTEMS, APPARATUS, ARTICLES OF MANUFACTURE, AND METHODS FOR DATA USAGE MONITORING TO IDENTIFY AND MITIGATE ETHICAL DIVERGENCE

FIELD OF THE DISCLOSURE

This disclosure relates generally to monitoring data and, more particularly, to systems, apparatus, articles of manufacture, and methods for data usage monitoring to identify and mitigate ethical divergence.

BACKGROUND

Data management systems gather data and/or otherwise monitor many different complex activities and processes. The consumption of data for a specific activity, tool, or task can be intensive and in many situations is performed by artificial intelligence (AI) and machine learning (ML) agents within an edge network. Such AI/ML agents perform various analytics efforts on consumed data. It is difficult to determine the reasons behind data consumption by AI/ML agents. Thus, it is complicated to determine whether such data consumption by AI/ML agents are for ethical or unethical purposes.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 illustrates an overview of an Edge cloud configuration for Edge computing.

FIG. 2 illustrates operational layers among endpoints, an Edge cloud, and cloud computing environments.

FIG. 3 illustrates an example approach for networking and services in an Edge computing system.

FIG. 4 illustrates deployment of a virtual Edge configuration in an Edge computing system operated among multiple Edge nodes and multiple tenants.

FIG. 5 is a schematic diagram of an example infrastructure processing unit (IPU).

FIG. 6 is a block diagram of an example adaptive data management (ADM) system to implement adaptive data management in a network environment.

FIG. 7 is a flowchart representative of example machine readable instructions and/or example operations that may be executed and/or instantiated by processor circuitry to generate an example recommendation to integrate a hardware, software, and/or firmware feature in a semiconductor-based device (e.g., a silicon-based device).

FIG. 8 is an illustration of an example edge network environment including an example edge gateway and an example edge switch that may implement the example ADM system of FIG. 6.

FIG. 9 is a block diagram of a portion of the example ADM system of FIG. 6 to implement examples disclosed herein.

FIG. 10 is a first example workflow that may be implemented by the ADM system of FIG. 6 to generate correlation factors for graph models captured in an index table for future use.

FIG. 11 is a second example workflow that may be implemented by the ADM system of FIG. 6 to generate correlation factors for dynamic query operation.

FIG. 12 is an illustration of a first example graph model and a second example graph model for depicting groups of related data and metadata connected linked via strength vectors.

FIG. 13 is a block diagram of example data usage monitoring circuitry to implement examples disclosed herein.

FIG. 14 is a block diagram of example data usage monitoring circuitry to implement deep data inspection examples disclosed herein.

FIG. 15 is a block diagram of example data usage monitoring circuitry to implement super node examples disclosed herein.

FIG. 16 is a block diagram of example data usage monitoring circuitry to implement examples disclosed herein.

FIG. 17 is a flowchart representative of example machine readable instructions and/or example operations that may be executed by example processor circuitry to implement the example data usage monitoring circuitry of FIG. 13 to effectuate identification and mitigation of ethical divergence.

FIG. 18 is a flowchart representative of example machine readable instructions and/or example operations that may be executed by example processor circuitry to implement the example data usage monitoring circuitry of FIG. 13 to ingest data from a data source.

FIG. 19 is a flowchart representative of example machine readable instructions and/or example operations that may be executed by example processor circuitry to implement the example data usage monitoring circuitry of FIG. 13 to orchestrate resources in an edge environment based on data.

FIG. 20 is a flowchart representative of example machine readable instructions and/or example operations that may be executed by example processor circuitry to implement the example data usage monitoring circuitry of FIG. 13 to execute a machine learning model with resources to generate outputs including at least one of a first value representative of a data stream characteristic or a second value representative of an AI application node characteristic.

FIG. 21 is a flowchart representative of example machine readable instructions and/or example operations that may be executed by example processor circuitry to implement the example sustainable storage circuitry of FIG. 13 to execute a machine learning model to generate outputs representative of a data stream characteristic and an AI application node characteristic.

FIG. 22 is a flowchart representative of example machine readable instructions and/or example operations that may be executed by example processor circuitry to implement the example data usage monitoring circuitry of FIG. 13 to determine a value representative of a data stream characteristic based on at least one of training data or ingested data.

FIG. 23 is a flowchart representative of example machine readable instructions and/or example operations that may be executed by example processor circuitry to implement the example data usage monitoring circuitry of FIG. 13 to determine a value representative of an AI application node characteristic based on at least one of training data or ingested data.

FIG. 24 is a flowchart representative of example machine readable instructions and/or example operations that may be executed by example processor circuitry to implement the example data usage monitoring circuitry of FIG. 13 to implement a super node.

FIG. 25 is a flowchart representative of example machine readable instructions and/or example operations that may be executed by example processor circuitry to implement the example data usage monitoring circuitry of FIG. 13 to implement a deep data inspection node.

FIG. 26 is a flowchart representative of example machine readable instructions and/or example operations that may be executed by example processor circuitry to implement the example data usage monitoring circuitry of FIG. 13 to generate and compare graph nodes.

FIG. 27 is a flowchart representative of example machine readable instructions and/or example operations that may be executed by example processor circuitry to implement the example data usage monitoring circuitry of FIG. 13 to cause operations at the nodes of an edge environment.

FIG. 28 is a flowchart representative of example machine readable instructions and/or example operations that may be executed by example processor circuitry to implement the example data usage monitoring circuitry of FIG. 13 to track data consumption in an edge environment.

FIG. 29 is a flowchart representative of example machine readable instructions and/or example operations that may be executed by example processor circuitry to implement the example data usage monitoring circuitry of FIG. 13 to modify a data stream or prohibit consumption of a data stream.

FIG. 30 is a block diagram of an example processing platform including processor circuitry structured to execute the example machine readable instructions and/or the example operations of FIGS. 7 and/or 17-29 to implement the example ADM system of FIG. 6.

FIG. 31 is a block diagram of an example processing platform including processor circuitry structured to execute the example machine readable instructions and/or the example operations of FIGS. 7 and/or 17-29 to implement the example data usage monitoring circuitry of FIG. 13.

FIG. 32 is a block diagram of an example implementation of the processor circuitry of FIGS. 30 and/or 31.

FIG. 33 is a block diagram of another example implementation of the processor circuitry of FIGS. 30 and/or 31.

FIG. 34 is a block diagram of an example software distribution platform (e.g., one or more servers) to distribute software (e.g., software corresponding to the example machine readable instructions of FIGS. 7 and/or 17-29) to client devices associated with end users and/or consumers (e.g., for license, sale, and/or use), retailers (e.g., for sale, re-sale, license, and/or sub-license), and/or original equipment manufacturers (OEMs) (e.g., for inclusion in products to be distributed to, for example, retailers and/or to other end users such as direct buy customers).

DETAILED DESCRIPTION

The figures are not to scale. In general, the same reference numbers will be used throughout the drawing(s) and accompanying written description to refer to the same or like parts.

As used herein, connection references (e.g., attached, coupled, connected, and joined) may include intermediate members between the elements referenced by the connection reference and/or relative movement between those elements unless otherwise indicated. As such, connection references do not necessarily infer that two elements are directly connected and/or in fixed relation to each other. As used herein, stating that any part is in “contact” with another part is defined to mean that there is no intermediate part between the two parts.

Unless specifically stated otherwise, descriptors such as “first,” “second,” “third,” etc., are used herein without imputing or otherwise indicating any meaning of priority, physical order, arrangement in a list, and/or ordering in any way, but are merely used as labels and/or arbitrary names to distinguish elements for ease of understanding the disclosed examples. In some examples, the descriptor “first” may be used to refer to an element in the detailed description, while the same element may be referred to in a claim with a different descriptor such as “second” or “third.” In such instances, it should be understood that such descriptors are used merely for identifying those elements distinctly that might, for example, otherwise share a same name.

As used herein “substantially real time” refers to occurrence in a near instantaneous manner recognizing there may be real world delays for computing time, transmission, etc. Thus, unless otherwise specified, “substantially real time” refers to real time+/−1 second. As used herein, the phrase “in communication,” including variations thereof, encompasses direct communication and/or indirect communication through one or more intermediary components, and does not require direct physical (e.g., wired) communication and/or constant communication, but rather additionally includes selective communication at periodic intervals, scheduled intervals, aperiodic intervals, and/or one-time events.

As used herein, “processor circuitry” is defined to include (i) one or more special purpose electrical circuits structured to perform specific operation(s) and including one or more semiconductor-based logic devices (e.g., electrical hardware implemented by one or more transistors), and/or (ii) one or more general purpose semiconductor-based electrical circuits programmed with instructions to perform specific operations and including one or more semiconductor-based logic devices (e.g., electrical hardware implemented by one or more transistors). Examples of processor circuitry include programmed microprocessors, Field Programmable Gate Arrays (FPGAs) that may instantiate instructions, Central Processor Units (CPUs), Graphics Processor Units (GPUs), Digital Signal Processors (DSPs), XPUs, or microcontrollers and integrated circuits such as Application Specific Integrated Circuits (ASICs). For example, an XPU may be implemented by a heterogeneous computing system including multiple types of processor circuitry (e.g., one or more FPGAs, one or more CPUs, one or more GPUs, one or more DSPs, etc., and/or a combination thereof) and application programming interface(s) (API(s)) that may assign computing task(s) to whichever one(s) of the multiple types of the processing circuitry is/are best suited to execute the computing task(s).

Network environments today have many different complex activities and processes, and the gathering of sufficient and relevant data to verify that a specific activity, tool, or task is performing as expected (or identifying a problem with the activity, tool, or task) can be intensive. As edge analytics efforts increase, the amount of data being utilized grows as well. In many situations, computer systems/devices/nodes use artificial intelligence (AI) and machine learning (ML) to process such large volumes of data. Although AI/ML provides huge value, efficiency, and productivity to data processing and analytics, few systems adequately consider the invasive nature of AI in data analytics. AI analysis of data cause a number of concerns in the areas of privacy, security, etc. Because of the way machine learning models operate, there is little traceability into how the insights from AI algorithms trigger decisions. For example, are surveillance solutions used exclusively to serve legitimate security purposes, or do they also identify and track gender, race, spoken languages and other sensitive topics? Even more problematic, AI implementations in the wrong hands can have a malicious purpose behind their monitoring activities. For example, in a healthcare environment, while AI could be used to inform people of disease tendencies in individuals based on genetic histories, environmental factors, etc., it could also be used to track human genetic traits to manipulate healthcare coverages and rates, which leads to ethical dilemmas.

Existing solutions to qualify AI operations focus primarily on accuracy and errors in perception. For example, failing to detect a red traffic light is a detection miss by an AI algorithm designed to detect such objects and states. For example, failing to trigger a functional control in a robot arm when facing unusual data for which the AI algorithm was not trained is a training issue resulting from not providing a broad enough global training data set. Such qualifications do not measure or detect usage context for good or bad purposes.

Examples disclosed herein include systems, apparatus, articles of manufacture, and methods for data usage monitoring to identify and mitigate ethical divergence in a data stream. For example, a machine learning model is implemented to monitor AI systems to determine the relevancy versus irrelevancy of AI application usage. For example, data usage and metadata usage may be tracked with a focus on the type of data being monitored and the context of the monitoring. For example, a process to determine whether AI applications that consume data are consuming it ethically is provided. Ethical data consumption means consuming data for a valid purpose. In some examples, the context of the data usage can be used to determine similar usages to other legitimate consumers of the data. Thus, examples disclosed herein describe a process to monitor data usage to discern good/ethical usage of data versus bad/unethical usage of the data. In some examples, the process to monitor data usage includes monitoring the specific data usage in question and also monitoring other nominal/baseline data usage of similar data sourced from similar or the same data source(s). In some examples, comparing a pattern of the data usage in question to nominal/baseline patterns of similar data usages may lead to a convergence of the compared patterns or a divergence of the compared patterns instantly or over time. Thus, if nominal/baseline patterns constitute ethical usages of data, a diverging data usage pattern to the nominal/baseline patterns may indicate an ethical divergence condition of the data stream, an ethical divergence condition of the data in the data stream, and/or an ethical divergence condition of the usage/consumption of the data/data stream by the data/data stream consumer (e.g., an artificial intelligence (AI) application node). In some examples, an ethical divergence condition may be also referred to as a deviation condition, an anomaly condition, a divergence condition, an anomalistic condition, an abnormality condition, a differentiation condition, or a discrepancy condition, among others. Additionally, in some examples, if the data usage is identified as bad/unethical, appropriate actions to mitigate such unethical usage are triggered. Thus, examples disclosed herein effectuate the identification and mitigation of an ethical divergence condition (e.g. a deviation condition) in a data stream.

In some examples, data is consumed by one or more nodes in an edge environment based on a stream of data (e.g., data stream) streamed from a data source (e.g., a source node of the data) to a consumer of the data (e.g., a consumer node of the data). A node means a computing device or other entity that has some amount of computer logic with capabilities to request data on a network and consume the requested data, such as a desktop computer, laptop, mobile phone, server, workstation, or embedded computer in any type of environment (e.g., autonomous vehicle, industrial system, central office, or any other type of edge device). In some examples, the consumption or attempted consumption of any such data stream may be measured for ethical usage. In some examples, in an edge environment that enables nodes to communicate with each other over one or more networks (e.g., one or more wired networks and/or one or more wireless networks, etc.), the communication may be in the form of network traffic. In some examples, such network traffic may include many concurrent data streams that are transferring data between many nodes.

In some examples, a measure of the ethical usage of a data stream (e.g., an amount of data within the data stream) can be based on at least one of a characteristic of the data stream or a characteristic of an AI application node that is consuming or attempting to consume the data stream. The term “data stream,” as used herein, means a movement of data of any type and in any amount from one location to another location across one or more networks. The example data stream may include data packets constructed in any known protocol that allows a data payload of any size to be transmitted between nodes/devices on networks. Thus, in some examples, a data stream may include one or more data packets of data. In some examples, data streams may transfer data over time (e.g., a series of data packets may be sent from a node A to a node B). In some examples, the AI application node is considered a data consumption node and includes some form of AI/ML application, running on the node, performing, instructing, or causing the attempts at consumption of the data stream (or portions thereof). The term “AI application node” herein is used to describe such a node.

In examples described herein, the data stream characteristic (e.g., to be measured for ethical usage), can be based on at least one of a content type of a data stream, a sensitive attribute of a data stream, a security level of a data stream, or a source location of a source node sourcing the data stream, among other possible characteristics. In some examples, a content type of the data stream may include classifying the content within the data stream as image data, audio data, textual data, telemetry data (e.g., sensor data, etc.), or any other type of data, or a combination of two or more types of such data. In some examples, a sensitive attribute of the data stream may include data that provides/describes sensitive topics, such as identification information of individuals (e.g., human resources data such as a home address, a social security number, etc.), health information of individuals (e.g., height, weight, body mass index, blood pressure, temperature, heart rate, blood cell counts, etc.), classification of individuals based on race, gender, or other or any other type of classification, information about financial statements, governmental records, confidential records, or one or more other sensitive attributes. In some examples, a security level of a data stream may include a level of known confidentiality based on content type, sensitive attributes, etc., such as confidential data, restricted data, top secret data, etc. In some examples, a source location of a source node sourcing the data stream may be described by an Internet Protocol (IP) address, a physical address, a room number, a building number, a floor number, an elevation, a global positioning system (GPS) set of coordinates, or another type of address that corresponds to a virtual or geographic location that may change how data is viewed. For example, a virtual source location IP address from a bank may cause heightened ethical scrutiny for a data stream with financial data. For example, a geographical source location at a military facility may cause heighted ethical scrutiny for image data. For example, a geographical source location in a recording studio may cause heighted ethical scrutiny for audio data.

In some examples, data in a data stream may allow layering characteristics, which can be utilized to filter data or modify data during a mitigation stage. For example, image data may relate to visual map data that has multiple layers (and therefore, potentially multiple filters) such as a first/highest map layer that shows only boundaries and roads, a next map layer may include buildings, a next map layer may include names of roads and buildings, a next map layer may add satellite imagery as an overlay but only with a pixel granularity that allows for detecting/discerning objects that are greater than 100 feet across, the next several map layers may tighten the focus to allow object detection at smaller granularities but certain classified objects within the visual map data may be blurred out or blacked out for privacy purposes, and a final map layer may reveal the blurred/blacked out areas. Each of these visual map layers may be associated with a filter at a node (e.g., a super node) for filtering out consumption requests or filtering in data monitoring of such data/data streams.

In some examples, the content type of a data stream, a sensitive attribute of a data stream, a security level of a data stream, a source location of a source node sourcing the data stream, etc., may be described as such characteristics within the data stream. For example, flags associated with such characteristics may be in the headers of data packets within the data stream. In some examples, data within the data stream may be tagged with metadata describing such characteristics. In some examples, data analysis logic within one or more nodes with access to the data stream may determine such data stream characteristics (e.g., characteristics regarding the content type of a data stream, a sensitive attribute of a data stream, a security level of a data stream, a source location of a data stream, etc.) by analyzing the data in the data stream (e.g., analyzing a data payload within one or more data packets in the data stream).

In examples described herein, the AI application node characteristic (e.g., to be measured for ethical usage), can be based on at least one of a service type attribute of the AI application node, a usage context of a data stream by the AI application node, or any one or more other AI application node characteristics. For example, a service type attribute may include a usage model of the AI application or service being performed by the AI application node. Examples of usage models of the AI application node include smart shelves in retail stores, monitoring high-risk intersections in cities, automated identification of individuals in airports or secure government buildings, predictive maintenance of equipment in factories, automated health screening of individuals, among a myriad of other service types of usage models. Thus, in some examples, the AI application node is classified as having one or more service types by labeling it as such with the service type attribute. In some examples, a voluntary labeling of the service type attribute is provided by the AI application node to other nodes within the edge environment as a way to declare itself and its services to peers, consumers, and administrators in the edge environment. In some examples, data analysis logic within one or more nodes with access to the AI application node may monitor the activity of the AI application node to either verify the voluntary labeling of the service type attribute is accurate or to provide an analyzed labeling of the AI application node's service type attribute if no such attribute was volunteered. In some examples, the usage context of the data stream may include characteristic information such as the monitoring of products consumption status for smart shelves in retail stores, the monitoring of the density of vehicles and pedestrians cross high-risk intersections in cities, identifying persons traveling or security records clearance for the automated identification of individuals in airports or secure government buildings, the continuous check of machines' health by a factory management system for the predictive maintenance of equipment in factories, or the fast-checking of a pre-visit medical visit for automated health screening of individuals, among other usage contexts. In some examples, the AI application node volunteers its usage context of the data stream. In some examples, data analysis logic within one or more nodes with access to the AI application node may monitor the activity of the AI application node to either verify the usage context of the data stream by the AI application node or to provide an analyzed labeling of the AI application node's usage context of the data stream if none was volunteered.

Example proactive data management and analytics systems and/or adaptive data management (ADM) techniques disclosed herein may monitor network traffic (e.g., content, context, usage frequencies, time used, etc.) across different nodes, and then highlight or otherwise identify the AI application node(s) consuming or attempting to consume data streams within the network traffic. For example, management may refer to causing a result to occur to achieve a desired goal, target, or objective. In some examples, management may be implemented by information in any form that may be ingested, processed, interpreted and/or otherwise manipulated by processor circuitry to produce a result. The produced result may itself be data. For example, management may itself be data that is representative of an action, activity, operation, etc., to be carried out at one or more nodes.

In some examples, data management (e.g., adaptive data management, data traffic management, etc.) may refer to decision making that achieves a goal, objective, or target. In some examples, the decision making can be an output (e.g., a data output, a numerical output, a dimensionless output, etc.) of a computing task or workload. For example, the output can be data that, when generated, may affect (e.g., directly affect) a node that generated the data. In some examples, the node that generated the data may be affected by being invoked to carry out and/or execute an activity, action, operation, etc., based on the generated data. For example, a node implemented by an electronic control unit (ECU) in a vehicle may generate an output representative of a decision for the vehicle to change lanes on a highway, and the output may be generated based on vehicle data (e.g., speed data, position or location data of the vehicle, position or location data of surrounding vehicle(s), etc.) associated with the vehicle. In some examples, the node implemented by the ECU may perform data management by generating the output based on ingested data (e.g., the vehicle data) to cause (e.g., directly cause) the vehicle to carry out and/or execute an operation in connection with the vehicle.

In some examples, the output can be data that, when generated, may affect (e.g., indirectly affect) a different node than the node that generated the output. For example, a first node performing data management may generate an output (e.g., a decision, a determination, etc.) based on ingested data and transmit the output to a second node to cause the second node to be affected. In some examples, the second node may be affected by being invoked to carry out and/or execute an activity, action, operation, etc., based on the received output.

In some examples, an example ADM system as disclosed herein may process data locally at the edges/nodes where data is sourced and consumed, then send the relevant data to a server, data center, etc., for further usage. In some examples, the ADM system as disclosed herein may implement data usage monitoring to identify and mitigate ethical divergence from a nominal/baseline data usage based on at least one of a characteristic of a data stream or a characteristic of an AI application node. In some examples, the ADM system may implement a “sink node” model, schema, technique, etc., where one(s) of the nodes can have a “logical sync node” consuming the data stream (e.g., obtaining, receiving, etc.) and deciding what to do with it, whether to communicate to the external world, upload somewhere, keep locally, or do not keep at all.

In some disclosed examples, the ADM system may carry out data analysis of the consumption or attempted consumption of the data stream by a “target” AI application node to prevent or mitigate unethical usage of the data within the data stream. As used herein, the “target AI application node” refers to the AI application node being monitored for ethical divergence of data usage to distinguish it from other AI application nodes in the edge environment. In some examples, the ADM system may utilize Artificial Intelligence/Machine Learning (AI/ML) modeling techniques and/or data graph techniques to map, associate, and/or otherwise correlate relevant datasets to one another. For example, in order to determine whether consumption/usage of the data in the data stream is ethical, comparisons of the target AI application node's consumption/usage (or attempts thereof) of the data stream to (known ethical) baseline/nominal consumption/usages from other AI application nodes may be implemented through the AI/ML modeling techniques and/or data graph techniques. In some disclosed examples, the ADM system may utilize AI/ML techniques to learn data stream characteristics and/or target AI application node characteristics on captured or ingested data for an observation period of the target AI application node for improved correlation determinations. Thus, datasets including known ethical consumption of similar data streams by other AI application nodes may be compared to a data set of the observed consumption or attempted consumption by the target AI application node.

In some examples, the ADM system may evaluate (e.g., continuously evaluate) data within one or more non-target data streams to establish a baseline pattern (e.g., a nominal pattern, etc.) of one or more data streams being consumed by other AI application nodes over time. In some examples, the baseline pattern (or patterns) may be used in subsequent comparisons for anomaly or deviation detection. For example, if a traffic feed camera on a busy intersection is counting pedestrians to provide a count for traffic control considerations, nominal data patterns coming from AI application nodes consuming such a video feed in an ethical manner may simply request a count of generic human bounding boxes to provide non-personal human counts. On the other hand, if the target data stream includes identification style bounding boxes that are trained to faces for facial recognition, and therefore include personal identification information, such divergence of data patterns between the nominal/baseline data pattern and the target data stream data pattern may identify an ethical divergence of the target data stream. In some examples, the ADM system may evaluate nominal data streams over time for greater accuracy in data to show tendencies, as opposed to a single instant anomaly for use in subsequent comparisons over time for further anomaly or deviation detection. As used herein, the “target data stream” refers to a specific data stream being consumed or attempted to be consumed by the target AI application node to distinguish from one or more other data streams in the edge environment. In some examples, the target data stream is one of several target data streams being analyzed as it is being consumed or attempted to be consumed by the target AI application. In some examples, the ADM system may evaluate the target data stream over time to gather a history of consumption of the target data stream for use in the subsequent comparisons to the baseline pattern to provide the anomaly or deviation detection. For example, the ADM system may utilize the baseline pattern to determine whether an event is a periodic normalcy (e.g., a cyclic event) or an aberration that requires attention.

In some disclosed examples, the ADM system may identify one or more characteristics representative of the target data stream and/or one or more characteristics representative of the target AI application node during the training/learning phase of the AI/ML approaches as disclosed herein. In some examples, the ADM system may utilize generated graph nodes representing combinations of multiple such characteristics that provide more accurate representations of baseline patterns of nominal data streams and baseline patterns of nominal AI application nodes for comparison against patterns of the target data stream and patterns of the target AI application node.

In some disclosed examples, the ADM system may assign metadata to the target data stream to cause orchestration of edge resources to monitor and potentially modify data usage/consumption of data within the target data stream by the target AI application node. In some examples, the ADM system may assemble metadata from multiple datasets into graph node model representations to allow comparison of graph model metadata to provide correlation factors for future monitoring and baseline comparison usages.

In some disclosed examples, the ADM system may mitigate risk associated with potential unethical usage/consumption of the target data stream by naming or tagging data in order to subsequently be able to use that identifier to search for the data. In some disclosed examples, the ADM system may also mitigate risk associated with potential unethical usage/consumption of the target data stream by implementing ledger techniques such as blockchain audit trails for data within the data stream. In some disclosed examples, the ADM system may also mitigate risk associated with potential unethical usage/consumption of the target data stream by disallowing the consumption of the target data stream by the target AI application node. Examples disclosed herein may monitor the relevant versus irrelevant AI/ML through monitoring the data and metadata usage. Examples disclosed herein monitor data usage to identify good, authorized, appropriate, etc., usage vs. bad, unauthorized, inappropriate, etc., usage, as in ethical versus unethical usage. Examples disclosed herein may trigger appropriate action(s) in response to a detection such unauthorized or unethical usage.

Examples disclosed herein enable monitoring of data usage and enforcing control (e.g., lawful control) to avoid irregular usage of data that can lead to unethical AI/ML. AI/ML itself is a tool to monitor data usage and imposes control (e.g., lawful control). In some examples, AI/ML-based systems use AI/ML to learn and verify the dominant features in the data used by the AI/ML algorithm during training and inference and if the features can represent any data stream characteristic, such as a sensitive attribute (e.g., person-related, gender, origin, geographic location, language, age, etc., associated with locality that can be subject to cultural norms). In some examples, AI/ML-based systems learn by counter example through observation of normal data usage to be able to identify irregular usage. In some examples, AI/ML-based systems automatically constrain or limit data provided to nodes seen to consume sensitive attributes in the data and flag the user/nodes that repeatedly access sensitive info.

Examples disclosed herein may leverage hardware, software, and/or firmware features to learn and/or mitigate irregular data usage. For example, an ADM system as disclosed herein may include a data consumption tracker that may be hardware, software, and/or firmware component(s) that track the features consumption in data (e.g., for video data track region of interest consumption, for audio data used in natural language processing track the language, etc. and compare against sensitive feature.

Examples disclosed herein may include Digital Right Management (DRM) for sensitive features in the data. For example, requiring decryption key access to access the data, or portion(s) thereof. Meticulously defined policies may need to be in place for monitoring and supervision of data usage to avoid single point of influence. Examples disclosed herein include AI/ML algorithms to learn the data content type and learn the nominal traffic on the network and node behavior (e.g., system and data access, data transfer and modification, etc.) and detect significant, periodic, unusual changes to nominal conditions and flag alerts. In some examples, the ADM system may include and/or otherwise implement an example contextual metadata/event-chain correlation manager that generates a graph node representation of the data stream based on multiple classified topics and objects detected and/or incorporates relationships/affinity to one another. In some such examples, the contextual metadata/event-chain correlation manager may enable more comprehensive comparisons of data and data streams to identify closely correlated content which may represent patterns in detection or targets to call out as non-random.

In some examples, the same graph node representations from above may be used to auto-generate initial levels of data authorization/confidentiality by comparing to other similar graph node tags for comparison. In some examples, the same graph node representations from above may also be used to autonomously update existing confidentiality flags across an entire operation's data landscape by setting a representative representation and using this to flag and change data throughout a database or network of databases.

In some examples, the ADM system may utilize a number of types of attributes to classify data and metadata. For example, sensor inputs, times and dates, classifications of visual objects in a video feed, chains of receivership, blockchaining for security, among other attributes to support a trusted “paper trail” for data use and ownership. In some examples, the ADM system may execute AI/ML algorithms to identify groups of nodes or traces of data routing which do not include human or perhaps always includes a person(s) which helps codify the ‘clues’ leading to bad outcome from the data insights.

In some examples, the ADM system may implement AI/ML algorithms in a hierarchal way to prevent single point of influence (or bad influence) that can be caused from wrong monitoring/supervision and lead to unfair decision. In some examples, the ADM system may set policies for monitoring/supervision of data through putting key indicators for relevant versus irrelevant usage and nominal versus non-nominal data and traffic in the network. In some examples, the ADM system may condition non-nominal traffic detection based on geographical policies and also environment context (e.g., looking for person's presence is common for surveillance applications but less common for factory room only having machinery). In some examples, the ADM system may utilize results that are produced by predictive analytics or anomaly detection algorithms, such as quality data of a certain factory process against individuals or group of people. For example, the ADM system may track the productivity of all individuals on a factory floor and highlight anomalies (e.g., people moving more slowly than the average movement speed).

In some examples, the ADM system may implement a plurality of super nodes to track data stream consumption across AI application nodes through sharing of data across the plurality of super nodes such as data stream patterns of nominal and non-nominal data streams. For example, one node that builds a nominal pattern alone has a data set limited by the data the node receives, but ten nodes that have ten different data sets of data streams and can share the data patterns being built may create a consensus data pattern applicable to an entire edge network or to a larger portion of the edge network than a single node would be able to access. In some examples, the ADM system may implement a plurality of deep data inspection (DDI) nodes to track data stream consumption across AI application nodes through sharing of data across the plurality of DDI nodes. In some examples, DDI nodes have access to a plurality of machine learning models, each trained with unique feature sets. In some examples, the DDI nodes may be deployed with a machine learning model that was trained on either the same features that are present in a target data stream or with a machine learning model that was trained on a feature set with the highest number/percentage of matching features to the features present in the target data stream. In some examples, the DDI nodes may communicate with each other in a similar format to the super nodes in that they may attempt to build one or more consensus data patterns that are most relevant to the target data stream(s) (e.g., the same characteristics/features are used in a comparison between a nominal data stream and the target data stream).

FIG. 1 is a block diagram 100 showing an overview of a configuration for Edge computing, which includes a layer of processing referred to in many of the following examples as an “Edge cloud”. For example, the block diagram 100 may implement an example adaptive data management system (also referred to herein as a smart data management system) as disclosed herein. As shown, the Edge cloud 110 is co-located at an Edge location, such as an access point or base station 140, a local processing hub 150, or a central office 120, and thus may include multiple entities, devices, and equipment instances. The Edge cloud 110 is located much closer to the endpoint (consumer and producer) data sources 160 (e.g., autonomous vehicles 161, user equipment 162, business and industrial equipment 163, video capture devices 164, drones 165, smart cities and building devices 166, sensors and IoT devices 167, etc.) than the cloud data center 130. Compute, memory, and storage resources which are offered at the edges in the Edge cloud 110 are critical to providing ultra-low latency response times for services and functions used by the endpoint data sources 160 as well as reduce network backhaul traffic from the Edge cloud 110 toward cloud data center 130 thus improving energy consumption and overall network usages among other benefits.

Compute, memory, and storage are scarce resources, and generally decrease depending on the Edge location (e.g., fewer processing resources being available at consumer endpoint devices, than at a base station, than at a central office). However, the closer that the Edge location is to the endpoint (e.g., user equipment (UE)), the more that space and power is often constrained. Thus, Edge computing attempts to reduce the amount of resources needed for network services, through the distribution of more resources which are located closer both geographically and in network access time. In this manner, Edge computing attempts to bring the compute resources to the workload data where appropriate, or, bring the workload data to the compute resources.

The following describes aspects of an Edge cloud architecture that covers multiple potential deployments and addresses restrictions that some network operators or service providers may have in their own infrastructures. These include, variation of configurations based on the Edge location (because edges at a base station level, for instance, may have more constrained performance and capabilities in a multi-tenant scenario); configurations based on the type of compute, memory, storage, fabric, acceleration, or like resources available to Edge locations, tiers of locations, or groups of locations; the service, security, and management and orchestration capabilities; and related objectives to achieve usability and performance of end services. These deployments may accomplish processing in network layers that may be considered as “near Edge”, “close Edge”, “local Edge”, “middle Edge”, or “far Edge” layers, depending on latency, distance, and timing characteristics.

Edge computing is a developing paradigm where computing is performed at or closer to the “Edge” of a network, typically through the use of a compute platform (e.g., x86 or ARM compute hardware architecture) implemented at base stations, gateways, network routers, or other devices which are much closer to endpoint devices producing and consuming the data. For example, Edge gateway servers may be equipped with pools of memory and storage resources to perform computation in real-time for low latency use-cases (e.g., autonomous driving or video surveillance) for connected client devices. Or as an example, base stations may be augmented with compute and acceleration resources to directly process service workloads for connected user equipment, without further communicating data via backhaul networks. Or as another example, central office network management hardware may be replaced with standardized compute hardware that performs virtualized network functions and offers compute resources for the execution of services and consumer functions for connected devices. Within Edge computing networks, there may be scenarios in services which the compute resource will be “moved” to the data, as well as scenarios in which the data will be “moved” to the compute resource. Or as an example, base station compute, acceleration and network resources can provide services in order to scale to workload demands on an as needed basis by activating dormant capacity (subscription, capacity on demand) in order to manage corner cases, emergencies or to provide longevity for deployed resources over a significantly longer implemented lifecycle.

FIG. 2 illustrates operational layers among endpoints, an Edge cloud, and cloud computing environments. For example, FIG. 2 may implement an example adaptive data management system as disclosed herein. Specifically, FIG. 2 depicts examples of computational use cases 205, utilizing the Edge cloud 110 among multiple illustrative layers of network computing. The layers begin at an endpoint (devices and things) layer 200, which accesses the Edge cloud 110 to conduct data creation, analysis, and data consumption activities. The Edge cloud 110 may span multiple network layers, such as an Edge devices layer 210 having gateways, on-premise servers, or network equipment (nodes 215) located in physically proximate Edge systems; a network access layer 220, encompassing base stations, radio processing units, network hubs, regional data centers (DC), or local network equipment (equipment 228); and any equipment, devices, or nodes located therebetween (in layer 212, not illustrated in detail). The network communications within the Edge cloud 110 and among the various layers may occur via any number of wired or wireless mediums, including via connectivity architectures and technologies not depicted.

Examples of latency, resulting from network communication distance and processing time constraints, may range from less than a millisecond (ms) when among the endpoint layer 200, under 5 ms at the Edge devices layer 210, to even between 10 to 40 ms when communicating with nodes at the network access layer 220. Beyond the Edge cloud 110 are core network 230 and cloud data center 240 layers, each with increasing latency (e.g., between 50-60 ms at the core network layer 230, to 100 or more ms at the cloud data center layer). As a result, operations at a core network data center 235 or a cloud data center 245, with latencies of at least 50 to 100 ms or more, will not be able to accomplish many time-critical functions of the use cases 205. Each of these latency values are provided for purposes of illustration and contrast; it will be understood that the use of other access network mediums and technologies may further reduce the latencies. In some examples, respective portions of the network may be categorized as “close Edge”, “local Edge”, “near Edge”, “middle Edge”, or “far Edge” layers, relative to a network source and destination. For instance, from the perspective of the core network data center 235 or a cloud data center 245, a central office or content data network may be considered as being located within a “near Edge” layer (“near” to the cloud, having high latency values when communicating with the devices and endpoints of the use cases 205), whereas an access point, base station, on-premise server, or network gateway may be considered as located within a “far Edge” layer (“far” from the cloud, having low latency values when communicating with the devices and endpoints of the use cases 205). It will be understood that other categorizations of a particular network layer as constituting a “close”, “local”, “near”, “middle”, or “far” Edge may be based on latency, distance, number of network hops, or other measurable characteristics, as measured from a source in any of the network layers 200-240.

The various use cases 205 may access resources under usage pressure from incoming streams, due to multiple services utilizing the Edge cloud. To achieve results with low latency, the services executed within the Edge cloud 110 balance varying requirements in terms of: (a) Priority (throughput or latency) and Quality of Service (QoS) (e.g., traffic for an autonomous car may have higher priority than a temperature sensor in terms of response time requirement; or, a performance sensitivity/bottleneck may exist at a compute/accelerator, memory, storage, or network resource, depending on the application); (b) Reliability and Resiliency (e.g., some input streams need to be acted upon and the traffic routed with mission-critical reliability, where as some other input streams may be tolerate an occasional failure, depending on the application); and (c) Physical constraints (e.g., power, cooling and form-factor, etc.).

The end-to-end service view for these use cases involves the concept of a service-flow and is associated with a transaction. The transaction details the overall service requirement for the entity consuming the service, as well as the associated services for the resources, workloads, workflows, and business functional and business level requirements. The services executed with the “terms” described may be managed at each layer in a way to assure real time, and runtime contractual compliance for the transaction during the lifecycle of the service. When a component in the transaction is missing its agreed to Service Level Agreement (SLA), the system as a whole (components in the transaction) may provide the ability to (1) understand the impact of the SLA violation, and (2) augment other components in the system to resume overall transaction SLA, and (3) implement steps to remediate.

Thus, with these variations and service features in mind, Edge computing within the Edge cloud 110 may provide the ability to serve and respond to multiple applications of the use cases 205 (e.g., object tracking, video surveillance, connected cars, etc.) in real-time or near real-time, and meet ultra-low latency requirements for these multiple applications. These advantages enable a whole new class of applications (e.g., Virtual Network Functions (VNFs), Function as a Service (FaaS), Edge as a Service (EaaS), standard processes, etc.), which cannot leverage conventional cloud computing due to latency or other limitations.

However, with the advantages of Edge computing comes the following caveats. The devices located at the Edge are often resource constrained and therefore there is pressure on usage of Edge resources. Typically, this is addressed through the pooling of memory and storage resources for use by multiple users (tenants) and devices. The Edge may be power and cooling constrained and therefore the power usage needs to be accounted for by the applications that are consuming the most power. There may be inherent power-performance tradeoffs in these pooled memory resources, as many of them are likely to use emerging memory technologies, where more power requires greater memory bandwidth. Likewise, improved security of hardware and root of trust trusted functions are also required, because Edge locations may be unmanned and may even need permissioned access (e.g., when housed in a third-party location). Such issues are magnified in the Edge cloud 110 in a multi-tenant, multi-owner, or multi-access setting, where services and applications are requested by many users, especially as network usage dynamically fluctuates and the composition of the multiple stakeholders, use cases, and services changes.

At a more generic level, an Edge computing system may be described to encompass any number of deployments at the previously discussed layers operating in the Edge cloud 110 (network layers 200-240), which provide coordination from client and distributed computing devices. One or more Edge gateway nodes, one or more Edge aggregation nodes, and one or more core data centers may be distributed across layers of the network to provide an implementation of the Edge computing system by or on behalf of a telecommunication service provider (“telco”, or “TSP”), internet-of-things service provider, cloud service provider (CSP), enterprise entity, or any other number of entities. Various implementations and configurations of the Edge computing system may be provided dynamically, such as when orchestrated to meet service objectives.

Consistent with the examples provided herein, a client compute node may be embodied as any type of endpoint component, device, appliance, or other thing capable of communicating as a producer or consumer of data. Further, the label “node” or “device” as used in the Edge computing system does not necessarily mean that such node or device operates in a client or agent/minion/follower role; rather, any of the nodes or devices in the Edge computing system refer to individual entities, nodes, or subsystems which include discrete or connected hardware or software configurations to facilitate or use the Edge cloud 110.

As such, the Edge cloud 110 is formed from network components and functional features operated by and within Edge gateway nodes, Edge aggregation nodes, or other Edge compute nodes among network layers 210-230. The Edge cloud 110 thus may be embodied as any type of network that provides Edge computing and/or storage resources which are proximately located to radio access network (RAN) capable endpoint devices (e.g., mobile computing devices, IoT devices, smart devices, etc.), which are discussed herein. In other words, the Edge cloud 110 may be envisioned as an “Edge” which connects the endpoint devices and traditional network access points that serve as an ingress point into service provider core networks, including mobile carrier networks (e.g., Global System for Mobile Communications (GSM) networks, Long-Term Evolution (LTE) networks, 5G/6G networks, etc.), while also providing storage and/or compute capabilities. Other types and forms of network access (e.g., Wi-Fi, long-range wireless, wired networks including optical networks, etc.) may also be utilized in place of or in combination with such 3GPP carrier networks.

The network components of the Edge cloud 110 may be servers, multi-tenant servers, appliance computing devices, and/or any other type of computing devices. For example, the Edge cloud 110 may include an appliance computing device that is a self-contained electronic device including a housing, a chassis, a case, or a shell. In some circumstances, the housing may be dimensioned for portability such that it can be carried by a human and/or shipped. Example housings may include materials that form one or more exterior surfaces that partially or fully protect contents of the appliance, in which protection may include weather protection, hazardous environment protection (e.g., electromagnetic interference (EMI), vibration, extreme temperatures, etc.), and/or enable submergibility. Example housings may include power circuitry to provide power for stationary and/or portable implementations, such as alternating current (AC) power inputs, direct current (DC) power inputs, AC/DC converter(s), DC/AC converter(s), DC/DC converter(s), power regulators, transformers, charging circuitry, batteries, wired inputs, and/or wireless power inputs. Example housings and/or surfaces thereof may include or connect to mounting hardware to enable attachment to structures such as buildings, telecommunication structures (e.g., poles, antenna structures, etc.), and/or racks (e.g., server racks, blade mounts, etc.). Example housings and/or surfaces thereof may support one or more sensors (e.g., temperature sensors, vibration sensors, light sensors, acoustic sensors, capacitive sensors, proximity sensors, infrared or other visual thermal sensors, etc.). One or more such sensors may be contained in, carried by, or otherwise embedded in the surface and/or mounted to the surface of the appliance. Example housings and/or surfaces thereof may support mechanical connectivity, such as propulsion hardware (e.g., wheels, rotors such as propellers, etc.) and/or articulating hardware (e.g., robot arms, pivotable appendages, etc.). In some circumstances, the sensors may include any type of input devices such as user interface hardware (e.g., buttons, switches, dials, sliders, microphones, etc.). In some circumstances, example housings include output devices contained in, carried by, embedded therein and/or attached thereto. Output devices may include displays, touchscreens, lights, light-emitting diodes (LEDs), speakers, input/output (I/O) ports (e.g., universal serial bus (USB)), etc. In some circumstances, Edge devices are devices presented in the network for a specific purpose (e.g., a traffic light), but may have processing and/or other capacities that may be utilized for other purposes. Such Edge devices may be independent from other networked devices and may be provided with a housing having a form factor suitable for its primary purpose; yet be available for other compute tasks that do not interfere with its primary task. Edge devices include Internet of Things devices. The appliance computing device may include hardware and software components to manage local issues such as device temperature, vibration, resource utilization, updates, power issues, physical and network security, etc. Example hardware for implementing an appliance computing device is described in conjunction with FIG. J1. The Edge cloud 110 may also include one or more servers and/or one or more multi-tenant servers. Such a server may include an operating system and implement a virtual computing environment. A virtual computing environment may include a hypervisor managing (e.g., spawning, deploying, commissioning, destroying, decommissioning, etc.) one or more virtual machines, one or more containers, etc. Such virtual computing environments provide an execution environment in which one or more applications and/or other software, code, or scripts may execute while being isolated from one or more other applications, software, code, or scripts.

In FIG. 3, various client endpoints 310 (in the form of mobile devices, computers, autonomous vehicles, business computing equipment, industrial processing equipment) exchange requests and responses that are specific to the type of endpoint network aggregation. For example, FIG. 3 may implement an example adaptive data management system as disclosed herein. For instance, client endpoints 310 may obtain network access via a wired broadband network, by exchanging requests and responses 322 through an on-premise network system 332. Some client endpoints 310, such as mobile computing devices, may obtain network access via a wireless broadband network, by exchanging requests and responses 324 through an access point (e.g., a cellular network tower) 334. Some client endpoints 310, such as autonomous vehicles may obtain network access for requests and responses 326 via a wireless vehicular network through a street-located network system 336. However, regardless of the type of network access, the TSP may deploy aggregation points 342, 344 within the Edge cloud 110 to aggregate traffic and requests. Thus, within the Edge cloud 110, the TSP may deploy various compute and storage resources, such as at Edge aggregation nodes 340, to provide requested content. The Edge aggregation nodes 340 and other systems of the Edge cloud 110 are connected to a cloud or data center 360, which uses a backhaul network 350 to fulfill higher-latency requests from a cloud/data center for websites, applications, database servers, etc. Additional or consolidated instances of the Edge aggregation nodes 340 and the aggregation points 342, 344, including those deployed on a single server framework, may also be present within the Edge cloud 110 or other areas of the TSP infrastructure.

FIG. 4 illustrates deployment and orchestration for virtualized and container-based Edge configurations across an Edge computing system operated among multiple Edge nodes and multiple tenants (e.g., users, providers) which use such Edge nodes. For example, FIG. 4 may implement an example adaptive data management system as disclosed herein. Specifically, FIG. 4 depicts coordination of a first Edge node 422 and a second Edge node 424 in an Edge computing system 400, to fulfill requests and responses for various client endpoints 410 (e.g., smart cities/building systems, mobile devices, computing devices, business/logistics systems, industrial systems, etc.), which access various virtual Edge instances. Here, the virtual Edge instances 432, 434 provide Edge compute capabilities and processing in an Edge cloud, with access to a cloud/data center 440 for higher-latency requests for websites, applications, database servers, etc. However, the Edge cloud enables coordination of processing among multiple Edge nodes for multiple tenants or entities.

In the example of FIG. 4, these virtual Edge instances include: a first virtual Edge 432, offered to a first tenant (Tenant 1), which offers a first combination of Edge storage, computing, and services; and a second virtual Edge 434, offered to a second tenant (Tenant 2), which offers a second combination of Edge storage, computing, and services. The virtual Edge instances 432, 434 are distributed among the Edge nodes 422, 424, and may include scenarios in which a request and response are fulfilled from the same or different Edge nodes. The configuration of the Edge nodes 422, 424 to operate in a distributed yet coordinated fashion occurs based on Edge provisioning functions 450. The functionality of the Edge nodes 422, 424 to provide coordinated operation for applications and services, among multiple tenants, occurs based on orchestration functions 460.

It should be understood that some of the devices in 410 are multi-tenant devices where Tenant 1 may function within a tenant1 ‘slice’ while a Tenant 2 may function within a tenant2 slice (and, in further examples, additional or sub-tenants may exist; and each tenant may even be specifically entitled and transactionally tied to a specific set of features all the way day to specific hardware features). A trusted multi-tenant device may further contain a tenant specific cryptographic key such that the combination of key and slice may be considered a “root of trust” (RoT) or tenant specific RoT. A RoT may further be computed dynamically composed using a DICE (Device Identity Composition Engine) architecture such that a single DICE hardware building block may be used to construct layered trusted computing base contexts for layering of device capabilities (such as a Field Programmable Gate Array (FPGA)). The RoT may further be used for a trusted computing context to enable a “fan-out” that is useful for supporting multi-tenancy. Within a multi-tenant environment, the respective Edge nodes 422, 424 may operate as security feature enforcement points for local resources allocated to multiple tenants per node. Additionally, tenant runtime and application execution (e.g., in instances 432, 434) may serve as an enforcement point for a security feature that creates a virtual Edge abstraction of resources spanning potentially multiple physical hosting platforms. Finally, the orchestration functions 460 at an orchestration entity may operate as a security feature enforcement point for marshalling resources along tenant boundaries.

Edge computing nodes may partition resources (memory, central processing unit (CPU), graphics processing unit (GPU), interrupt controller, input/output (I/O) controller, memory controller, bus controller, etc.) where respective partitionings may contain a RoT capability and where fan-out and layering according to a DICE model may further be applied to Edge Nodes. Cloud computing nodes often use containers, FaaS engines, servlets, servers, or other computation abstraction that may be partitioned according to a DICE layering and fan-out structure to support a RoT context for each. Accordingly, the respective RoTs spanning devices 410, 422, and 440 may coordinate the establishment of a distributed trusted computing base (DTCB) such that a tenant-specific virtual trusted secure channel linking all elements end to end can be established.

Further, it will be understood that a container may have data or workload specific keys protecting its content from a previous Edge node. As part of migration of a container, a pod controller at a source Edge node may obtain a migration key from a target Edge node pod controller where the migration key is used to wrap the container-specific keys. When the container/pod is migrated to the target Edge node, the unwrapping key is exposed to the pod controller that then decrypts the wrapped keys. The keys may now be used to perform operations on container specific data. The migration functions may be gated by properly attested Edge nodes and pod managers (as described above).

In further examples, an Edge computing system is extended to provide for orchestration of multiple applications through the use of containers (a contained, deployable unit of software that provides code and needed dependencies) in a multi-owner, multi-tenant environment. A multi-tenant orchestrator may be used to perform key management, trust anchor management, and other security functions related to the provisioning and lifecycle of the trusted ‘slice’ concept in FIG. 4. For instance, an Edge computing system may be configured to fulfill requests and responses for various client endpoints from multiple virtual Edge instances (and, from a cloud or remote data center). The use of these virtual Edge instances may support multiple tenants and multiple applications (e.g., augmented reality (AR)/virtual reality (VR), enterprise applications, content delivery, gaming, compute offload, etc.) simultaneously. Further, there may be multiple types of applications within the virtual Edge instances (e.g., normal applications; latency sensitive applications; latency-critical applications; user plane applications; networking applications; etc.). The virtual Edge instances may also be spanned across systems of multiple owners at different geographic locations (or, respective computing systems and resources which are co-owned or co-managed by multiple owners).

For instance, each Edge node 422, 424 may implement the use of containers, such as with the use of a container “pod” 426, 428 providing a group of one or more containers. In a setting that uses one or more container pods, a pod controller or orchestrator is responsible for local control and orchestration of the containers in the pod. Various Edge node resources (e.g., storage, compute, services, depicted with hexagons) provided for the respective Edge slices 432, 434 are partitioned according to the needs of each container.

With the use of container pods, a pod controller oversees the partitioning and allocation of containers and resources. The pod controller receives instructions from an orchestrator (e.g., orchestrator 460) that instructs the controller on how best to partition physical resources and for what duration, such as by receiving key performance indicator (KPI) targets based on SLA contracts. The pod controller determines which container requires which resources and for how long in order to complete the workload and satisfy the SLA. The pod controller also manages container lifecycle operations such as: creating the container, provisioning it with resources and applications, coordinating intermediate results between multiple containers working on a distributed application together, dismantling containers when workload completes, and the like. Additionally, the pod controller may serve a security role that prevents assignment of resources until the right tenant authenticates or prevents provisioning of data or a workload to a container until an attestation result is satisfied.

Also, with the use of container pods, tenant boundaries can still exist but in the context of each pod of containers. If each tenant specific pod has a tenant specific pod controller, there will be a shared pod controller that consolidates resource allocation requests to avoid typical resource starvation situations. Further controls may be provided to ensure attestation and trustworthiness of the pod and pod controller. For instance, the orchestrator 460 may provision an attestation verification policy to local pod controllers that perform attestation verification. If an attestation satisfies a policy for a first tenant pod controller but not a second tenant pod controller, then the second pod could be migrated to a different Edge node that does satisfy it. Alternatively, the first pod may be allowed to execute and a different shared pod controller is installed and invoked prior to the second pod executing.

FIG. 5 depicts an example of an infrastructure processing unit (IPU) 500. In some examples, the IPU 500 can effectuate and/or otherwise facilitate proactive and/or adaptive data management and analytics as described herein. Different examples of IPUs disclosed herein enable improved performance, management, security and coordination functions between entities (e.g., cloud service providers), and enable infrastructure offload and/or communications coordination functions. As disclosed in further detail below, IPUs may be integrated with smart NICs and storage or memory (e.g., on a same die, system on chip (SoC), or connected dies) that are located at on-premises systems, base stations, gateways, neighborhood central offices, and so forth. Different examples of one or more IPUs disclosed herein can perform an application including any number of microservices, where each microservice runs in its own process and communicates using protocols (e.g., an HTTP resource API, message service or gRPC). Microservices can be independently deployed using centralized management of these services. A management system may be written in different programming languages and use different data storage technologies.

Furthermore, one or more IPUs can execute platform management, networking stack processing operations, security (crypto) operations, storage software, identity and key management, telemetry, logging, monitoring and service mesh (e.g., control how different microservices communicate with one another). The IPU can access an xPU to offload performance of various tasks. For instance, an IPU exposes XPU, storage, memory, and CPU resources and capabilities as a service that can be accessed by other microservices for function composition. This can improve performance and reduce data movement and latency. An IPU can perform capabilities such as those of a router, load balancer, firewall, TCP/reliable transport, a service mesh (e.g., proxy or API gateway), security, data-transformation, authentication, quality of service (QoS), security, telemetry measurement, event logging, initiating and managing data flows, data placement, or job scheduling of resources on an XPU, storage, memory, or CPU.

In the illustrated example of FIG. 5, the IPU 500 includes or otherwise accesses secure resource managing circuitry 502, network interface controller (NIC) circuitry 504, security and root of trust circuitry 506, resource composition circuitry 508, time stamp managing circuitry 510, memory and storage 512, processing circuitry 514, accelerator circuitry 516, and/or translator circuitry 518. Any number and/or combination of other structure(s) can be used such as but not limited to compression and encryption circuitry 520, memory management and translation unit circuitry 522, compute fabric data switching circuitry 524, security policy enforcing circuitry 526, device virtualizing circuitry 528, telemetry, tracing, logging and monitoring circuitry 530, quality of service circuitry 532, searching circuitry 534, network functioning circuitry (e.g., routing, firewall, load balancing, network address translating (NAT), etc.) 536, reliable transporting, ordering, retransmission, congestion controlling circuitry 538, and high availability, fault handling and migration circuitry 540 shown in FIG. 5. Different examples can use one or more structures (components) of the example IPU 500 together or separately. For example, compression and encryption circuitry 520 can be used as a separate service or chained as part of a data flow with vSwitch and packet encryption.

In some examples, the IPU 500 includes a field programmable gate array (FPGA) 570 structured to receive commands from an CPU, XPU, or application via an API and perform commands/tasks on behalf of the CPU, including workload management and offload or accelerator operations. The illustrated example of FIG. 5 may include any number of FPGAs configured and/or otherwise structured to perform any operations of any IPU described herein.

Example compute fabric circuitry 550 provides connectivity to a local host or device (e.g., server or device (e.g., xPU, memory, or storage device)). Connectivity with a local host or device or smartNIC or another IPU is, in some examples, provided using one or more of peripheral component interconnect express (PCIe), ARM AXI, Intel® QuickPath Interconnect (QPI), Intel® Ultra Path Interconnect (UPI), Intel® On-Chip System Fabric (IOSF), Omnipath, Ethernet, Compute Express Link (CXL), HyperTransport, NVLink, Advanced Microcontroller Bus Architecture (AMBA) interconnect, OpenCAPI, Gen-Z, CCIX, Infinity Fabric (IF), and so forth. Different examples of the host connectivity provide symmetric memory and caching to enable equal peering between CPU, XPU, and IPU (e.g., via CXL.cache and CXL.mem).

Example media interfacing circuitry 560 provides connectivity to a remote smartNIC or another IPU or service via a network medium or fabric. This can be provided over any type of network media (e.g., wired or wireless) and using any protocol (e.g., Ethernet, InfiniBand, Fiber channel, ATM, to name a few).

In some examples, instead of the server/CPU being the primary component managing IPU 500, IPU 500 is a root of a system (e.g., rack of servers or data center) and manages compute resources (e.g., CPU, xPU, storage, memory, other IPUs, and so forth) in the IPU 500 and outside of the IPU 500. Different operations of an IPU are described below.

In some examples, the IPU 500 performs orchestration to decide which hardware or software is to execute a workload based on available resources (e.g., services and devices) and considers service level agreements and latencies, to determine whether resources (e.g., CPU, xPU, storage, memory, etc.) are to be allocated from the local host or from a remote host or pooled resource. In examples when the IPU 500 is selected to perform a workload, secure resource managing circuitry 502 offloads work to a CPU, xPU, or other device and the IPU 500 accelerates connectivity of distributed runtimes, reduce latency, CPU and increases reliability.

In some examples, secure resource managing circuitry 502 runs a service mesh to decide what resource is to execute workload, and provide for L7 (application layer) and remote procedure call (RPC) traffic to bypass kernel altogether so that a user space application can communicate directly with the example IPU 500 (e.g., the IPU 500 and application can share a memory space). In some examples, a service mesh is a configurable, low-latency infrastructure layer designed to handle communication among application microservices using application programming interfaces (APIs) (e.g., over remote procedure calls (RPCs)). The example service mesh provides fast, reliable, and secure communication among containerized or virtualized application infrastructure services. The service mesh can provide critical capabilities including, but not limited to service discovery, load balancing, encryption, observability, traceability, authentication and authorization, and support for the circuit breaker pattern.

In some examples, infrastructure services include a composite node created by an IPU at or after a workload from an application is received. In some cases, the composite node includes access to hardware devices, software using APIs, RPCs, gRPCs, or communications protocols with instructions such as, but not limited, to iSCSI, NVMe-oF, or CXL.

In some cases, the example IPU 500 dynamically selects itself to run a given workload (e.g., microservice) within a composable infrastructure including an IPU, xPU, CPU, storage, memory, and other devices in a node.

In some examples, communications transit through media interfacing circuitry 560 of the example IPU 500 through a NIC/smartNIC (for cross node communications) or loopback back to a local service on the same host. Communications through the example media interfacing circuitry 560 of the example IPU 500 to another IPU can then use shared memory support transport between xPUs switched through the local IPUs. Use of IPU-to-IPU communication can reduce latency and jitter through ingress scheduling of messages and work processing based on service level objective (SLO).

For example, for a request to a database application that requires a response, the example IPU 500 prioritizes its processing to minimize the stalling of the requesting application. In some examples, the IPU 500 schedules the prioritized message request issuing the event to execute a SQL query database and the example IPU constructs microservices that issue SQL queries and the queries are sent to the appropriate devices or services.

FIG. 6 is a block diagram of an example adaptive data management (ADM) system 600 to implement adaptive data management in a network environment. The ADM system 600 of FIG. 6 may be instantiated by processor circuitry such as a central processing unit executing instructions. Additionally or alternatively, the ADM system 600 of FIG. 6 may be instantiated by an ASIC or an FPGA structured to perform operations corresponding to the instructions.

In the illustrated example of FIG. 6, the ADM system 600 includes an example ADM console 602, example data sources 604, and an example data ingestion manager 606, which includes an example pre-processing manager 608. The ADM system 600 includes an example data query manager 610, which includes an example data query handler 612, an example query cache cluster manager 614, and an example metadata cluster manager 616. The ADM system 600 includes an example data publishing manager 618, which includes an example scheduler 620. The ADM system 600 includes an example node manager 622, which includes an example preferred nodes table 624. The ADM system 600 includes an example network plane 626, which includes an example data plane 628 and an example control plane 630. The ADM system 600 includes an example data security manager 632, an example algorithm manager/recommender (AMR) 634, and an example analytics manager 636, which includes example algorithms 638 (identified by Algo1, Algo2, Algo3), and an example metadata/data enrichment manager 640. The ADM system 600 includes an example resource manager 642 and an example distributed datastore 644, which includes an example metadata storage 646, and an example raw datastore 648.

In the illustrated example, the ADM system 600 includes the ADM console 102 to setup and/or otherwise configure portion(s) of the ADM system 600 (e.g., the data ingestion manager 606, the node manager 622, etc.). For example, the ADM console 602 may configure the metadata/data enrichment manager 640. In some examples, the ADM console 602 may implement metadata tagging (e.g., add, remove, and/or modify metadata). In some examples, the ADM console 602 may implement security policies (e.g., add, remove, and/or modify access policies of the data security manager 632. In some examples, the ADM console 602 may implement data management settings (e.g., locality, expiration date, etc., of data). In some examples, the ADM console 602 may be implemented by one or more user experience (UX) and/or user interface (UI) consoles.

In the illustrated example, the ADM system 600 includes the data ingestion manager 606 to ingest, receive, and/or otherwise obtain data from one(s) of the data sources 604. For example, the data sources 604 may be implemented by any hardware, software, and/or firmware as described herein (e.g., hardware, software, and/or firmware of an autonomous guided vehicle (AGV), a server, an IoT device, etc.). In some examples, the data ingestion manager 606 includes the pre-processing manager 608 to pre-process data obtained from the data sources 604.

In the illustrated example, the ADM system 600 includes the data query manager 610 to queue and/or and process data search requests from users and/or applications. For example, the data query handler 612 may queue and/or process the data search requests. In some such examples, the data query handler 612 may return results associated with the data search results to the requester. In some examples, the data query manager 610 utilizes the existence of metadata tables extracted from data files (e.g., media, alpha-numeric, spatial, etc.) that have been pre-generated by the example metadata/data enrichment manager 640. In some examples, the data query manager 610 may be implemented to do a search and match of topical terms to metadata tags or use weighted topics and phrases with Boolean operations to perform complex contextual matches, prioritizations, and sequence of topics mapping.

In some examples, the data query manager 610 manages multiple metadata context resulting from metadata generating engines, sub-metadata tables specific to user/applications with unique context, permissions, etc. In some examples, the data query manager 610 may scan a metadata file for primary search and recommendation of most appropriate data file links. For example, the data query manager 610 may include the metadata cluster manager 616 to scan the metadata file and/or return a number of excerpts (e.g., a user selectable number of excerpts) for final selection. In some examples, the data query manager 610 may check selections for permission level appropriateness. For example, different departments, regions, etc., of an environment may have security and access control. In some examples, the data query manager 610 may link a user/application to a returned and/or otherwise identified source file. In some examples, the data query manager 610 and a metadata database (e.g., the metadata storage 646) need not be co-resident.

In some examples, the data query manager 610 may include the query cache cluster manager 614 to execute selective caching. For example, the query cache cluster manager 614 may activate and/or otherwise enable caching for frequent topics, most recently used search terms with user selected and preferred source file links, file linkages that have a high correlation to one another (e.g., occurs frequently), etc., and/or combination(s) thereof.

In some examples, the data query manager 610 implements capacity scaling for demand volume and to serve local teams. For example, the data query manager 610 may launch additional instances of the data query manager 610 near and/or otherwise proximate to demand sources (e.g., department server or individual personal computer) that may be associated with the data sources 604.

Advantageously, in some examples, a locality of the metadata storage 646 to the data query manager 610 may reduce network traffic and latency to ensure that even if a file is unavailable, the existence of the file may be confirmed. In some examples, the data query manager 610 may effectuate synchronization with other managers of the ADM system 600 more frequently for metadata (e.g., metadata tables) of the metadata storage 646 that is/are accessed most frequently or having significant changes made (e.g., usually another feature of frequent use or recent capture). In some examples, the data query manager 610 may effectuate interactive and/or programmatic access to portion(s) of the ADM system 600.

In the illustrated example, the ADM system 600 includes the data publishing manager 618 to implement publish-subscribe messaging. For example, a subscriber (e.g., a data subscriber, a device subscriber, etc.) may coordinate with the scheduler 620 to subscribe to changes, updates, etc., of data of the metadata storage 646, the raw datastore 648, and/or one(s) of the data sources 604. In some such examples, the data publishing manager 618 may publish data of interest to the appropriate subscribers.

In the illustrated example, the ADM system 600 includes the node manager 622 to enable edge nodes to maintain lists (e.g., a friends list), a neighboring nodes list, a trusted or verified node list, etc.) on the network. In some examples, the list(s) may include(s) the preferred nodes table 624. For example, the preferred nodes table 624 may be implemented by a routing table in networking examples. In some examples, the node manager 622 may maintain the table/list/index as dynamic and/or evolving table/list/index by considering previous interactions and/or transactions between neighboring nodes. For example, the node manager 622 may control the table/list/index to rate neighboring nodes based on the context of data requested, frequency of data requested, Quality-of-Service (QoS) of past interactions, etc., and/or combination(s) thereof. In some examples, the table/list/index may exist in the distributed datastore 644, which may be quickly accessible upon a request from the data query manager 610.

In the illustrated example, the ADM system 600 includes the network plane 626 to facilitate the transmission, distribution, and/or otherwise propagation of data. In some examples, the network plane 626 may be implemented by one or more networks. For example, the network plane 626 may be implemented by the Internet. However, the network plane 626 may be implemented using any suitable wired and/or wireless network(s) including, for example, one or more data buses, one or more Local Area Networks (LANs), one or more wireless LANs, one or more cellular networks, one or more private networks, one or more public networks, one or more fiber networks, one or more satellite networks, etc., and/or combination(s) thereof. In the illustrated example, component(s) of the ADM system 600 may provide, deliver, propagate, etc., data in the ADM system 100 by the data plane 628. In the illustrated example, component(s) of the ADM system 600 may provide, deliver, propagate, etc., controls, commands, directions, instructions, etc., in the ADM system 100 by the control plane 630.

In the illustrated example, the ADM system 600 includes the data security manager 632 to control (e.g., add, remove, modify, etc.) one(s) of the access policies. For example, the data security manager 632 may control the access policies rather than the node/platform level security components or user authentication to the network. In some examples, the data security manager 632 may be accessed through the ADM console 102.

In some examples, the data security manager 632 assigns initial security/access levels to data files/streams based on user provided policy or explicit settings. In some examples, the data security manager 632 facilitates autonomous control of access policies, where content may inherit security levels from other similar files based on metadata/topics. In some examples, the data security manager 632 ensures compatible security level data match for the user/application/service level security with appropriate data levels. In some examples, the data security manager 632 defines the scope of data availability (e.g., geographic, topical, personnel, security level, etc.). In some examples, the data security manager 632 logs audits for a query log (e.g., request, copies, moves, success/fail, reason, etc.) maintained by the data query manager 610. In some examples, the data security manager 632 ensures data/metadata from high security areas are not copied/moved to lower-level security environments. In some examples, the data security manager 632 enforces top secret level access to confidential areas or non-top secret level access to unsecured servers. In some examples, the data security manager 632 monitors with the data query manager 610 the data request traffic for potential irregularities. In some examples, the data security manager 632 may implement and/or otherwise provide encryption services, keys as a service, etc., and/or combination(s) thereof.

In the illustrated example, the ADM system 600 includes the AMR 634 to monitor the data on the data plane 628 and/or in the distributed datastore 644 (e.g., when triggered to run analytics) and/or upon a gap in the existing algorithms 638 of the analytics manager 636. In some examples, the AMR 634 may interface with the resource manager 642 by an example interface 650. For example, the resource manager 642 may implement an orchestration agent that receives a request for new one(s) of the algorithms 638 to act on of data of interest. For example, if a node that was previously monitoring a video stream and now has some additional time series data, the AMR 634 can request the resource manager/orchestrator agent 642 for a new one of the algorithms 638 to update the analytics manager 636 to run insights on both modalities.

In some examples, the analytics manager 636 includes a metadata agent (e.g., a metadata agent that may be implemented by the metadata/data enrichment manager 640) that may request analytics to be executed by the AMR 634 for generating metadata from source streams/files. In some such examples, the AMR 634 may invoke one(s) of the algorithms 638 to generate the analytics. In some such examples, the analytics may be generated by an Artificial Intelligence/Machine Learning (M/ML) model (e.g., a neural network (NN)) for classification, neural natural language processing (NNLP) to parse documentation, etc., and/or combination(s) thereof.

In some examples, in the data or user domain, an application or a direct user request may request an analytics container (e.g., an analytics container that may be implemented by the analytics manager 636) with an appropriate configuration and optimizations for the requested task. In some examples, certain one(s) of the algorithms 638 or AI/ML models may be preferred over time as a function of accuracy and performance.

In some examples, the resource manager/orchestration agent 642 may orchestrate new one(s) of the algorithms 638 either from a centralized algorithm library based on example algorithm ratings, scores, etc., if available, or through a pointer to a local datastore (e.g., if available in a library) of the analytics manager 636 for faster access or the distributed datastore 644.

Artificial intelligence (AI), including machine learning (ML), deep learning (DL), and/or other artificial machine-driven logic, enables machines (e.g., computers, logic circuits, etc.) to use a model to process input data to generate an output based on patterns and/or associations previously learned by the model via a training process. For instance, one(s) of the algorithms 638 may be trained with data to recognize patterns and/or associations and follow such patterns and/or associations when processing input data such that other input(s) result in output(s) consistent with the recognized patterns and/or associations.

Many different types of machine-learning models and/or machine-learning architectures exist. In some examples, the analytics manager 636 may generate one(s) of the algorithms 638 as neural network model(s). In some examples, the resource manager/orchestration agent 642 may obtain and/or generate one(s) of the algorithms 638. Using a neural network model enables the analytics manager 636 to execute AI/ML workload(s). In general, machine-learning models/architectures that are suitable to use in the example approaches disclosed herein include recurrent neural networks. However, other types of machine learning models could additionally or alternatively be used such as supervised learning ANN models, clustering models, classification models, etc., and/or a combination thereof. Example supervised learning ANN models may include two-layer (2-layer) radial basis neural networks (RBN), learning vector quantization (LVQ) classification neural networks, etc. Example clustering models may include k-means clustering, hierarchical clustering, mean shift clustering, density-based clustering, etc. Example classification models may include logistic regression, support-vector machine or network, Naive Bayes, etc. In some examples, the analytics manager 636 may compile and/or otherwise generate one(s) of the algorithm(s) 638 as lightweight machine-learning models.

In general, implementing an ML/AI system involves two phases, a learning/training phase and an inference phase. In the learning/training phase, a training algorithm is used to train the one(s) of the algorithms 638 to operate in accordance with patterns and/or associations based on, for example, training data. In general, the one(s) of the algorithms 638 include(s) internal parameters (e.g., indices, raw data, metadata, insights, models, weights, etc.) that guide how input data is transformed into output data, such as through a series of nodes and connections within the one(s) of the algorithms 638 to transform input data into output data. Additionally, hyperparameters are used as part of the training process to control how the learning is performed (e.g., a learning rate, a number of layers to be used in the machine learning model, etc.). Hyperparameters are defined to be training parameters that are determined prior to initiating the training process.

Different types of training may be performed based on the type of ML/AI model and/or the expected output. For example, the analytics manager 636 may invoke supervised training to use inputs and corresponding expected (e.g., labeled) outputs to select parameters (e.g., by iterating over combinations of select parameters) for the one(s) of the algorithms 638 that reduce model error. As used herein, “labeling” refers to an expected output of the machine learning model (e.g., a classification, an expected output value, etc.). Alternatively, the analytics manager 636 may invoke unsupervised training (e.g., used in deep learning, a subset of machine learning, etc.) that involves inferring patterns from inputs to select parameters for the one(s) of the algorithms 638 (e.g., without the benefit of expected (e.g., labeled) outputs).

In some examples, the analytics manager 636 trains the one(s) of the algorithms 638 using unsupervised clustering of operating observables. For example, the operating observables may include data from the data sources 604, metadata in the metadata storage 646, data in the raw datastore 648, etc., and/or combination(s) thereof. However, the analytics manager 636 may additionally or alternatively use any other training algorithm such as stochastic gradient descent, Simulated Annealing, Particle Swarm Optimization, Evolution Algorithms, Genetic Algorithms, Nonlinear Conjugate Gradient, etc.

In some examples, the analytics manager 636 may train the one(s) of the algorithms 638 until the level of error is no longer reducing. In some examples, the analytics manager 636 may train the one(s) of the algorithms 638 locally on the analytics manager 636 and/or remotely at an external computing system (e.g., on external computing device(s) in communication with the resource manager/orchestration agent 642) communicatively coupled to the analytics manager 636. In some examples, the analytics manager 636 trains the one(s) of the algorithms 638 using hyperparameters that control how the learning is performed (e.g., a learning rate, a number of layers to be used in the machine learning model, etc.). In some examples, the analytics manager 636 may use hyperparameters that control model performance and training speed such as the learning rate and regularization parameter(s). The analytics manager 636 may select such hyperparameters by, for example, trial and error to reach an optimal model performance. In some examples, the analytics manager 636 Bayesian hyperparameter optimization to determine an optimal and/or otherwise improved or more efficient network architecture to avoid model overfitting and improve the overall applicability of the one(s) of the algorithms 638. Alternatively, the analytics manager 636 may use any other type of optimization. In some examples, the analytics manager 636 may perform re-training. The analytics manager 636 may execute such re-training in response to override(s) by a user of the ADM 600, a receipt of new training data, change(s) to node(s), change(s) observed and/or otherwise identified by node(s), etc.

In some examples, the analytics manager 636 facilitates the training of the one(s) of the algorithms 638 using training data. In some examples, the analytics manager 636 utilizes training data that originates from locally generated data, such as one(s) of the data from the data sources 604, metadata in the metadata storage 646, data in the raw datastore 648, etc., and/or combination(s) thereof. In some examples, the analytics manager 636 utilizes training data that originates from externally generated data, such as data from the data sources 604, data from the resource manager/orchestration agent 642, etc., and/or combination(s) thereof. In some examples where supervised training is used, the analytics manager 636 may label the training data (e.g., label training data or portion(s) thereof with appropriate metadata). Labeling is applied to the training data by a user manually or by an automated data pre-processing system. In some examples, the analytics manager 636 may pre-process the training data. In some examples, the analytics manager 636 sub-divides the training data into a first portion of data for training the one(s) of the algorithms 638, and a second portion of data for validating the one(s) of the algorithms 638.

Once training is complete, the analytics manager 636 may deploy the one(s) of the algorithms 638 for use as an executable construct that processes an input and provides an output based on the network of nodes and connections defined in the one(s) of the algorithms 638. The analytics manager 636 may store the one(s) of the algorithms 638 in the analytics manager 636. In some examples, the analytics manager 636 may invoke the interface 650 to transmit the one(s) of the algorithms 638 to one(s) of the external computing systems in communication with the resource manager/orchestration agent 642. In some such examples, in response to transmitting the one(s) of the algorithms 638 to the one(s) of the external computing systems, the one(s) of the external computing systems may execute the one(s) of the algorithms 638 to execute AI/ML workloads with at least one of improved efficiency or performance.

Once trained, the deployed one(s) of the algorithms 638 may be operated in an inference phase to process data. In the inference phase, data to be analyzed (e.g., live data) is input to the one(s) of the algorithms 638, and the one(s) of the algorithms 638 execute(s) to create an output. This inference phase can be thought of as the AI “thinking” to generate the output based on what it learned from the training (e.g., by executing the one(s) of the algorithms 638 to apply the learned patterns and/or associations to the live data). In some examples, input data undergoes pre-processing before being used as an input to the one(s) of the algorithms 638. Moreover, in some examples, the output data may undergo post-processing after it is generated by the one(s) of the algorithms 638 to transform the output into a useful result (e.g., a display of data, a detection and/or identification of an object, an instruction to be executed by a machine, etc.).

In some examples, output of the deployed one(s) of the algorithms 638 may be captured and provided as feedback. By analyzing the feedback, an accuracy of the deployed one(s) of the algorithm(s) 638 can be determined. If the feedback indicates that the accuracy of the deployed model is less than a threshold or other criterion, training of an updated model can be triggered using the feedback and an updated training data set, hyperparameters, etc., to generate an updated, deployed model.

In the illustrated example, the ADM system 600 includes the metadata/data enrichment manager 640 to schedule and/or execute metadata creation and/or post-processing routines intended to extract context and meaning from the source stream/files to enhance source files to decrease noise and/or clarify/focus subjects of interest. In some examples, the metadata/data enrichment manager 640 includes a metadata or enhancement request routine, an online metadata agent, and/or an offline metadata agent. In some examples, the metadata or enhancement request routine may be configured and/or otherwise generated to take inputs from a user or process/application to articulate the types of metadata/enhancement and determine what operations may be done in real time and what must be done offline based on complexity of the request, type of data, available processing resources, priority/urgency of the operation.

In some examples, the online metadata agent may access existing metadata or enhancement functionality within a node or launch a selected algorithm package to perform real time metadata/enhancement actions on the data stream and create a source-file linked metadata record that may be passed to the data query manager 610 for incorporation and synchronization with other authorized and/or relevant instances of the data query manager 610. In the example of source file enhancement, the original file may be archived and linked with appropriate metadata record while the modified file is returned to the requestor.

In some examples, the offline metadata agent may be implemented as the real time agent instantiated on the server/file storage that runs metadata/enhancement routines offline due to resource availability, complexity of operations, and/or lower priority setting. Subsequent behavior may be similar to the online metadata once post-processing has been completed.

In some examples, the metadata/data enrichment manager 640 evaluates metadata/enhancement requests and priority. In some examples, the metadata/data enrichment manager 640 may select appropriate operations for base metadata and enhancement operations. In some examples, the metadata/data enrichment manager 640 may invoke and/or otherwise call to the AMR 634 that heuristically may suggest proper algorithms for advanced data operations and can recommend algorithm combinations (e.g., algorithm recipes) from prior operations that may be preferred by different customers or operate best on certain hardware.

In some examples, the metadata/data enrichment manager 640 identifies real-time operations from post-processed operations and confirm with the user allowing user to modify. In some examples, the metadata/data enrichment manager 640 launch modules (e.g., AI, NNLP, analytics, statistics, etc.) to generate metadata and/or enhance existing data (e.g., hyper-resolution or false image enhancements) on local node or supporting compute platform(s). In some examples, the metadata/data enrichment manager 640 manages archiving of sources, appending metadata records to the data query manager 610 (with source links), etc. In some examples, a single instance of the metadata/data enrichment manager 640 may manage multiple metadata/enhancement operations.

In the illustrated example, the ADM system 600 includes the distributed datastore 644 to record data. For example, the distributed datastore 644 may include the metadata storage 646 to record and/or otherwise store metadata. In some examples, the distributed datastore 644 may include the raw datastore 648 to record raw and/or otherwise unprocessed data. The distributed datastore 644 may be implemented by one or more volatile memories (e.g., a Synchronous Dynamic Random Access Memory (SDRAM), Dynamic Random Access Memory (DRAM), RAMBUS Dynamic Random Access Memory (RDRAM), etc.), one or more non-volatile memories (e.g., flash memory), and/or combination(s) thereof. The distributed datastore 644 may additionally or alternatively be implemented by one or more double data rate (DDR) memories, such as DDR, DDR2, DDR3, DDR4, mobile DDR (mDDR), etc. The distributed datastore 644 may additionally or alternatively be implemented by one or more mass storage devices such as hard disk drive(s) (HDD(s)), compact disk (CD) drive(s), digital versatile disk (DVD) drive(s), solid-state disk (SSD) drive(s), etc. While in the illustrated example the distributed datastore 644 is illustrated as a single datastore, the distributed datastore 644 may be implemented by any number and/or type(s) of datastores. Furthermore, the data stored in the distributed datastore 644 may be in any data format such as, for example, binary data, comma delimited data, tab delimited data, structured query language (SQL) structures, etc.

In some examples, the metadata storage 646, the raw datastore 648, and/or, more generally, the distributed datastore 644 may implement one or more databases. The term “database” as used herein means an organized body of related data, regardless of the manner in which the data or the organized body thereof is represented. For example, the organized body of related data may be in the form of one or more of a table, a map, a grid, a packet, a datagram, a frame, a file, an e-mail, a message, a document, a report, a list or in any other form.

As used herein, data is information in any form that may be ingested, processed, interpreted and/or otherwise manipulated by processor circuitry to produce a result. The produced result may itself be data.

As used herein “threshold” is expressed as data such as a numerical value represented in any form, that may be used by processor circuitry as a reference for a comparison operation.

As used herein, a model is a set of instructions and/or data that may be ingested, processed, interpreted and/or otherwise manipulated by processor circuitry to produce a result. Often, a model is operated using input data to produce output data in accordance with one or more relationships reflected in the model. The model may be based on training data.

While an example manner of implementing the ADM system 600 is illustrated in FIG. 6, one or more of the elements, processes, and/or devices illustrated in FIG. 6 may be combined, divided, re-arranged, omitted, eliminated, and/or implemented in any other way. Further, the ADM console 602, the data sources 604, the data ingestion manager 606, the pre-processing manager 608, the data query manager 610, the data query handler 612, the query cache cluster manager 614, the metadata cluster manager 616, the data publishing manager 618, the scheduler 620, the node manager 622, the preferred nodes table 624, the network plane 626, the data plane 628, the control plane 630, the data security manager 632, the AMR 634, the analytics manager 636, the algorithms 638, the metadata/data enrichment manager 640, the resource manager 642, the distributed datastore 644, the metadata storage 646, the raw datastore 648, the interface 650, and/or, more generally, the ADM system 600 of FIG. 6, may be implemented by hardware alone or by hardware in combination with software and/or firmware. Thus, for example, any of the ADM console 602, the data sources 604, the data ingestion manager 606, the pre-processing manager 608, the data query manager 610, the data query handler 612, the query cache cluster manager 614, the metadata cluster manager 616, the data publishing manager 618, the scheduler 620, the node manager 622, the preferred nodes table 624, the network plane 626, the data plane 628, the control plane 630, the data security manager 632, the AMR 634, the analytics manager 636, the algorithms 638, the metadata/data enrichment manager 640, the resource manager 642, the distributed datastore 644, the metadata storage 646, the raw datastore 648, the interface 650, and/or, more generally, the ADM 600, could be implemented by processor circuitry, analog circuit(s), digital circuit(s), logic circuit(s), programmable processor(s), programmable microcontroller(s), graphics processing unit(s) (GPU(s)), digital signal processor(s) (DSP(s)), application specific integrated circuit(s) (ASIC(s)), programmable logic device(s) (PLD(s)), and/or field programmable logic device(s) (FPLD(s)) such as Field Programmable Gate Arrays (FPGAs). Further still, the ADM system 600 of FIG. 6 may include one or more elements, processes, and/or devices in addition to, or instead of, those illustrated in FIG. 6, and/or may include more than one of any or all of the illustrated elements, processes and devices.

Flowcharts representative of example hardware logic circuitry, machine readable instructions, hardware implemented state machines, and/or any combination thereof for implementing the ADM system 600 of FIG. 6 are shown in FIGS. 7, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, and/or 27. The machine readable instructions may be one or more executable programs or portion(s) of an executable program for execution by processor circuitry, such as the processor circuitry J112 shown in the example processor platform J100 discussed below in connection with FIG. J1 and/or the example processor circuitry discussed below in connection with FIGS. J2 and/or J3. The program may be embodied in software stored on one or more non-transitory computer readable storage media such as a CD, a floppy disk, a hard disk drive (HDD), a solid-state drive (SSD), a DVD, a Blu-ray disk, a volatile memory (e.g., Random Access Memory (RAM) of any type, etc.), or a non-volatile memory (e.g., electrically erasable programmable read-only memory (EEPROM), FLASH memory, an HDD, an SSD, etc.) associated with processor circuitry located in one or more hardware devices, but the entire program and/or parts thereof could alternatively be executed by one or more hardware devices other than the processor circuitry and/or embodied in firmware or dedicated hardware. The machine readable instructions may be distributed across multiple hardware devices and/or executed by two or more hardware devices (e.g., a server and a client hardware device). For example, the client hardware device may be implemented by an endpoint client hardware device (e.g., a hardware device associated with a user) or an intermediate client hardware device (e.g., a radio access network (RAN)) gateway that may facilitate communication between a server and an endpoint client hardware device). Similarly, the non-transitory computer readable storage media may include one or more mediums located in one or more hardware devices. Further, although the example program is described with reference to the flowcharts illustrated in FIGS. 7, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, and/or 27, many other methods of implementing the example ADM system 600 may alternatively be used. For example, the order of execution of the blocks may be changed, and/or some of the blocks described may be changed, eliminated, or combined. Additionally or alternatively, any or all of the blocks may be implemented by one or more hardware circuits (e.g., processor circuitry, discrete and/or integrated analog and/or digital circuitry, an FPGA, an ASIC, a comparator, an operational-amplifier (op-amp), a logic circuit, etc.) structured to perform the corresponding operation without executing software or firmware. The processor circuitry may be distributed in different network locations and/or local to one or more hardware devices (e.g., a single-core processor (e.g., a single core central processor unit (CPU)), a multi-core processor (e.g., a multi-core CPU), etc.) in a single machine, multiple processors distributed across multiple servers of a server rack, multiple processors distributed across one or more server racks, a CPU and/or a FPGA located in the same package (e.g., the same integrated circuit (IC) package or in two or more separate housings, etc.).

The machine readable instructions described herein may be stored in one or more of a compressed format, an encrypted format, a fragmented format, a compiled format, an executable format, a packaged format, etc. Machine readable instructions as described herein may be stored as data or a data structure (e.g., as portions of instructions, code, representations of code, etc.) that may be utilized to create, manufacture, and/or produce machine executable instructions. For example, the machine readable instructions may be fragmented and stored on one or more storage devices and/or computing devices (e.g., servers) located at the same or different locations of a network or collection of networks (e.g., in the cloud, in edge devices, etc.). The machine readable instructions may require one or more of installation, modification, adaptation, updating, combining, supplementing, configuring, decryption, decompression, unpacking, distribution, reassignment, compilation, etc., in order to make them directly readable, interpretable, and/or executable by a computing device and/or other machine. For example, the machine readable instructions may be stored in multiple parts, which are individually compressed, encrypted, and/or stored on separate computing devices, wherein the parts when decrypted, decompressed, and/or combined form a set of machine executable instructions that implement one or more operations that may together form a program such as that described herein.

In another example, the machine readable instructions may be stored in a state in which they may be read by processor circuitry, but require addition of a library (e.g., a dynamic link library (DLL)), a software development kit (SDK), an application programming interface (API), etc., in order to execute the machine readable instructions on a particular computing device or other device. In another example, the machine readable instructions may need to be configured (e.g., settings stored, data input, network addresses recorded, etc.) before the machine readable instructions and/or the corresponding program(s) can be executed in whole or in part. Thus, machine readable media, as used herein, may include machine readable instructions and/or program(s) regardless of the particular format or state of the machine readable instructions and/or program(s) when stored or otherwise at rest or in transit.

The machine readable instructions described herein can be represented by any past, present, or future instruction language, scripting language, programming language, etc. For example, the machine readable instructions may be represented using any of the following languages: C, C++, Java, C#, Perl, Python, JavaScript, HyperText Markup Language (HTML), Structured Query Language (SQL), Swift, etc.

As mentioned above, the example operations of FIGS. 7 and 17-29 may be implemented using executable instructions (e.g., computer and/or machine readable instructions) stored on one or more non-transitory computer and/or machine readable media such as optical storage devices, magnetic storage devices, an HDD, a flash memory, a read-only memory (ROM), a CD, a DVD, a cache, a RAM of any type, a register, and/or any other storage device or storage disk in which information is stored for any duration (e.g., for extended time periods, permanently, for brief instances, for temporarily buffering, and/or for caching of the information). As used herein, the terms non-transitory computer readable medium and non-transitory computer readable storage medium is expressly defined to include any type of computer readable storage device and/or storage disk and to exclude propagating signals and to exclude transmission media.

“Including” and “comprising” (and all forms and tenses thereof) are used herein to be open ended terms. Thus, whenever a claim employs any form of “include” or “comprise” (e.g., comprises, includes, comprising, including, having, etc.) as a preamble or within a claim recitation of any kind, it is to be understood that additional elements, terms, etc., may be present without falling outside the scope of the corresponding claim or recitation. As used herein, when the phrase “at least” is used as the transition term in, for example, a preamble of a claim, it is open-ended in the same manner as the term “comprising” and “including” are open ended. The term “and/or” when used, for example, in a form such as A, B, and/or C refers to any combination or subset of A, B, C such as (1) A alone, (2) B alone, (3) C alone, (4) A with B, (5) A with C, (6) B with C, or (7) A with B and with C. As used herein in the context of describing structures, components, items, objects and/or things, the phrase “at least one of A and B” is intended to refer to implementations including any of (1) at least one A, (2) at least one B, or (3) at least one A and at least one B. Similarly, as used herein in the context of describing structures, components, items, objects and/or things, the phrase “at least one of A or B” is intended to refer to implementations including any of (1) at least one A, (2) at least one B, or (3) at least one A and at least one B. As used herein in the context of describing the performance or execution of processes, instructions, actions, activities and/or steps, the phrase “at least one of A and B” is intended to refer to implementations including any of (1) at least one A, (2) at least one B, or (3) at least one A and at least one B. Similarly, as used herein in the context of describing the performance or execution of processes, instructions, actions, activities and/or steps, the phrase “at least one of A or B” is intended to refer to implementations including any of (1) at least one A, (2) at least one B, or (3) at least one A and at least one B.

As used herein, singular references (e.g., “a”, “an”, “first”, “second”, etc.) do not exclude a plurality. The term “a” or “an” object, as used herein, refers to one or more of that object. The terms “a” (or “an”), “one or more”, and “at least one” are used interchangeably herein. Furthermore, although individually listed, a plurality of means, elements or method actions may be implemented by, e.g., the same entity or object. Additionally, although individual features may be included in different examples or claims, these may possibly be combined, and the inclusion in different examples or claims does not imply that a combination of features is not feasible and/or advantageous.

FIG. 7 is a flowchart representative of example machine readable instructions and/or example operations 700 that may be executed and/or instantiated by processor circuitry to generate an example recommendation to integrate a hardware, software, and/or firmware feature in a semiconductor-based device (e.g., a silicon-based device). The machine readable instructions and/or the operations 700 of FIG. 7 begin at block 702, at which a candidate algorithm (e.g., an AI/ML algorithm) is identified. At block 704, a business impact of the candidate algorithm may be quantified. At block 706 end-to-end use case(s) that may utilize the candidate algorithm are defined. At block 708 current and future candidate platforms that may be utilized for the end-to-end use case(s) may be identified. At block 710, target key performance indicators (KPIs) may be defined and technical benchmarking may be implemented at block 712. At block 714, potential changes to a semiconductor-based device may be identified and computer-aided design (e.g., CAD, UCAD, etc.) experiments may be conducted at block 716. At block 718, an impact of different hypotheses are quantified, which may result in generating a semiconductor-device based recommendation at block 720. For example, a recommendation to change (e.g., add, remove, and/or modify) hardware, software, and/or firmware associated with a semiconductor-based device may be generated at block 720. In some such examples, the semiconductor-based device may be manufactured based on the recommendation. In some such examples, the manufactured semiconductor-based device may execute and/or otherwise implement the candidate algorithm identified at block 702 using hardware, software, and/or firmware of the manufactured semiconductor-based device.

FIG. 8 is an illustration of an example edge network environment 800 including an example edge gateway 802 and an example edge switch 804 that may implement the ADM system 600 of FIG. 6 and/or otherwise as disclosed herein. In some examples, the edge gateway 802 and/or the edge switch 804 may implement resources of the edge devices layer 210 of FIG. 2. In some such examples the edge gateway 802 and/or the edge switch 804 may implement the access point or base station 140 of FIG. 1, the local processing hub 150 of FIG. 1, and/or the nodes 215 of FIG. 2. For example, the edge gateway 802 and/or the edge switch 804 may implement the edge devices layer 210 of FIG. 2.

The edge network environment 800 of the illustrated example includes an example public network 806, an example private network 808, and an example edge cloud 810. In this example, the public network 806 may implement a telephone service provider (TSP) network (e.g., a Long-Term Evolution (LTE) network, a 5G network, a Telco network, etc.). For example, the public network 806 may implement the network access layer 220 of FIG. 2, the core network 230 of FIG. 2, and/or the cloud data center layer 240 of FIG. 2. In this example, the private network 808 may implement an enterprise network (e.g., a close campus network, a private LTE network, a private 5G network, etc.). For example, the private network 808 may implement the endpoint layer 200 of FIG. 2 and/or the edge devices layer 210 of FIG. 2. In some examples, the edge cloud 810 may be implemented by one or more hardware, software, and/or firmware resources. For example, the edge cloud 810 may be implemented by one or more computer servers. In this example, the edge cloud 810 may implement an enterprise edge cloud. For example, the edge cloud 810 may implement the edge cloud 110 of FIGS. 1, 2, and/or 3.

In the illustrated example of FIG. 8, the edge network environment 800 may implement a smart factory (e.g., a smart industrial factory), a process control environment, etc. For example, the edge network environment 800 may implement one(s) of the computational use cases 205 of FIG. 2, such as a manufacturing, smart building, logistics, vehicle, and/or video computational use cases.

The edge network environment 800 of the illustrated example includes an example process control system 812, example robots (e.g., collaborative robots, robot arms, etc.) 814, a first example industrial machine (e.g., an autonomous industrial machine) 816, a second example industrial machine 818, a third example industrial machine 820, a fourth example industrial machine 822, an example predictive maintenance system 824, an example vehicle (e.g., a truck, an autonomous truck, an autonomous vehicle, etc.) 826, a first example monitoring sensor 828, a second example monitoring sensor 830, and example endpoint devices 832, 834, 836. In some examples, the process control system 812 may include one or more industrial machines such as a silo, a smokestack, a conveyor belt, a mixer, a pump, etc., and/or a combination thereof. For example, the process control system 812 may implement the business and industrial equipment 163 of FIG. 1, the smart cities and building devices 166 of FIG. 1, etc.

In some examples, the robots 814 may implement hydraulic and/or electromechanical robots that may be configured to execute manufacturing tasks (e.g., lifting equipment, assembling components, etc.), industrial tasks, etc. For example, the robots 814 may implement the business and industrial equipment 163 of FIG. 1, the smart cities and building devices 166 of FIG. 1, etc. In some examples, the industrial machines 816, 818, 820, 822 are autonomous machines, such as AGVs, autonomous forklifts, scissor lifts, etc. For example, the industrial machines 816, 818, 820 may implement the business and industrial equipment 163 of FIG. 1, the drones 165 of FIG. 1, the smart cities and building devices 166 of FIG. 1, etc. In some examples, the predictive maintenance system 824 may implement one or more computing devices, servers, etc., that identify maintenance alerts, fault predictions, etc., associated with equipment of the edge network environment 800 based on sensor data (e.g., prognostic health data). For example, the predictive maintenance system 824 may implement the business and industrial equipment 163 of FIG. 1, the smart cities and building devices 166 of FIG. 1, the sensors and IoT devices 167 of FIG. 1, etc.

In some examples, the vehicle 826 may implement one of the autonomous vehicles 161 of FIG. 1. In some examples, the first monitoring sensor 828 and/or the second monitoring sensor 830 are video cameras. For example, the first monitoring sensor 828 and/or the second monitoring sensor 830 may implement the business and industrial equipment 163 of FIG. 1, the video capture devices 164 of FIG. 1, the smart cities and building devices 166 of FIG. 1, the sensors and IoT devices 167 of FIG. 1, etc. Alternatively, the first monitoring sensor 828 and/or the second monitoring sensor 830 may implement a thermal camera (e.g., an infrared camera), an air pollution sensor, a carbon dioxide sensor, a temperature sensor, a humidity sensor, an air pressure sensor, etc., or any other type of sensor.

In this example, the endpoint devices 832, 834, 836 include a first example endpoint device 832, a second example endpoint device 834, and a third example endpoint device 836. In some examples, one(s) of the endpoint devices 832, 834, 836 may implement consumer computing devices, user equipment, etc. For example, one or more of the endpoint devices 832, 834, 836 may implement the user equipment 162 of FIG. 1. In some such examples, one or more of the endpoint devices 832, 834, 836 may be implemented by a smartphone, a tablet computer, a desktop computer, a laptop computer, a wearable device (e.g., a headset or headset display, an augmented reality (AR) headset, a smartwatch, smart glasses, etc.), etc.

In the illustrated example of FIG. 8, the edge gateway 802 may facilitate communication, data transfers, etc., between different networks, such as communication from a source service, a source appliance, etc., of the public network 806 to a target service, a target appliance, etc., of the private network 808. For example, the edge gateway 802 may receive a data stream including one or more data packets from a source (e.g., a data source), a producer (e.g., a data producer), etc. In some such examples, the edge gateway 802 may receive the data stream from the vehicle 826, the second endpoint device 834, the third endpoint device 836, etc., to be transmitted to a target service, a target appliance, etc., which may be implemented by the cloud data center 130 of FIG. 1, the cloud data center 245 of FIG. 2, the cloud or data center 360 of FIG. 3, etc.

In some examples, the edge gateway 802 may facilitate communication, data transfers, etc., between a source service, a source appliance, etc., of the private network 808 to a target service, a target appliance, etc., of the public network 806. For example, the edge gateway 802 may receive a data stream including one or more data packets from a source (e.g., a data source), a producer (e.g., a data producer), etc., which may be implemented by the cloud data center 130 of FIG. 1, the cloud data center 245 of FIG. 2, the cloud or data center 360 of FIG. 3, etc. In some such examples, the edge gateway 802 may receive the data stream from the cloud data center 130 of FIG. 1, the cloud data center 245 of FIG. 2, the cloud or data center 360 of FIG. 3, etc., to be transmitted to the vehicle 826, the second endpoint device 834, the third endpoint device 836, etc.

In the illustrated example of FIG. 8, the edge switch 804 may facilitate communication, data transfers, etc., between different sources and targets within a network, such as communication from a source service, a source appliance, etc., of the private network 808 to a target service, a target appliance, etc., of the private network 808. For example, the edge switch 804 may receive a data stream from the edge gateway 802, the edge cloud 810, the process control system 812, one(s) of the robots 814, one(s) of the industrial machines 816, 818, 820, 822, the predictive maintenance system 824 (or sensor(s) thereof), the first monitoring sensor 828, the second monitoring sensor 830, the first endpoint device 832, the second endpoint device 834, the third endpoint device 836, etc. In some such examples, the edge switch 804 may transmit the data stream to a destination within the private network 808. For example, the edge switch 804 may transmit the data stream to at least one of the edge gateway 802, the edge cloud 810, the process control system 812, one(s) of the robots 814, one(s) of the industrial machines 816, 818, 820, 822, the predictive maintenance system 824 (or sensor(s) thereof), the vehicle 826, the first monitoring sensor 828, the second monitoring sensor 830, the first endpoint device 832, the second endpoint device 834, or the third endpoint device 836.

In some examples, the edge gateway 802 and/or the edge switch 804 may implement adaptive data management based on global observability at the edge, which may be implemented by the edge network environment 800 or portion(s) thereof. In some examples, the edge network environment 800 may implement a large number and/or different types of applications, such as machine vision applications implemented by the robots 814, autonomous driving applications implemented by the vehicle 826, etc. In some such examples, the data generated by the private network 808 is relatively diverse because of the vast range of data sources, such as sensors, controllers, services, and/or user input that may be processed and analyzed to identify anomalies and trends in the data. For example, the edge gateway 802 and/or the edge switch 804 may facilitate the transmission of data including sensor data or measurements, video feeds, still images, predictive maintenance alerts or control commands, robotic control commands, etc., and/or a combination thereof.

In some examples, the edge gateway 802 and/or the edge switch 804 may transfer data to components of the ADM system 600 of FIG. 6 to execute one(s) of the algorithms 638 to implement ADM as disclosed herein. In some examples, the edge gateway 802 and/or the edge switch 804 may execute the one(s) of the algorithms 638 to implement ADM as disclosed herein. In some examples, there are a plurality of the edge gateways 802 and/or a plurality of the edge switches 804 in the edge network environment 800. The algorithms 638 may be executed in multiple places of the edge network environment 800 (e.g., by ones of the edge gateways 802, the edge switches 804, etc., or any other device(s)). In some examples, the different ones of the edge gateways 802, the edge switches 804, etc., may have more or less observability based on the data that they process and/or otherwise encounter. Accordingly, two different ones of the devices of FIG. 8 may develop, train, and/or otherwise generate different one(s) of the algorithms 638 based on the data processed by each of the different ones of the devices. For example, a first one of the edge switches 804 may observe 10% of the edge network environment 800 and a second one of the edge switches 804 may observe 90% of the edge network environment 800, which may become the basis for the differences in data outputs generated by the algorithms 638 executed by the first and second one of the devices. In some such examples, the first one of the devices may transmit and/or otherwise propagate data outputs from its execution of the algorithms 638 to the second one of the devices. In some such examples, the second one of the devices may transmit and/or otherwise propagate model outputs from its execution of the algorithms 638 to the first one of the devices for cross-training and/or cross-correlation of the algorithms 638.

In some examples, data generated by the private network 808 may be immense. In some examples, a data source, such as the process control system 812, one(s) of the robots 814, one(s) of the industrial machines 816, 818, 820, 822, the predictive maintenance system 824 (or sensor(s) thereof), the vehicle 826, the first monitoring sensor 828, the second monitoring sensor 830, the first endpoint device 832, the second endpoint device 834, and/or the third endpoint device 836, may have insufficient computing resources (e.g., one or more processors, one or more accelerators, one or more memories, one or more mass storage discs, etc.) to analyze the data generated by the data source. In some such examples, the data source may be unable to identify redundant data, less important or less significant data, etc., due to insufficient computing resources and therefore may flood the private network 808 with a significant quantity of data at relatively short intervals. Advantageously, the edge gateway 802, the edge switch 804, and/or, more generally, the ADM 600 of FIG. 6, may implement ADM as disclosed herein to process data based on at least one of a content and/or a context of the data.

FIG. 9 is a block diagram of an example portion 900 of the example ADM system 600 of FIG. 6 to effectuate identification and mitigation of ethical divergence in a data stream. The portion 900 of the illustrated example is a system including the data ingestion manager 606, the data query manager 610, the data publishing manager 618, the node manager 622, the preferred nodes table 624, the network plane 626, the AMR 634, the analytics manager 636, the algorithms 638, the metadata/data enrichment manager 640, the resource manager/orchestration agent 642, and the distributed datastore 644 of FIG. 6.

In the illustrated example, the portion 900 of the ADM system 600 can implement a first example operation 902 (identified by a circle that encloses “1”), a second example operation 904 (identified by a circle that encloses “2”), a third example operation 906 (identified by a circle that encloses “3”), and a fourth example operation 908 (identified by a circle that encloses “4”). During the first operation 902, the data ingestion manager 606 may capture data from the data sources 604. The data ingestion manager 606 may pre-process the data by metadata tagging with data management settings (e.g., a locality or location of the data, expiration date, a source of the data, a type of the data, etc.). The AMR 634 may monitor the data on the network plane 626 and/or invoke the distributed datastore 644 to execute analytics using a model (e.g., an AI/ML model, a data graph model, etc.). The AMR 634 may synchronize with the resource manager/orchestration agent 642 to identify new one(s) of the algorithms 638 that may be executed and/or instantiated at a node. For example, the AMR 634 can provide metadata associated with ingested data to the resource manager/orchestration agent 642. In some examples, the resource manager/orchestration agent 642 can identify an AI/ML model that corresponds to the metadata and provide the AI/ML model to the AMR 634 for execution and/or instantiation at the node to execute a workload associated with the ingested data or data to be subsequently ingested.

During the second operation 904, the resource manager/orchestration agent 642 can identify criteria for global decisions (e.g., decisions, policies or policy determinations, etc., that may be applicable to a substantial or entire portion of an environment), provide guidance on the criteria, and/or execute resource management and/or orchestration of resources (e.g., hardware, software, and/or firmware resources) of the ADM system 600. In some examples, the resource manager/orchestration agent 642 can orchestrate resources and allocate different one(s) of the algorithms 638, select preferred node(s) of the preferred nodes table 624 to monitor, etc., and/or combination(s) thereof. In some examples, the resource manager/orchestration agent 642 can orchestrate a new one of the algorithms 638 to be provided from an algorithm library based on algorithm ratings, scores, etc., if available, or point to a local datastore (if available locally) for faster access.

During the third operation 906, the analytics manager 636 monitors data from the data sources 604 for further processing. For example, the analytics manager 636 can learn characteristics of a data stream and/or characteristics of an AI application node (consuming or attempting to consume the data stream) for an observation period and associate with at least one of location, action recognition, type, data, size, owner, classification tags, frequency of appearance, etc., and/or combination(s) thereof.

During the fourth operation 908, the portion 900 of the ADM system 600, and/or, more generally, the ADM system 600, can implement example data orchestration and nodes activities 910. Example activities, actions, operations, tasks, etc., can include identification and mitigation of ethical divergence of a data stream. Example activities, operations, tasks, etc., can include tagging data within the datastream with one or more metadata tags, such as injecting a hash that represents time, date, location, identification of data stream ownership, etc. into the data stream, blockchaining data within the data stream, modifying the data stream to provide only a portion of the data, obscuring data within the data stream (e.g., a sensitive portion of image data in the data stream may be blurred out), or prohibiting the consumption of the data stream by one or more AI application nodes, among other operations.

FIG. 10 is a first example workflow 1000 that may be implemented by the ADM system architecture 600 to generate correlation factors and/or delta scores for graph models captured in an index table for future use. In some examples, the first workflow 1000 may implement a static graph model correlation index and/or delta score workflow. In some examples, the first workflow 1000 may generate at least one graph node representation of a data stream based on at least one or more data points in the data stream.

During a first operation, an example metadata manager 1002 may execute contextual analytics on a first example dataset 1004 (identified by dataset A) and a second example dataset 1006 (identified by dataset B). For example, the metadata manager 1002 may be implemented by the metadata/data enrichment manager 640 of FIG. 6. In some examples, the first dataset 1004 and/or the second dataset 1006 may be implemented by data stored in at least one of the metadata storage 646 or the raw datastore 648 of FIG. 6.

During a second operation, the metadata manager 1002 may generate example graph node model representations 1008 based on the contextual analytics associated with the first dataset 1004 and the second dataset 1006. The graph node model representations 1008 may include a first example graph node model representation 1010 and a second example graph node model representation 1012. For example, the first graph node model representation 1010 may correspond to the first dataset 1004 and the second graph node model representation 1012 may correspond to the second dataset 1006. In some examples, the first workflow 1000 may generate at least one graph node representation (e.g., the first graph node model representation 1010) of a target data stream based on at least one or more data points in the data stream. In some examples, the dataset A 1004 may include at least a portion of the data in a target data stream consumed or attempted to be consumed by an AI application node. The example dataset A 1004 may include one or more example data points representative of information within the target data stream. In some examples, the first workflow 1000 may generate at least one graph node representation (e.g., the second graph node model representation 1012) of a nominal data stream based on at least one or more data points in the nominal data stream. In some examples, the dataset B 1006 may include at least a portion of the data in a nominal data stream consumed by an AI application node. The example dataset B 1006 may include one or more example data points representative of information within the nominal data stream.

During a third operation, an example graph model metadata comparator 1014 may execute context and variance calculations based on at least one of the first graph node model representation 1010 or the second graph node model representation 1012. In some examples, the graph model metadata comparator 1014 may be implemented by the metadata/data enrichment manager 640. In some examples, the graph model metadata comparator 1014 may utilize the context and variance calculations to compare the first graph node model representation 1010, representative of information within the target data stream, to the second graph node model representation 1012, representative of information within a nominal data stream.

During a fourth operation, the graph model metadata comparator 1014 may generate, determine, and/or otherwise output correlation factors for the first graph node model representation 1010 and/or the second graph node model representation 1012 and/or output one or more delta scores representative of a difference between the first graph node model representation 1010 and/or the second graph node model representation 1012.

In some examples, the graph model metadata comparator 1014 may calculate a delta score of the difference between the first graph node model representation 1010 and the second graph node model representation 1012. In some examples, the graph model metadata comparator 1014 may convert the first graph node model representation 1010 and the second graph node model representation 1012 from a cluster of data points to a centroid of the cluster of data points. In some examples, the graph model metadata comparator 1014 may convert the first graph node model representation 1010 and the second graph node model representation 1012 into vector representations through a linear transformation of data points within the first graph node model representation 1010 and the second graph node model representation 1012. In some examples, the graph model metadata comparator 1014 may calculate a similarity score that utilizes a distance metric common in database queries. The example similarity score represents a distance between a vector representative of the first graph node model representation 1010 and a vector representative of the second graph node model representation 1012. In some examples, the graph model metadata comparator 1014 calculates the delta score by normalizing the similarity score to a value between 0 and 1 and subtracting that value from 1 (e.g., delta score=1−similarity score).

In some examples, the correlation factors and/or the delta scores may be stored in an index table (or other form of data representation) that may be utilized for future use. For example, the analytics manager 636 may execute the algorithms 638 based on the correlation factors and/or the delta scores. In response to determining the correlation factors and/or delta scores 1016, the first workflow 1000 of the illustrated example of FIG. 10 concludes.

FIG. 11 is a second example workflow 1100 that may be implemented by the ADM system 600 to generate correlation factors and or delta scores for dynamic query operation. In some examples, the second workflow 1100 may implement an on-demand graph model correlation and/or delta score workflow.

During a first operation, an example data query manager 1102 may obtain a data query from a node (e.g., a data requester). The data query manager 1102 may obtain a first example graph node model representation 1104 from a first example metadata storage 1106. The data query manager 1102 may obtain a second example graph node model representation 1108 from a second example metadata storage 1110. In some examples, the first graph node model representation 1104 may be implemented by the first graph node model representation 1010 of FIG. 10. In some examples, the second graph node model representation 1108 may be implemented by the second graph node model representation 1012 of FIG. 10. In some examples, the first metadata storage 1106 and/or the second metadata storage 1108 may be implemented by the metadata storage 646 of FIG. 6. In some examples, the first graph node model representation 1104 may include at least a portion of the data in a target data stream consumed or attempted to be consumed by an AI application node. The example first graph node model representation 1104 may include one or more example data points representative of information within the target data stream. In some examples, the second graph node model representation 1108 may include at least a portion of the data in a nominal data stream consumed by an AI application node. The example second graph node model representation 1108 may include one or more example data points representative of information within the nominal data stream.

During a second operation, the data query manager 1102 provides at least one of the data query, the first graph node model representation 1104, or the second graph node model representation 1108 to an example graph model metadata comparator 1112. In some examples, the graph model metadata comparator 1112 may be implemented by the metadata/data enrichment manager 640 of FIG. 6.

During a third operation, the graph model metadata comparator 1112 may compare the at least one of the data query, the first graph node model representation 1104, or the second graph node model representation 1108 and generate value(s) based on the comparison(s). In some examples, the graph model metadata comparator 1112 may generate one or more delta scores (e.g., the generated value(s)) of the difference between the first graph node model representation 1104 and the second graph node model representation 1108. The graph model metadata comparator 1112 may provide the value(s) 1114 to the data query manager 1102. In some examples, the value(s) may be implemented by a correlation factor and/or a delta score for dynamic query operation 1116. In response to determining the correlation factor and/or the delta score for dynamic query operation, the second workflow 1100 of FIG. 11 concludes.

FIG. 12 is an illustration of a first example graph model 1202 and a second example graph model 1204 for depicting groups of related data (e.g., ingested data, stored data, etc.) and metadata connected linked via example strength/distance vectors 1205, 1207. In some examples, at least one of the first graph model 1202 or the second graph model 1204 can implement the data graph model 1326 (FIG. 13), or portion(s) thereof. In some examples, the first graph model 1202 and the second graph model 1204 are part of a third example graph model 1200.

In some examples, the first graph model 1202 and/or the second graph model 1204 is/are contextual data graph model(s). The first graph model 1202 includes a first example major node 1206, a first example adjacent node 1208, and a second example adjacent node 1210. The second graph model 1204 includes a second example major node 1212, a third example adjacent node 1214, a fourth example adjacent node 1216, and an example adjacent node grouping 1218. The patterns (e.g., solid, dotted, striped, hashed, etc.) of the various major nodes and adjacent nodes illustrated in FIG. 12 depict the various descriptors (e.g., keywords) of the metadata associated with raw data, ingested data, stored data, etc. stored in the distributed datastore 644 of FIG. 6, the datastore 1060 of FIG. 10, etc. The example adjacent nodes illustrated in FIG. 12 represent the metadata (e.g., the metadata 1324 of FIG. 13) associated with the ingested data, stored data, raw data etc. stored in the distributed datastore 644 of FIG. 6, the datastore 1320 of FIG. 13, etc. Additionally and/or alternatively, the graph models 1202, 1204 of metadata descriptors illustrated in FIG. 12 can represent data points, raw data, ingested data, stored data, etc., stored in other memory or other storage devices in a cloud (e.g., the edge cloud 110, the cloud data center 130, etc.). For example, the nodes in the graph models 1202 and 1204 may represent data points from ingested data (e.g., data within a target data stream).

The lines connecting the major nodes 1206, 1212 to the adjacent nodes or the adjacent nodes to each other represent the strength vectors 1205, 1207 between the two nodes. In some examples, the strength vectors 1205, 1207 are of a single dimension and/or multiple dimensions and includes descriptors (e.g., keywords, similarity ratings (e.g., 1 through 5 with 1 being relatively not similar and 5 being substantially similar and/or identical, low/medium/high, etc.), characters, strings, numerical values, etc.) that represent how strongly the metadata of the raw data, ingested data, stored data, etc., match. In some examples, the length of the strength vectors 1205, 1207 shown in the illustrated example of FIG. 12 symbolize the level of commonality, similarity, association, etc., two connected nodes share with one another. In other words, the strength vector length can depict the correlation between two connected nodes based on how frequently descriptors (e.g., data points, keywords, terms, numerical values, data types, data categories, etc.) appear in both connected nodes. For example, the metadata of the first adjacent node 1208 can have fewer matching data (e.g., descriptors, data blocks, etc.) with the metadata of the first major node 1206 than the metadata of the third adjacent node 1214 has with the second major node 1212. In some examples, the ML circuitry 1306 (FIG. 13), and/or, more generally, the data usage monitoring circuitry 1300 (FIG. 13), can determine the number of matching metadata (e.g., metadata descriptors) that exists between adjacent nodes. The data usage monitoring circuitry 1300 can generate the strength vector 1205, 1207 based on the amount of metadata descriptors that match (e.g., fully matching, partially matching, and/or not matching) between the adjacent nodes. The graph models described herein (e.g., the data graph model 1066, the first graph model 1202, the second graph model 1204, etc.) are formed and/or predicted based on the metadata descriptions (e.g., the metadata descriptors) because parsing through the contents of the raw data to find similarities would consume much more processing resources, memory bandwidth, and compute time than parsing through the metadata. In some examples, the more two graph models match, the lower the delta score between the two graph models.

In some examples, the first major node 1206 and the second major node 1212 have the same metadata descriptors. In some examples, the weighting of (e.g., length of the strength vectors 1205, 1207 between) the adjacent nodes (e.g., the first adjacent node 1208, the third adjacent node 1214, etc.) provide context and associated metadata descriptors. In some examples, the provided context of the graph models 1202, 1204 are tailored to a particular department and/or discipline within a company, an organization, a group, etc. For example, the first major node 1206 and the second major node 1212 can both include metadata that describe the associated raw data as belonging to the design engineering department of a bicycle company. In some examples, the first adjacent node 1208 and the third adjacent node 1214 can both include metadata that describe the raw data as belonging to the gear design segment of the design engineering department. However, since the strength vector connecting the second major node 1212 to the third adjacent node 1214 is shorter than the strength vector connecting the first major node 1206 and the first adjacent node 1208, it can be inferred that and/or otherwise be indicative of the second graph model 1204 having a stronger association with the gear design segment of the design engineering department than does the first graph model 1202.

In some examples, the first major node 1206 and the second major node 1212 have different metadata descriptors but are connected to adjacent nodes with similar metadata. In some examples, there is an implication that the first major node 1206 and the second major node 1212 have the same contextual definition. A user and/or operator can establish the example contextual definitions prescriptively depending on the metadata associations in the graph model(s). Additionally or alternatively, the data usage monitoring circuitry 1300 can perform predictions/operations/insights on the graph models 1202, 1204 to determine the contextual definitions based on events, task, and/or objects in common among adjacent nodes. In some examples, nodes of the graph models 1202, 1204 are grouped together such as an example grouping 1218 of the second graph model 1204. Example groupings of nodes can reinforce the contextual definition(s), descriptor(s), subject area(s), etc. of the major node (e.g., the second major node 1212).

In some examples, the term “to associate” is defined as to correlate, link, couple, and/or connect two or more datasets, data points, raw data, metadata, etc., in the form of a strength vector (e.g., the strength vectors 1205, 1207) based on similarities between the two or more datasets, data points, raw data, metadata, etc. By way of example, if a first metadata set has a sufficient number of same terms (e.g., over half of the total number of terms) as a second metadata set, then the strength vector is said to associate the first metadata and the second metadata. For example, if first raw data that the first metadata describes gets copied into storage on a different node, then the strength vector that associates the first metadata and the second metadata indicates to the data usage monitoring circuitry 1300 that second raw data is also to be copied into the same storage on the same node as the first raw data.

In some examples, a dataset, data point, raw data, metadata, etc. are associated with the factors the data usage monitoring circuitry 1300 determines as inputs to executed algorithms (e.g., one(s) of the algorithms 638 of FIG. 6, the ML model 1068 of FIG. 10, etc.). For example, if the data usage monitoring circuitry 1300 recognizes a cyclical trend of cyclical events (e.g., a recurring usage (e.g., file opening, reading, manipulating, etc.) every month) of a raw dataset, then the data usage monitoring circuitry 1300 can associate the cyclical trend with the raw dataset. In some examples, the data usage monitoring circuitry 1300 can represent the cyclical trend in the form of a data block (e.g., a cyclical trend metadata block that includes a time (e.g., “monthly”, “30 days”, etc.) and an action (e.g., “read from”, “write new”, “remove data”, “manipulate existing”, etc.)) included in metadata of the raw dataset. Therefore, in some examples, if the data usage monitoring circuitry 1300 analyzes the metadata of the raw data for a retention decision, the data usage monitoring circuitry 1300 can factor the cyclical trend associated with the metadata/raw data into the analysis, prediction, learning, and/or retention decision.

In some examples, the term “association” refers to a correlation, linkage, coupling, and/or connection between two or more datasets, data points, raw data, metadata, etc. In some examples, a strength vector associating two adjacent nodes of a graph model can represent and/or define the association between the two adjacent nodes. In some examples, the data usage monitoring circuitry 1300 determines factor(s) (e.g., uniqueness, retention cost, cyclical event, etc.) of ingested data and/or stored data, and the factor(s) is/are associated with the ingested data and/or the stored data for which the factor(s) were determined. By way of example, if first stored data (with associated first metadata) is found to have a uniqueness relative to second stored data (e.g., 1:2 ratio or 50 percent uniqueness), then the uniqueness associated with the first stored data is written into the first metadata (along with a storage location and/or an identifier of the stored data with which the uniqueness of the first data is compared (e.g., the second stored data)).

In some examples, the ML circuitry 1306, and/or, more generally, the data usage monitoring circuitry 1300 can receive a stream of first metadata and/or first raw data and a stream of second metadata and/or second raw data to generate a first graph model (e.g., the data graph model 1066, the first graph model 1202, the second graph model 1204, etc.) and a second graph model (e.g., the data graph model 1066, the first graph model 1202, the second graph model 1204, etc.). The data usage monitoring circuitry 1300 can contextually compress the first metadata and/or the second metadata that have correlating metadata and/or raw data content. In some examples, the data usage monitoring circuitry 1300 can perform context and variance calculations on the adjacent node(s) of the first graph model and/or the second graph model (e.g., the data graph model 1066, the first graph model 1202, the second graph model 1204, etc.) to contextually compress (e.g., further reduce the number of nodes in) the graph model(s). The data usage monitoring circuitry 1300 can convert the representative graph model(s) into first index table(s). The first index table(s) of the graph model representation(s) tabulate the raw data and/or the metadata that are depicted in the graph model(s). The data usage monitoring circuitry 1300 can also generate second index table(s) including correlation factors between the first graph model and the second graph model. The data usage monitoring circuitry 1300 can use the first index table(s) and/or the second index table(s) to execute learning, predictions, insights, operations, etc. on the ingested data, stored data, raw data, metadata, etc. that retain, move, modify, and/or discard the data being analyzed.

FIG. 13 is a block diagram of data usage monitoring circuitry 1300 to effectuate identification and mitigation ethical divergence. Alternatively, the data usage monitoring circuitry 1300 can be referred to as data consumption monitoring circuitry, ethical data usage monitoring circuitry, and/or ethical data consumption monitoring circuitry. The data usage monitoring circuitry 1300 of FIG. 13 may be instantiated (e.g., creating an instance of, bring into being for any length of time, materialize, implement, etc.) by processor circuitry such as a central processing unit executing instructions. Additionally or alternatively, the data usage monitoring circuitry 1300 of FIG. 13 may be instantiated (e.g., creating an instance of, bring into being for any length of time, materialize, implement, etc.) by an ASIC or an FPGA structured to perform operations corresponding to the instructions. It should be understood that some or all of the data usage monitoring circuitry 1300 of FIG. 13 may, thus, be instantiated at the same or different times. Some or all of the data usage monitoring circuitry 1300 may be instantiated, for example, in one or more threads executing concurrently on hardware and/or in series on hardware. Moreover, in some examples, some or all of the data usage monitoring circuitry 1300 of FIG. 13 may be implemented by one or more virtual machines and/or containers executing on the microprocessor.

In some examples, the data usage monitoring circuitry 1300 can be implemented by and/or otherwise included in one(s) of the endpoint data sources 160 of FIG. 1. In some examples, the data usage monitoring circuitry 1300 can be implemented by and/or otherwise included in the edge cloud 110, the central office 120, the cloud data center 130, the access point or base station 140, and/or the local processing hub 150 of FIG. 1.

In some examples, the data usage monitoring circuitry 1300 can execute and/or otherwise implement data consumption inspection logic to identify and mitigate ethical divergences of data consumption in edge environments. For example, the data usage monitoring circuitry 1300 can generate output(s) representative of determining whether an AI/ML actor (e.g., application) is using a data stream ethically or unethically. In some examples, the data usage monitoring circuitry 1300 can generate output(s) representative of baseline data patterns in nominal data streams against which to compare monitored data streams. In some examples, the data usage monitoring circuitry 1300 can generate output(s) representative of delta values between data patterns in a monitored data stream and one or more of the baseline data patterns. For example, a delta value may be a numerical value to be compared against a threshold value to determine if there is a significant enough discrepancy from a baseline data pattern to generate an ethical divergence alert. In some examples, the data usage monitoring circuitry 1300 can generate output to mitigate such ethical divergences. For example, the data usage monitoring circuitry 1300 may implement metadata tagging/tracking of data in a monitored data stream, implement a blockchain to track data in a monitored data stream, disallowing the consumption of the monitored data stream by one or more AI applications, among other mitigation efforts. In some examples, in lieu of a threshold, the data usage monitoring circuitry 1300 may implement a decision from the execution of a machine learning model in the form of one or more outputs. In some examples, the data usage monitoring circuitry 1300 may additionally provide a confidence level corresponding to the outputs (e.g., 95% confidence that there is an anomaly).

In some examples, the data usage monitoring circuitry 1300 can be a portion or a part of a larger system, environment, or collection of hardware, software, and/or firmware that can effectuate the identification and mitigation of ethical divergences. For example, the data usage monitoring circuitry 1300 can effectuate the identification and mitigation of ethical divergences by generating output(s) that mitigate unethical data consumption directly. In some examples, the data usage monitoring circuitry 1300 can effectuate the identification and mitigation of ethical divergences by generating output(s) that, when ingested as input(s) by other hardware, software, and/or firmware, can cause the other hardware, software, and/or firmware to mitigate unethical data consumption. For example, the data usage monitoring circuitry 1300 can effectuate the identification and mitigation of ethical divergences either directly or indirectly (e.g., through other hardware, software, and/or firmware). The data usage monitoring circuitry 1300 of the illustrated example includes example interface circuitry 1302, example resource manager orchestration circuitry 1304, example machine learning (ML) circuitry 1306, example metadata manager circuitry 1308, example operation execution circuitry 1310, example algorithm manager circuitry 1312, example deep data inspection circuitry 1314, example data consumption tracker circuitry 1316, example digital rights management circuitry 1318, an example datastore 1320, and an example bus 1332. The datastore 1320 of the illustrated example includes an example policy 1322, example metadata 1324, an example data graph model 1326, an example ML model A 1328, and an example ML model B 1330.

In the illustrated example of FIG. 13, the interface circuitry 1302, the resource manager orchestration circuitry 1304, the ML circuitry 1306, the metadata manager circuitry 1308, the operation execution circuitry 1310, the algorithm manager circuitry 1312, the deep data inspection circuitry 1314, and the datastore 1320 are in communication with one(s) of each other via the bus 1332. For example, the bus 1332 can be implemented by at least one of an Inter-Integrated Circuit (I2C) bus, a Serial Peripheral Interface (SPI) bus, a Peripheral Component Interconnect (PCI) bus, or a Peripheral Component Interconnect Express (PCIe or PCIE) bus. Additionally or alternatively, the bus 1332 can be implemented by any other type of computing or electrical bus.

The data usage monitoring circuitry 1300 of the illustrated example includes the interface circuitry 1302 to receive and/or ingest data that is generated and/or otherwise produced in an environment, such as the edge network environment 800 of FIG. 8. In some examples, the interface circuitry 1302 can implement the pre-processing manager 608, and/or, more generally, the data ingestion manager 606 of FIG. 6.

In some examples, the interface circuitry 1302 ingests data from a data source, such as the data sources 604 of FIG. 6. In some examples, the interface circuitry 1302 ingests data from a data source at a node, such as the logical entity 601 of FIG. 1. In some examples, the interface circuitry 1302 tags portion(s) of the data with metadata. For example, the interface circuitry 1302 can generate metadata and associate the metadata with corresponding portion(s) of the data. In some examples, the interface circuitry 1302 queries an orchestrator for an ML model, such as one(s) of the algorithms 638, the ML model A 1328, and/or the ML model B 1330, etc., that is associated with the metadata. For example, the interface circuitry 1302 can provide metadata to the resource manager/orchestration agent 642 of FIG. 6 and query the resource manager/orchestration agent 642 for one(s) of the algorithms 638 that is/are associated with the metadata.

In some examples, the interface circuitry 1302 tags portions of carries out metadata management by generating and managing metadata corresponding to a data stream and/or to an AI application node attempting to consume the data stream. For example, the interface circuitry 1302 can access a data stream and hash metadata tags (e.g., time, date, location, identification of a security stakeholder of the data in the data stream, etc.) into the data stream for tracking or other identification purposes. For example, hashed metadata tags may be injected into image data in a data stream. An example data stream can include image data that is made up of pixel data within each image in the stream as well as other run off data outside of the visible frames. The interface circuitry 1302 can inject hashed metadata tags/information into pixel data and/or into run off data associated with the image data.

The data usage monitoring circuitry 1300 of the illustrated example includes the resource manager orchestration circuitry 1304 to orchestrate resources in an edge environment based on data. In some examples, the resource manager orchestration circuitry 1304 can implement the resource manager/orchestration agent 642 of FIG. 6.

In some examples, the resource manager orchestration circuitry 1304 obtains an orchestration policy indicative of at least one of a quantity or a type of workload(s) to be executed in an edge environment. For example, the policy 1322 can be an orchestration policy created, defined, and/or otherwise generated by an organization (e.g., a business entity or company, a hospital, a government or other regulatory department, a university, etc.). In some examples, the resource manager orchestration circuitry 1304 can generate the orchestration policy to include a quantity and/or type(s) of workloads to be executed by resources associated with the organization. For example, the resource manager orchestration circuitry 1304, and/or, more generally, the organization, can instantiate the edge network environment 800 to execute workloads (e.g., acceleration, compute, network, storage, etc., workloads). In some examples, the resource manager orchestration circuitry 1304, and/or, more generally, the organization, can generate the policy 1322 to be an orchestration policy that includes data, information, parameters, etc., that define type(s) of the workloads and/or an expected number of workloads to be executed during a time period (e.g., a number of workloads to be executed per hour, day, week, month, year, etc.). In some examples, the resource manager orchestration circuitry 1304, and/or, more generally, the organization, can generate the policy 1322 to determine a number and/or type of resources (e.g., hardware, software, and/or firmware resources) to execute the quantity and/or type(s) of workloads. In some examples, the resource manager orchestration circuitry 1304, and/or, more generally, the organization, can generate the policy 1322 to define quality-of-service requirements (e.g., latency, throughput, etc., requirements), regulatory requirements, service level agreements (SLAs), etc., and/or any combination(s) thereof, that is/are to be satisfied to effectively run the organization.

In some examples, the resource manager orchestration circuitry 1304 can determine that the types of workloads include acceleration workloads such as AI/ML workloads, image or video processing, AR/VR processing, etc. In some examples, the resource manager orchestration circuitry 1304 can determine that the types of workloads include compute workloads such as sensor data processing workloads, image recognition workloads, audio recognition workloads, productivity software (e.g., database, word processing, slide presentation, spreadsheet generation, etc., software), etc. In some examples, the resource manager orchestration circuitry 1304 can determine that the types of workloads include network workloads such as receiving/transmitting data in a network, virtual resource migration (e.g., moving data or applications from a first VM or container to a second VM or container, etc.). In some examples, the resource manager orchestration circuitry 1304 can determine that the types of workloads include storage workloads such as storing data in a datastore (e.g., the distributed datastore 644), a database, etc.

In some examples, the resource manager orchestration circuitry 1304 instantiates resources in an edge environment to execute workload(s) based on an orchestration policy. For example, the resource manager orchestration circuitry 1304 can allocate and/or otherwise deploy acceleration, compute, network, security, storage, etc., resources to execute workloads in the edge network environment 800. In some examples, the resource manager orchestration circuitry 1304 generates a topology associated with the resources to at least one of execute a workload or route data in the edge environment with the resources. For example, the resource manager orchestration circuitry 1304 can generate a network topology associated with a plurality of resources by creating connections (e.g., communication connections, network connections, etc.) between one(s) of the plurality of the resources to one(s) of each other.

In some examples, the resource manager orchestration circuitry 1304 identifies one or more nodes as preferred nodes in an edge environment based on a topology. For example, the resource manager orchestration circuitry 1304 can identify one or more nodes as preferred nodes in the preferred nodes table 624 of FIG. 6. In some examples, the resource manager orchestration circuitry 1304 can identify the one or more nodes as preferred nodes to execute a workload in connection with adjacent, neighboring, and/or otherwise surrounding nodes. For example, the resource manager orchestration circuitry 1304 can identify one or more preferred nodes as super nodes to monitor data streams between data sources and one or more data stream consumers, such as one or more AI application nodes. For example, the resource manager orchestration circuitry 1304 can identify one or more preferred nodes as deep data inspection nodes to perform deep data inspection at one or more other nodes in the edge environment. For example, the resource manager orchestration circuitry 1304 can cause resources of the one or more nodes to execute workload(s). For example, the resource manager orchestration circuitry 1304 can identify the edge cloud 810 of FIG. 8 as a preferred node, which can cause the edge cloud 810 to execute an AI/ML workload (e.g., offload the AI/ML workload from a sensor, other node(s), etc.) based on video data obtained from the first monitoring sensor 828 and/or the second monitoring sensor 830. In some examples, the resource manager orchestration circuitry 1304 deploys ML model(s), such as one(s) of the algorithms 638 and/or the ML model A 1328 and/or the ML model B 1330, at a node to determine at least one of a first value of a data stream characteristic or a second value of an AI application node characteristic of data ingested at the node (or a different node), stored in the distributed datastore 644 or locally at the node (or a group of nodes that may include the node).

The data usage monitoring circuitry 1300 of the illustrated example includes the ML circuitry 1306 to execute a ML model. In some examples, the ML circuitry 1306 executes the ML model A 1328 with resources (e.g., acceleration, compute, storage, network, security, etc., resources, software resources, firmware resources, etc.) to generate outputs including at least one of a first value representative of a data stream characteristic or a second value representative of an AI application node characteristic (e.g., ingested and/or stored at a node). In some examples, the ML circuitry 1306 can implement the analytics manager 636 of FIG. 6.

In some examples, the ML circuitry 1306 executes the ML model A 1328 in a training phase or an inference phase. For example, during a training phase, the ML circuitry 1306 can obtain training data associated with an observation period. In some examples, during the training phase, the ML circuitry 1306 can execute the ML model A 1328 using the training data, including nominal network traffic data and/or nominal node behavior data, to generate outputs representative of baseline data for a data stream characteristic and/or an AI application node characteristic.

In some examples, the ML circuitry 1306 executes the ML model A 1328 in a training phase to train the machine learning model with nominal traffic (e.g., baseline network traffic) and/or nominal node behavior (e.g., baseline node behavior). In some examples, the nominal traffic is indicative of one or more nominal data streams with one or more expected data points. For example, the ML circuitry executes the ML model A 1328 in a training phase based on nominal (network) traffic ingested by the node (e.g., standard node) hosting the deployed ML model A 1328 and/or the ML model B 1330. In some examples, nominal traffic includes one or more nominal data streams to generate outputs that are indicative of values representative of a characteristic of a data stream (e.g., a nominal data stream). In some examples, the ML circuitry 1306 executes the ML model A 1328 in a training phase to train the machine learning model with one or more target data streams to be consumed by the target AI application. In some examples, the training phase to train a machine learning model with the one or more target data streams to be consumed by the target AI application takes place over a period of time to build a history of data streams sent to the target AI application node.

In some examples, the nominal node behavior is indicative of one or more expected data consumption patterns by one or more nominal nodes. In some examples, node behavior may include one or more AI application node characteristics (e.g., key indicators) that describe such behavior. For example, node behavior may include a service type attribute of the AI application node or a usage context of the application node. For example, the ML circuitry executes the ML model A 1328 in a training phase based nominal node behavior to generate outputs that are indicative of values representative of a characteristic of an AI application node (e.g., a nominal AI application node). In some examples, the ML circuitry 1306 executes the ML model A 1328 in a training phase to train the machine learning model with information indicating node behavior of the target AI application. In some examples, the training phase to train a machine learning model with information indicating node behavior of the target AI application takes place over a period of time to build a history of node behavior of the target AI application node.

In some examples, the ML circuitry 1306 executes the ML model A 1328 during a training phase for a super node. In some examples, the training phase for a super node includes a similar process to the training phase for a standard node as described above and further includes additional data streams ingested by at least one other super node. For example, the ML circuitry executes the ML model A 1328 in a training phase based a combination of nominal traffic (including one or more nominal data streams) ingested by the super node hosting the deployed ML model A 1328 as well as nominal traffic (including one or more nominal data streams) ingested by at least one other super node to generate outputs representative of a characteristic of a data stream. In some examples, nominal traffic from each super node includes one or more nominal data streams to generate outputs that are indicative of values representative of a characteristic of a data stream (e.g., a nominal data stream). In some examples, the ML circuitry executes the ML model A 1328 in a training phase based a combination of nominal traffic (including one or more nominal data streams) ingested by the super node hosting a first instantiation of the ML model A 1328 and outputs generated by the at least one other super node while executing a second instantiation of the ML model A 1328 based on nominal traffic (including one or more nominal data streams) ingested by the at least one other super node.

In some examples, the ML circuitry 1306 executes the ML model B 1330 during a training phase for a deep data inspection (DDI) node. In some examples, the algorithm manager circuitry 1312 selects a second machine learning model separate from the first machine learning model to be deployed on a DDI node. In some examples, the resource manager orchestration circuitry 1304 deploys the ML model B 1330, selected by the algorithm manager circuitry 1312, on a DDI node. In some examples, AI algorithms designated for DDI (e.g., at least one of algorithms 1-3 (638) in FIG. 6) are trained with at least one or more different features in a feature set to reflect diverse data categories and/or an environment/geographical context. In some examples, the diverse data categories and/or an environment/geographical context are identified with metadata tags in the target data stream.

In some examples, upon deployment of the DDI node, the algorithm manager circuitry 1312 selects an AI algorithm with matching features. For example, the interface circuitry 1302 may tag data within a target data stream with metadata that corresponds to one or more features (e.g., characteristics such as data stream characteristics and/or AI application node characteristics) representative of the data. In some examples, the algorithm manager circuitry 1312 may have access to a plurality of AI algorithms (e.g., algorithms 1-3 (638)), each of which may focus on a subset of an overall set of features available. Thus, in some examples, the algorithm manager circuitry 1312 may select a second algorithm used to train ML model B 1330 (as opposed to a first algorithm used to train ML model A 1328). In some examples, ML model A 1328 and ML model B 1330 are trained to different feature sets. In some examples, algorithm manager circuitry 1312 will select an ML model trained with either an algorithm that utilizes the set of features in the target data stream, or, alternatively, an algorithm that utilizes a closest feature set available to the set of features in the target data stream.

In some examples, the training phase for a DDI node includes a similar process to the training phase for a super node as described above and further includes additional data streams ingested by at least one other DDI node. For example, the ML circuitry executes the ML model B 1330 in a training phase based a combination of nominal traffic (including one or more nominal data streams) ingested by the DDI node hosting the deployed ML model B 1330 as well as nominal traffic (including one or more nominal data streams) ingested by at least one other DDI node to generate outputs representative of a characteristic of a data stream. In some examples, nominal traffic from each DDI node includes one or more nominal data streams to generate outputs that are indicative of values representative of a characteristic of a data stream (e.g., a nominal data stream). In some examples, the ML circuitry executes the ML model B 1330 in a training phase based a combination of nominal traffic (including one or more nominal data streams) ingested by the DDI node hosting a first instantiation of the ML model B 1330 and outputs generated by the at least one other DDI node while executing a second instantiation of the ML model B 1330 based on nominal traffic (including one or more nominal data streams) ingested by the at least one other super DDI.

In some examples, the ML circuitry 1306 determines to execute the ML model A 1328 during an inference phase. For example, the ML circuitry 1306 can provide ingested data, or portion(s) thereof, to the ML model A 1328 as inputs (e.g., data inputs, model inputs, etc.) to generate outputs (e.g., data outputs, model outputs, etc.), which can be decisions, determinations, insights, etc. For example, the ML circuitry 1306 can execute the ML model A 1328 with ingested data to generate an output, which can be indicative and/or otherwise representative of a decision or determination to update baseline data based on the ingested data. In some examples, the ML circuitry 1306 can determine to update portion(s) of baseline data, such as change, update, and/or otherwise adjust metadata associated with baseline data. For example, the ML circuitry 1306 can update metadata associated with baseline data by changing a first value representative of a data stream characteristic, a second value representative of an AI application characteristic, etc., that can be included in the metadata.

In some examples, the ML circuitry 1306 can tag ingested data to undergo identification and mitigation of ethical divergence in data consumption operations. For example, the ML circuitry 1306, in response to a determination not to update baseline data in view of an output of the ML model A 1328 based on ingested data, can determine to annotate, assign, and/or otherwise tag the ingested data for one or more identification and mitigation of ethical divergence in data consumption operations. In some examples, the interface circuitry 1302 add metadata to the ingested data that, in response to being added, can invoke the operation execution circuitry 1310 or the deep data inspection circuitry 1314 to process the ingested data. For example, the operation execution circuitry 1310 can prohibit (e.g., disallow) the ingested data, or portion(s) thereof, from being consumed by one or more data consumers, such as one or more AI application nodes.

In some examples, the ML circuitry 1306 determines a value representative a data stream characteristic based on at least one of training data or ingested data. For example, the ML circuitry 1306 can execute the ML model A 1328 to determine a value representative a data stream characteristic based on training data (e.g., using training data during a training phase) or inference data (e.g., using ingested data during an inference phase). In some examples, the ML circuitry 1306 can execute the ML model A 1328 with one or more inputs, such as at least one of a content type of a data stream, a sensitive attribute of a data stream, a security level of a data stream, or a source location of a data stream, among other possible data stream characteristics. In some examples, the ML circuitry 1306 can execute the ML model A 1328 with the one or more inputs to determine a value representative of the data stream characteristic.

In some examples, the ML circuitry 1306 determines a value of a characteristic of the consumer of the data (e.g., an AI application node characteristic) based on at least one of training data or ingested data. For example, the ML circuitry 1306 can execute the ML model A 1328 to determine a value representative of a characteristic of the AI application node based on training data (e.g., using training data during a training phase) or inference data (e.g., using ingested data during an inference phase). In some examples, the ML circuitry 1306 can execute the ML model A 1328 with one or more inputs, such as a service type attribute of the AI application node, a usage context of a data stream by the AI application node, or any one or more other AI application node characteristics. For example, the ML circuitry 1306 can execute the ML model A 1328 with the one or more inputs to generate an output, which can include a determination of a value representative of a characteristic of the AI application node based on the one or more inputs.

In some examples, the ML circuitry 1306 can execute the ML model A 1328 to determine a content type characteristic of data in the data stream (e.g., ingested or stored data at a node). For example, a content type of the data stream may include a characteristic classifying the content within the data stream as image data, audio data, textual data, telemetry data (e.g., sensor data, etc.), or any other type of data.

In some examples, the ML circuitry 1306 can execute the ML model A 1328 to determine a sensitive attribute characteristic of data in the data stream (e.g., ingested or stored data at a node). For example, a sensitive attribute of the data stream may include a characteristic that provides/describes sensitive topics, such as identification information of individuals (e.g., human resources data), classification of individuals based on race, gender, or other or any other type of classification, information about financial statements, governmental records, confidential records, or one or more other sensitive attributes.

In some examples, the ML circuitry 1306 can execute the ML model A 1328 to determine a security level characteristic of data in the data stream (e.g., ingested or stored data at a node). For example, a security level of a data stream may include a level of known confidentiality based on content type, sensitive attributes, etc. such as confidential data, restricted data, top secret data, etc.

In some examples, the ML circuitry 1306 can execute the ML model A 1328 to determine a source location characteristic of data in the data stream (e.g., ingested or stored data at a node). For example, a source location of a data stream may be described by an Internet Protocol (IP) address, a physical address, or another type of address that corresponds to a virtual or geographic location that may change how data is viewed. For example, a virtual source location IP address from a bank may cause heightened ethical scrutiny for a data stream with financial data. For example, a geographical source location at a military facility may cause heighted ethical scrutiny for image data. For example, a geographical source location in a recording studio may cause heighted ethical scrutiny for audio data.

In some examples, the ML circuitry 1306 can execute the ML model A 1328 to determine a service type attribute characteristic of the consumer of data (e.g., the AI application node) in the data stream (e.g., ingested or stored data at a node). For example, a service type attribute may include a usage model of the AI application or service being performed by the AI application node. Examples of usage models of the AI application node include smart shelves in retail stores, monitoring high-risk intersections in cities, automated identification of individuals in airports or secure government buildings, predictive maintenance of equipment in factories, automated health screening of individuals, among a myriad of other service types of usage models. Thus, in some examples, the AI application node is classified as having one or more service types by labeling it as such with the service type attribute.

In some examples, the ML circuitry 1306 can execute the ML model A 1328 to determine a usage context characteristic of the consumer of data (e.g., the AI application node) in the data stream (e.g., ingested or stored data at a node). For example, a usage context of the data in the data stream may include characteristic information such as the monitoring of products consumption status for smart shelves in retail stores, the monitoring of the density of vehicles and pedestrians cross high-risk intersections in cities, identifying persons traveling or security records clearance for the automated identification of individuals in airports or secure government buildings, the continuous check of machines' health by a factory management system for the predictive maintenance of equipment in factories, or the fast-checking of a pre-visit medical visit for automated health screening of individuals, among other usage contexts.

The data usage monitoring circuitry 1300 of the illustrated example includes the metadata manager circuitry 1308 to manage metadata related to a data stream. In some examples, the metadata manager circuitry 1308 may aggregate, combine, and/or otherwise merge data stored at one or more nodes by instantiating and/or generating the data graph model 1326. In some examples, the metadata manager circuitry 1308 can reduce the quantity of data by replacing portion(s) of the data itself with metadata that can be expressed by the data graph model 1326. In some examples, the metadata manager circuitry 1308 can implement the metadata/data enrichment manager 640 of FIG. 6.

The data usage monitoring circuitry 1300 of the illustrated example includes the operation execution circuitry 1310 to cause operation(s) at node(s) of an edge environment based on data (e.g., data ingested, processed, and/or stored at one or more nodes). In some examples, the operation execution circuitry 1310 can implement the data ingestion manager 606 of FIG. 6. In some examples, the operation execution circuitry 1310 can invoke the ML circuitry 1306 to execute ML model A 1328 and/or ML model B 1330 to generate output(s) based on input, such as ingested data (e.g., sensor data or any other type of data generated or produced in a network environment). For example, the output(s) can include at least one of first output(s) representative of object detection (e.g., a detection of a person, animal, device, object, etc.), second output(s) representative of object classification (e.g., an identification and/or a classification of a type and/or description of a person, animal, device, object, etc.), third output(s) representative of object tracking (e.g., maintain an object identifier in addition to a motion and/or velocity vector that defines a path of motion of a person, animal, device, object, etc.), fourth output(s) representative of key performance indicators (KPIs) or other parameters or metrics, or fifth output(s) representative of parameters or metrics associated with an environment based on the policy 1322. In some examples, the operation execution circuitry 1310 can generate a command, direction, an instruction, etc., that, when obtained by a node, causes the node to carry out an operation at the node based on at least one of the first output(s), the second output(s), the third output(s), the fourth output(s), or the fifth output(s). For example, the operation execution circuitry 1310 can command an autonomous guided vehicle to avoid an object in response to a detection of the object based on sensor data of the autonomous guided vehicle or a different node associated with the autonomous guided vehicle. In some examples, the operation execution circuitry 1310 generates an alert and transmits the alert to the node or different node(s) to cause the operation(s) to be executed.

The data usage monitoring circuitry 1300 of the illustrated example includes the algorithm manager circuitry 1312 to select an AI algorithm (e.g., AI algorithms 1-3 (638 in FIG. 6)) used to train an ML model, such as ML model A 1328 and/or ML model B 1330. In some examples, the algorithm manager circuitry 1312 can invoke the ML circuitry 1306 to execute ML model A 1328 and/or ML model B 1330 to generate outputs based on a target data stream to determine a set of features (e.g., characteristics) within the target data stream. In some examples, the algorithm manager circuitry 1312 select an AI algorithm that is representative of a set of features in the target data stream. In some examples, the selected AI algorithm may correspond to a specific ML model, such as ML model A 1328 or ML model B 1330. In some examples, the algorithm manager circuitry 1312 can implement the algorithm manager/recommender 634 of FIG. 6.

The data usage monitoring circuitry 1300 of the illustrated example includes the deep data inspection (DDI) circuitry 1314 to monitor data usage at a node. In some examples, the resource manager orchestration circuitry 1304 may deploy/instantiate a DDI node to effectuate identification and/or mitigation of ethical divergence of monitored data usage. In some examples, the DDI circuitry 1314 can implement the data ingestion manager 606, the data query manager 610, and/or the node manager 622 of FIG. 6. In some examples, the DDI circuitry 1314 can examine data in a data stream by way of lawful inspection upon ingest and/or upon query. In some examples, the DDI circuitry 1314 can invoke an ML model, such as ML model B 1330, to generate outputs to compare characteristics of a target data stream to characteristics of one or more nominal/baseline data streams. In some examples, the DDI circuitry 1314 may share data streams (e.g., nominal data streams) and/or outputs generated from a ML model (e.g., ML model B 1330) with at least one other DDI node. In some examples, the DDI circuitry 1314 may obtain data streams (e.g., nominal data streams) and/or outputs generated from at least one ML model instantiated at another DDI node. For example, multiple DDI nodes may share data and or outputs associated with a plurality of nominal data streams to create one or more consensus data patterns corresponding to (affiliated with) the plurality of nominal data streams. In some examples, DDI circuitry 1314 causes the learning and designing of one or more data patterns to compare and contrast nominal data versus non-nominal data.

The data usage monitoring circuitry 1300 of the illustrated example includes data consumption tracker circuitry 1316 to track data consumption in an edge environment. In some examples, the data consumption tracker circuitry 1316 can implement the data ingestion manager 606 of FIG. 6. In some examples, the data consumption tracker circuitry 1316 tracks data by filtering a target data stream to obtain metadata. In some examples, the data consumption tracker circuitry 1316 analyzes the target AI application node behavior corresponding to obtained metadata. In some examples, the data consumption tracker circuitry 1316 adds to a metadata count and a target AI application behavior count for obtained metadata in the target data stream and the corresponding behavior. In some examples, the data consumption tracker circuitry 1316 generates an alert to be processed by the data usage monitoring circuitry 1300, such as in a designated alert processing node, when the metadata count and a target AI application behavior count for obtained metadata satisfies a threshold value. In some examples, in lieu of a threshold, the data usage monitoring circuitry 1300 may implement a decision from the execution of a machine learning model in the form of one or more outputs. In some examples, the data usage monitoring circuitry 1300 may additionally provide a confidence level corresponding to the outputs (e.g., 85% confidence that an ethical divergence has been detected).

The data usage monitoring circuitry 1300 of the illustrated example includes digital rights management (DRM) circuitry 1318 to manage the digital rights of data streams. In some examples, the DRM circuitry 1318 determines a DRM policy 1322 to manage target data stream. In some examples, the DRM circuitry 1318 can implement the node manager 622 and/or the data security manager 632 of FIG. 6. In some examples, the DRM policies provides details on which entities within the edge environment have access rights (and what level access rights) to the target data stream (e.g., the data in the target data stream). In some examples, the DRM circuitry 1318 may employ one or more mitigation techniques in case a target AI application node exhibits an ethical divergence in its consumption efforts of the target data stream when compared against other AI application nodes consuming other data streams (e.g., nominal/baseline data streams). For example, the DRM circuitry 1318 may A) tag data in the data stream with metadata for tracking the data, B) hash tracking information into the data in the target data stream (e.g., hash time/date/location/ID information into the pixel data of frames in a video stream), C) implement a blockchain for data in the target data stream, and/or D) prohibit consumption of the target data stream by the target AI application node, among other possible mitigation techniques.

The data usage monitoring circuitry 1300 of the illustrated example includes the datastore 1320 to record data, such as the policy 1322 the metadata 1324, the data graph model 1326, the ML model A 1328, the ML model B 1330, and/or the like. In some examples, the ML model A 1328 and/or the ML model B 1330 is each a neural network model. Additionally and/or alternatively, the ML model A 1328 and/or the ML model B 1330 may be any other type of AI/ML model. In some examples, the policy 1322 is representative of, corresponds to, and/or otherwise includes intents (or intentions), goals, objectives, targets, etc. For example, the policy 1322 can be generated by an ML model, such as the ML model A 1328 and/or the ML model B 1330, or by a user (e.g., an IT manager, an HR manager, a developer, a system architect, etc.) to carry out intentions on how data is to be ingested, stored, and/or otherwise processed to achieve reduced environment impacts. In some examples, the policy 1322 can include data (e.g., data objects, metadata, etc.) that, when analyzed by the data usage monitoring circuitry 1300, can carry out operations to effectuate, facilitate, and/or otherwise carry out identification and mitigation of the ethical divergence of a data stream in accordance with the intentions or desires of the creator of the policy 1322.

In some examples, the interface circuitry 1302 can determine that an intention of the policy 1322 includes techniques or processes to ingest data from the data sources 604 with ethical divergence (e.g., tag the ingested data with metadata to indicate that the ingested data is limited or prohibited, etc.). In some examples, the resource manager orchestration circuitry 1304 can identify an intention of the policy 1322 as representative of how the resource manager orchestration circuitry 1304 is to orchestrate and/or otherwise instantiate resources in the edge network environment 800 to achieve identification and mitigation of ethical divergence in data consumption. In some examples, the ML circuitry 1306 can provide intentions or desires of the policy 1322 as input(s) to the ML model 1326 to generate output(s), which can include actions, determinations, etc., to cause nodes in the edge network environment 800 to operate with reduced environment impact.

In some examples, the datastore 1320 can be implemented by a volatile memory (e.g., a Synchronous Dynamic Random Access Memory (SDRAM), Dynamic Random Access Memory (DRAM), RAMBUS Dynamic Random Access Memory (RDRAM), etc.) and/or a non-volatile memory (e.g., flash memory). The datastore 1320 may additionally or alternatively be implemented by one or more double data rate (DDR) memories, such as DDR, DDR2, DDR3, DDR4, DDR5, mobile DDR (mDDR), DDR SDRAM, etc. The datastore 1320 may additionally or alternatively be implemented by one or more mass storage devices such as hard disk drive(s) (HDD(s)), compact disk (CD) drive(s), digital versatile disk (DVD) drive(s), solid-state disk (SSD) drive(s), Secure Digital (SD) card(s), CompactFlash (CF) card(s), etc. While in the illustrated example the datastore 1320 is illustrated as a single datastore, the datastore 1320 may be implemented by any number and/or type(s) of databases. Furthermore, the data stored in the datastore 1320 may be in any data format such as, for example, binary data, comma delimited data, tab delimited data, structured query language (SQL) structures, etc. In some examples, the datastore 1320 can implement one or more databases of data. The term “database” as used herein means an organized body of related data, regardless of the manner in which the data or the organized body thereof is represented. For example, the organized body of related data may be in the form of one or more of a table, a map, a grid, a packet, a datagram, a frame, a file, an e-mail, a message, a document, a report, a list or in any other form.

In some examples, the interface circuitry 1302 is instantiated by processor circuitry executing interface circuitry instructions and/or configured to perform operations such as those represented by the flowcharts of FIGS. 17, 18, 24, and 25.

In some examples, the apparatus includes ingesting data from a data source. For example, the means for ingesting may be implemented by interface circuitry 1302. In some examples, the interface circuitry 1302 may be instantiated by processor circuitry such as the example processor circuitry 3012 of FIG. 30. For instance, the interface circuitry 1302 may be instantiated by the example microprocessor 3200 of FIG. 32 executing machine executable instructions such as those implemented by at least blocks 1702 of FIG. 17, 1802, 1804, 1806 of FIG. 18, 2404 of FIGS. 24, and 2504 of FIG. 25. In some examples, the interface circuitry 1302 may be instantiated by hardware logic circuitry, which may be implemented by an ASIC, XPU, or the FPGA circuitry 3300 of FIG. 33 structured to perform operations corresponding to the machine readable instructions. Additionally or alternatively, the interface circuitry 1302 may be instantiated by any other combination of hardware, software, and/or firmware. For example, the interface circuitry 1302 may be implemented by at least one or more hardware circuits (e.g., processor circuitry, discrete and/or integrated analog and/or digital circuitry, an FPGA, an ASIC, an XPU, a comparator, an operational-amplifier (op-amp), a logic circuit, etc.) structured to execute some or all of the machine readable instructions and/or to perform some or all of the operations corresponding to the machine readable instructions without executing software or firmware, but other structures are likewise appropriate.

In some examples, the means for ingesting is to ingest data from multiple ones (e.g., a plurality of ones) of data sources at one or more nodes, the multiple ones of the data sources including a data source. In some examples, the means for ingesting is to tag a portion of the data with metadata. In some examples, the means for ingesting is to query an orchestrator to identify a machine learning model as associated with metadata. In some examples, the means for ingesting is to execute the machine learning model at the one or more nodes to determine at least one of a first value of a characteristic of a data stream or a second value of a characteristic of an AI application node.

In some examples, the resource manager orchestration circuitry 1304 is instantiated by processor circuitry executing resource manager orchestration circuitry instructions and/or configured to perform operations such as those represented by the flowcharts of FIGS. 17, 19, 24, and 25.

In some examples, the apparatus includes means orchestrating resources in an edge environment based on data ingested from one or more data sources. For example, the means for orchestrating may be implemented by resource manager orchestration circuitry 1304. In some examples, the resource manager orchestration circuitry 1304 may be instantiated by processor circuitry such as the example processor circuitry 3012 of FIG. 30. For instance, the resource manager orchestration circuitry 1304 may be instantiated by the example microprocessor 3200 of FIG. 32 executing machine executable instructions such as those implemented by at least blocks 1704 of FIG. 17, 1902, 1904, 1906, 1908, 1910 of FIG. 19, 2402 of FIG. 24, 2502 of FIG. 25. In some examples, the resource manager orchestration circuitry 1304 may be instantiated by hardware logic circuitry, which may be implemented by an ASIC, XPU, or the FPGA circuitry 3300 of FIG. 33 structured to perform operations corresponding to the machine readable instructions. Additionally or alternatively, the resource manager orchestration circuitry 1304 may be instantiated by any other combination of hardware, software, and/or firmware. For example, the resource manager orchestration circuitry 1304 may be implemented by at least one or more hardware circuits (e.g., processor circuitry, discrete and/or integrated analog and/or digital circuitry, an FPGA, an ASIC, an XPU, a comparator, an operational-amplifier (op-amp), a logic circuit, etc.) structured to execute some or all of the machine readable instructions and/or to perform some or all of the operations corresponding to the machine readable instructions without executing software or firmware, but other structures are likewise appropriate.

In some examples, the means for orchestrating is to obtain an orchestration policy indicative of at least one of a quantity or a type of workloads to be executed in an edge environment. In some examples, the means for orchestrating is to instantiate resources in the edge environment to execute a workload based on the orchestration policy, the resources including at least one of compute resources or network resources. In some examples, the means for orchestrating is to generate a topology associated with the resources to at least one of execute the workload with one or more of the compute resources or route data in the edge environment with one or more of the network resources. In some examples, the means for orchestrating is to identify one or more nodes as one or more preferred nodes in the edge environment based on the topology, the one or more preferred nodes to generate local determinations associated with the data. In some examples, the means for orchestrating is to deploy a machine learning model to the one or more nodes in response to an identification of the one or more nodes as the one or more preferred nodes.

In some examples, the ML circuitry 1306 is instantiated by processor circuitry executing ML circuitry instructions and/or configured to perform operations such as those represented by the flowcharts of FIGS. 17, 18, 20, 21, 22, 23, 24, 25, and 26.

In some examples, the apparatus includes means executing a machine learning (ML) model based on data to generate outputs. In some examples, the outputs include at least one of a first value representative of a characteristic of a data stream or a second value representative of a characteristic of an AI application node. For example, the means for executing may be implemented by ML circuitry 1306. In some examples, the ML circuitry 1306 may be instantiated by processor circuitry such as the example processor circuitry 3012 of FIG. 30. For instance, the ML circuitry 1306 may be instantiated by the example microprocessor 3200 of FIG. 32 executing machine executable instructions such as those implemented by at least blocks 1706 of FIG. 17, blocks 1808 of FIG. 18, blocks 2002, 2004, 2006, 2008, 2010, 2012, 2014 of FIG. 20, blocks 2102, 2104, 2106, 2108, 2110 of FIG. 21, blocks 2202, 2204, 2206, 2208, 2210 of FIG. 22, blocks 2302, 2304, 2306 of FIG. 23, 2406, 2408, 2410, 2412 of FIG. 24, block 2508 of FIG. 25, and block 2604 of FIG. 26. In some examples, the ML circuitry 1306 may be instantiated by hardware logic circuitry, which may be implemented by an ASIC, XPU, or the FPGA circuitry 3300 of FIG. 33 structured to perform operations corresponding to the machine readable instructions. Additionally or alternatively, the ML circuitry 1306 may be instantiated by any other combination of hardware, software, and/or firmware. For example, the ML circuitry 1306 may be implemented by at least one or more hardware circuits (e.g., processor circuitry, discrete and/or integrated analog and/or digital circuitry, an FPGA, an ASIC, an XPU, a comparator, an operational-amplifier (op-amp), a logic circuit, etc.) structured to execute some or all of the machine readable instructions and/or to perform some or all of the operations corresponding to the machine readable instructions without executing software or firmware, but other structures are likewise appropriate.

In some examples, the means for executing is to determine a first value representative of a characteristic of a data stream that includes at least one of a content type of the target data stream, a sensitive attribute of the target data stream, a security level of the target data stream, or a source location of the target data stream. The means for executing is to determine a second value representative of a characteristic of an AI application node that includes at least one of a service type attribute or a usage context.

In some examples, the metadata manager circuitry 1308 is instantiated by processor circuitry executing metadata manager circuitry instructions and/or configured to perform operations such as those represented by the flowchart of FIG. 26.

In some examples, the apparatus includes means for generating at least one target graph node representation of the target data stream. For example, the means for generating may be implemented by metadata manager circuitry 1308. In some examples, the metadata manager circuitry 1308 may be instantiated by processor circuitry such as the example processor circuitry 3012 of FIG. 30. For instance, the metadata manager circuitry 1308 may be instantiated by the example microprocessor 3200 of FIG. 32 executing machine executable instructions such as those implemented by at least blocks 2602 of FIG. 26. In some examples, the metadata manager circuitry 1308 may be instantiated by hardware logic circuitry, which may be implemented by an ASIC, XPU, or the FPGA circuitry 3300 of FIG. 33 structured to perform operations corresponding to the machine readable instructions. Additionally or alternatively, the metadata manager circuitry 1308 may be instantiated by any other combination of hardware, software, and/or firmware. For example, the metadata manager circuitry 1308 may be implemented by at least one or more hardware circuits (e.g., processor circuitry, discrete and/or integrated analog and/or digital circuitry, an FPGA, an ASIC, an XPU, a comparator, an operational-amplifier (op-amp), a logic circuit, etc.) structured to execute some or all of the machine readable instructions and/or to perform some or all of the operations corresponding to the machine readable instructions without executing software or firmware, but other structures are likewise appropriate.

In some examples, the operation execution circuitry 1310 is instantiated by processor circuitry executing operation execution circuitry instructions and/or configured to perform operations such as those represented by the flowcharts of FIGS. 17 and 27.

In some examples, the apparatus includes means for causing an operation at a node of an edge environment based on at least one of data (e.g., ingested data) or outputs (e.g., outputs from an ML model). In some examples, the data is associated with the node. For example, the means for causing may be implemented by operation execution circuitry 1310. In some examples, the operation execution circuitry 1310 may be instantiated by processor circuitry such as the example processor circuitry 3012 of FIG. 30. For instance, the operation execution circuitry 1310 may be instantiated by the example microprocessor 3200 of FIG. 32 executing machine executable instructions such as those implemented by at least blocks 1710 of FIG. 17, and blocks 2702 and 2704 of FIG. 27. In some examples, the operation execution circuitry 1310 may be instantiated by hardware logic circuitry, which may be implemented by an ASIC, XPU, or the FPGA circuitry 3300 of FIG. 33 structured to perform operations corresponding to the machine readable instructions. Additionally or alternatively, the operation execution circuitry 1310 may be instantiated by any other combination of hardware, software, and/or firmware. For example, the operation execution circuitry 1310 may be implemented by at least one or more hardware circuits (e.g., processor circuitry, discrete and/or integrated analog and/or digital circuitry, an FPGA, an ASIC, an XPU, a comparator, an operational-amplifier (op-amp), a logic circuit, etc.) structured to execute some or all of the machine readable instructions and/or to perform some or all of the operations corresponding to the machine readable instructions without executing software or firmware, but other structures are likewise appropriate.

In some examples, the algorithm manager circuitry 1312 is instantiated by processor circuitry executing analytics manager circuitry instructions and/or configured to perform operations such as those represented by the flowchart of FIG. 25.

In some examples, the apparatus includes means selecting a machine learning model trained on a feature set. For example, the means for selecting may be implemented by algorithm manager circuitry 1312. In some examples, the algorithm manager circuitry 1312 may be instantiated by processor circuitry such as the example processor circuitry 3012 of FIG. 30. For instance, the algorithm manager circuitry 1312 may be instantiated by the example microprocessor 3200 of FIG. 32 executing machine executable instructions such as those implemented by at least blocks 2506 of FIG. 25. In some examples, the algorithm manager circuitry 1312 may be instantiated by hardware logic circuitry, which may be implemented by an ASIC, XPU, or the FPGA circuitry 3300 of FIG. 33 structured to perform operations corresponding to the machine readable instructions. Additionally or alternatively, the algorithm manager circuitry 1312 may be instantiated by any other combination of hardware, software, and/or firmware. For example, the algorithm manager circuitry 1312 may be implemented by at least one or more hardware circuits (e.g., processor circuitry, discrete and/or integrated analog and/or digital circuitry, an FPGA, an ASIC, an XPU, a comparator, an operational-amplifier (op-amp), a logic circuit, etc.) structured to execute some or all of the machine readable instructions and/or to perform some or all of the operations corresponding to the machine readable instructions without executing software or firmware, but other structures are likewise appropriate.

In some examples, the deep data inspection circuitry 1314 is instantiated by processor circuitry executing deep data inspection circuitry instructions and/or configured to perform operations such as those represented by the flowchart of FIG. 25.

In some examples, the apparatus includes means determining second outputs satisfy a second threshold value. For example, the means for determining may be implemented by deep data inspection circuitry 1314. In some examples, the deep data inspection circuitry 1314 may be instantiated by processor circuitry such as the example processor circuitry 3012 of FIG. 30. For instance, the deep data inspection circuitry 1314 may be instantiated by the example microprocessor 3200 of FIG. 32 executing machine executable instructions such as those implemented by at least blocks 2510, 2512, 2514, 2516, 2518 of FIG. 25. In some examples, the deep data inspection circuitry 1314 may be instantiated by hardware logic circuitry, which may be implemented by an ASIC, XPU, or the FPGA circuitry 3300 of FIG. 33 structured to perform operations corresponding to the machine readable instructions. Additionally or alternatively, the deep data inspection circuitry 1314 may be instantiated by any other combination of hardware, software, and/or firmware. For example, the deep data inspection circuitry 1314 may be implemented by at least one or more hardware circuits (e.g., processor circuitry, discrete and/or integrated analog and/or digital circuitry, an FPGA, an ASIC, an XPU, a comparator, an operational-amplifier (op-amp), a logic circuit, etc.) structured to execute some or all of the machine readable instructions and/or to perform some or all of the operations corresponding to the machine readable instructions without executing software or firmware, but other structures are likewise appropriate.

In some examples, the means for determining is to determine whether ethical divergence of a target data stream occurs greater than a frequency threshold. In some examples, the means for determining is to determine whether ethical divergence of occurs at other DDI nodes greater than a frequency threshold. In some examples, the means for determining is to determine whether ethical divergence at target data stream meets a constraint.

In some examples, the data consumption tracker circuitry 1316 is instantiated by processor circuitry executing data consumption tracker circuitry instructions and/or configured to perform operations such as those represented by the flowcharts of FIGS. 24, 25, and 28.

In some examples, the apparatus includes means for tracking data consumption. For example, the means for tracking may be implemented by data consumption tracker circuitry 1316. In some examples, the data consumption tracker circuitry 1316 may be instantiated by processor circuitry such as the example processor circuitry 3012 of FIG. 30. For instance, the data consumption tracker circuitry 1316 may be instantiated by the example microprocessor 3200 of FIG. 32 executing machine executable instructions such as those implemented by at least block 2412 of FIG. 24, 2512 of FIGS. 25, and 2802, 2804, 2806, 2808, 2810 of FIG. 28. In some examples, the data consumption tracker circuitry 1316 may be instantiated by hardware logic circuitry, which may be implemented by an ASIC, XPU, or the FPGA circuitry 3300 of FIG. 33 structured to perform operations corresponding to the machine readable instructions. Additionally or alternatively, the data consumption tracker circuitry 1316 may be instantiated by any other combination of hardware, software, and/or firmware. For example, the data consumption tracker circuitry 1316 may be implemented by at least one or more hardware circuits (e.g., processor circuitry, discrete and/or integrated analog and/or digital circuitry, an FPGA, an ASIC, an XPU, a comparator, an operational-amplifier (op-amp), a logic circuit, etc.) structured to execute some or all of the machine readable instructions and/or to perform some or all of the operations corresponding to the machine readable instructions without executing software or firmware, but other structures are likewise appropriate.

In some examples, the digital rights management circuitry 1318 is instantiated by processor circuitry executing deep data inspection circuitry instructions and/or configured to perform operations such as those represented by the flowcharts of FIGS. 25 and 29.

In some examples, the apparatus includes means managing digital rights of a target data stream. For example, the means for managing may be implemented by digital rights management circuitry 1318. In some examples, the digital rights management circuitry 1318 may be instantiated by processor circuitry such as the example processor circuitry 3012 of FIG. 30. For instance, the digital rights management circuitry 1318 may be instantiated by the example microprocessor 3200 of FIG. 32 executing machine executable instructions such as those implemented by at least block 2522 of FIG. 25 and blocks 2902, 2904, 2906, and 2908 of FIG. 29. In some examples, the digital rights management circuitry 1318 may be instantiated by hardware logic circuitry, which may be implemented by an ASIC, XPU, or the FPGA circuitry 3300 of FIG. 33 structured to perform operations corresponding to the machine readable instructions. Additionally or alternatively, the digital rights management circuitry 1318 may be instantiated by any other combination of hardware, software, and/or firmware. For example, the deep digital rights management circuitry 1318 may be implemented by at least one or more hardware circuits (e.g., processor circuitry, discrete and/or integrated analog and/or digital circuitry, an FPGA, an ASIC, an XPU, a comparator, an operational-amplifier (op-amp), a logic circuit, etc.) structured to execute some or all of the machine readable instructions and/or to perform some or all of the operations corresponding to the machine readable instructions without executing software or firmware, but other structures are likewise appropriate.

FIG. 14 is an illustration of an example ADM system 1400. In the illustrated example, the ADM system 1400 includes the analytics manager 636 of FIG. 6, which implements example Deep Data Inspection (DDI) modules 1402A/B, 1404A/B, and 1406A/B at example nodes 1408, 1410, and 1412. For example, the ADM system 1400 of the illustrated example may be implemented by the ADM system 600 of FIG. 6. In some examples, the nodes 1408, 1410, and 1412 may implement one(s) of the data sources/producers 604 of FIG. 6.

In the illustrated example, the nodes 1408, 1410, and 1412 may ingest data, which may be representative of a data query (e.g., data query 1414, 1416, and 1418). The nodes 1408, 1410, and 1412 may provide the data query 1414, 1416, and 1418 to the data query manager 610 of FIG. 6. The data query manager 610 may process the data query 1414, 1416, and 1418 by retrieving data from datastores 1420, 1422, and 1424 and return the results to the nodes 1408, 1410, and 1412, which may provide the results back to the data requester.

In the illustrated example, the analytics manager 636 may execute DDI via the DDI “A” modules 1402A, 1404A, and 1406A on the ingested data 1426, 1428, and 1430. For example, the analytics manager 636 may execute one(s) of the algorithms 638 to examine data by way of lawful inspection upon ingest and/or upon query, to the data query manager 610 to analyze the data content and decide whether the data content meets and/or otherwise satisfies the ethical data norm, matches the nominal data/traffic across other nodes or over time, and/or violates ethical data policies established by the owner(s) of the system whose ethical values have been captured as conditional statements, boundaries, or rules (built from data attributes) a priori as input to the system. In some examples, the DDI modules 1402A/B, 1404A/B, and 1406A/B across the nodes 1408, 1410, and 1412 may trigger federated learning between the nodes 1408, 1410, and 1412 through info sharing 1432 and 1434 across the nodes 1408, 1410, and 1412 and info sharing 1436 with an example central node or cloud 1438 to augment the scope of learning and designing a pattern for nominal data versus non-nominal data.

In some examples, the central node and/or cloud 1438 may be implemented by portion(s) of the ADM system 600 of FIG. 6. In some examples, the central node and/or cloud 1438 may be implemented by the cloud data center 130 of FIG. 1, the cloud data center 245 of FIG. 2, the cloud/data center 360 of FIG. 3, the cloud/data center 440 of FIG. 4, etc.

In the illustrated example of FIG. 14, the central node and/or cloud 1438 provides causes the management and/or coalescing of data/information exchange/sharing among nodes 1408, 1410, and 1412, including nodes 1, 2, through node n. In some examples, the central node and/or cloud 1438 shares such data/information with analytics manager 636 of FIG. 6. In some examples, the analytics manager 636 may execute and/or otherwise implement offline learning and insights on shared data from one or more DDI instantiations, such as DDI 1402A/B, 1404A/B, and 1406A/B. For example, the analytics manager 636 may execute one(s) of the algorithms 638 of FIG. 6 on the data, statuses, information, etc., from the central node and/or cloud 1438.

In some examples, from a deployment perspective, DDI may function on clear data and/or encrypted data and consider several data categories to identify features in the content of data queried by an AI/ML algorithm or consumed by an AI/ML algorithm.

Example DDI functions with encrypted data may include techniques such as Homomorphic Encryption or Searchable Encryption. Example Data Categories that DDI may detect may include people specific data, data on person's presence, sensitive data in industrial environment, etc.

In some examples, the AI/ML algorithms for DDI may be trained with different set(s) of features to reflect the diverse data categories and the environment/geo context. Upon deployment of the DDI, the service may choose the AI/ML algorithms trained with the matching features (e.g., for AI algorithm 1, it can have two flavors, flavor 1(a) trained with the feature set a and flavor 1(b) trained with the feature set b, and upon deployment the service will on-board the algorithm flavor matching the environment context and geo). In some examples, if any rules/policies are being changed, the AI/ML algorithm may be retrained with additional feature set to have a new/updated flavor. In some examples, an adversarial avoidance module, which may be implemented by one(s) of the algorithms 638, and/or, more generally, the analytics manager 636, may learn the defect pattern or quality pattern over a period through DDI. In some such examples, any change from the nominal level may be flagged and taken into consideration for model update or hyper parameter optimization during the continuous learning cycle. In some such examples, this check, verification, etc., may be particularly applicable in fully autonomous industrial systems to securely update models.

FIG. 15 is an illustration of an example ADM system 1400 implementing super nodes. In some examples, super nodes may be deployed by inspecting data streams that are intended for consumption by AI application nodes. In the illustrated example, the ADM system 1500 includes a network plane 1502, which communicatively couples (e.g., allows data transmission) between one or more data sources 1504 and one or more AI application nodes, such as AI application node 1 (1506), AI application node 2 (1508), and AI application node 3 (1510). In some examples, data sources 1504 may be implemented by data sources 604 in FIG. 6. In some examples, resource manager orchestration circuitry 1304 orchestrates resources in the edge environment to deploy one or more super nodes, such as super node A 1512 and/or super node B 1514.

The example data usage monitoring circuitry 1300 may be alerted that an ethical divergence is occurring with one or more data streams being sourced by data sources 1504. In such circumstances, the example data usage monitoring circuitry 1300 may cause the resource manager orchestration circuitry 1304 to deploy one or more super nodes to monitor such questionable data streams. In some examples, a super node may be a logical node deployed within a data consumer resource, such as within AI application node 1 (1506), AI application node 2 (1508), and/or AI application node 3 (1510). In some examples, a super node may be deployed as a physical node located between the data sources and one or more consumers of the data (e.g., AI application node 1 (1506), AI application node 2 (1508), and/or AI application node 3 (1510)), or at a minimum at a location with access to the data stream.

In some examples, once deployed, super node A 1512 and/or super node B 1514 assess if the data streams (e.g., data streams 1516 and 1518) satisfy a nominal pattern or not. In some examples, super node A 1512 is deployed with a first instantiation of a machine learning model that assesses one or more of the data streams 1516 by generating outputs of a value representing a characteristic of one or more of the data streams 1516 and/or a value representing a characteristic of AI application node 1 (1506) and/or AI application node 2 (1508). In some examples, super node B 1512 is deployed with a second instantiation of a machine learning model that assesses one or more of the data streams 1518 by generating outputs of a value representing a characteristic of one or more of the data streams 1518 and/or a value representing a characteristic of AI application node 3 (1510). The example super node A (1512) and example super node B (1514), then exchange the outputs from their respective machine learning models. Then either one or both of example super node A (1512) and example super node B (1514) combine both sets of outputs (e.g., concatenate) and feed the combined outputs back into their respective machine learning models as training data to work towards a consensus of a nominal data pattern across such data streams 1516 and 1518.

In some examples, a super node may monitor a cluster of nodes. For example, super node A 1512 may be deployed to monitor a cluster of nodes that include AI application node 1 (1506) and AI application node 2 (1508). In some examples, the consensus data patterns may be represented by one or more graph nodes and/or one or more correlation factors, as discussed above in FIG. 10 and FIG. 11. In some examples, super nodes may include a nested tree of super nodes, to provide consensus data patterns from a larger number of data streams in bigger networks. In some examples, the consensus data patterns that represent a larger number of data streams may be consensus patterns of multiple consensus patterns across clusters. In some examples, super node A (1512) and/or super node B (1514) may generate an alert for the ADM system 1500 in response to detecting an ethical divergence from a nominal pattern in a given data stream.

In some examples, different AI algorithms (e.g., AI algorithms 1-3 (638)) may be selected in a hierarchy of super nodes. In some examples, a hierarchy of super nodes may provide a tree topology of filtering questionable data streams at ever-greater levels of scrutiny. For example, a first super node may detect anomalies in the data stream (vs. a nominal/baseline data stream). If the first super node determines that an anomaly is present, then the first super node may pass the data stream onto a second super node that is trained to classify anomalies. For example, some anomalies do not create an ethical divergence of the data stream but rather have a valid reason for showing. In some examples, there may be multiple levels of anomaly classification. In some examples, if each super node detects such anomalies, then an additional super node may be deployed to perform gradation on the detected anomalies. For example, not only may a first super node verify an anomaly has taken place and a second super node (e.g., next tier in the hierarchy) then may then classify the anomaly, but a third super node then may grade the anomaly (e.g., a geographical location that is top secret or risky with respect to privacy). Thus, multiple thresholds may exist to allow ML models deployed on super nodes to classify data streams more accurately.

FIG. 16 is a block diagram of example data usage monitoring circuitry 1300 to instantiate/deploy one or more example node types. In some examples, a processing node 1602 is present in an edge environment. The example processing node 1602 has access to an incoming data stream 1604 (e.g., a data stream within network traffic in the edge environment). In some examples, a resource manager orchestration circuitry 1606 orchestrates resources in the edge network to deploy one of a plurality of types of nodes. In some examples, the orchestration is based on data within the incoming data stream 1604.

For example, the resource manager orchestration circuitry 1606 may deploy a “standard” node 1608 to monitor data usage in the edge network, and more specifically to monitor data usage in the incoming data stream 1604. In some examples, the standard node 1608 for monitoring utilizes a deployed primary machine learning (ML) model 1610A to generate outputs corresponding to data within the data stream.

The example resource manager orchestration circuitry 1606 may deploy a super node 1612 to monitor a cluster of AI application nodes that consume data. In some examples, the super node 1612 utilizes another instantiation of the primary machine learning (ML) model 1610B to generate outputs corresponding to data within the data stream. In some examples, the super node 1612 shares data with other super nodes that are also monitoring data streams. In some examples, a group of super nodes that share data may develop a consensus data pattern to use as an example nominal data stream to which a target data stream is compared to effectuate identification and mitigation of ethical divergence of data streams.

The example resource manager orchestration circuitry 1606 may deploy a DDI node 1614 to monitor an ingested data stream with a different (e.g., secondary) machine learning model (e.g., secondary ML model 1616). The secondary ML model 1616 may be utilized because a feature set used to train the secondary ML model 1616 is a closest match to one or more data points or metadata tags corresponding to the incoming data stream 1604. In some examples, a group of DDI nodes performs federated learning to build a consensus nominal stream pattern to compare against any target data stream.

While an example manner of implementing the ADM system 600 of FIG. 6 is illustrated in FIG. 13, one or more of the elements, processes, and/or devices illustrated in FIG. 13 may be combined, divided, re-arranged, omitted, eliminated, and/or implemented in any other way. Further, the example interface circuitry 1302, the example resource manager orchestration circuitry 1304, the example ML circuitry 1306, the example metadata manager circuitry 1308, the example operation execution circuitry 1310, the example algorithm manager circuitry 1312, the example deep data inspection circuitry 1314, the example data consumption tracking circuitry 1316, the example digital rights management (DRM) circuitry 1318, the example datastore 1320, the example bus 1332, and/or, more generally, the example ADM system 600 of FIG. 6, may be implemented by hardware alone or by hardware in combination with software and/or firmware. Thus, for example, any of the example interface circuitry 1302, the example resource manager orchestration circuitry 1304, the example ML circuitry 1306, the example metadata manager circuitry 1308, the example operation execution circuitry 1310, the example algorithm manager circuitry 1312, the example deep data inspection circuitry 1314, the example data consumption tracking circuitry 1316, the example DRM circuitry 1318, the example datastore 1320, the example bus 1332, and/or, more generally, the example ADM system 600 of FIG. 6, could be implemented by processor circuitry, analog circuit(s), digital circuit(s), logic circuit(s), programmable processor(s), programmable microcontroller(s), graphics processing unit(s) (GPU(s)), digital signal processor(s) (DSP(s)), application specific integrated circuit(s) (ASIC(s)), programmable logic device(s) (PLD(s)), and/or field programmable logic device(s) (FPLD(s)) such as Field Programmable Gate Arrays (FPGAs). Further still, the example ADM system 600 of FIG. 6 may include one or more elements, processes, and/or devices in addition to, or instead of, those illustrated in FIG. 13, and/or may include more than one of any or all of the illustrated elements, processes and devices.

FIG. 17 is a flowchart representative of example machine readable instructions and/or example operations 1700 that may be executed and/or instantiated by processor circuitry to effectuate identification and mitigation of an ethical divergence in a data stream. The example machine readable instructions and/or the example operations 1700 of FIG. 17 begin at block 1702, at which the data usage monitoring circuitry 1300 ingests data in a data stream from a data source. For example, the interface circuitry 1302 (FIG. 13) can obtain data from one(s) of the data sources 604 of FIG. 6, the first monitoring sensor 828 of FIG. 8, etc. In some examples, the interface circuitry 1302 can extract data of interest from the data. In some examples, the interface circuitry 1302 can effectuate and/or facilitate identification and mitigation of an ethical divergence in a data stream by ingesting data that can be provided as input(s) to one or more nodes to cause the one or more nodes to generate output(s) representative of achievements in identifying and mitigating ethical divergence in a data stream in an edge environments. An example process that may be executed and/or instantiated by processor circuitry to implement block 1702 is described below in connection with FIG. 18.

At block 1704, the data usage monitoring circuitry 1300 orchestrates resources in an edge environment based on the data ingested from the data source. For example, the resource manager orchestration circuitry 1304 (FIG. 13) can instantiate hardware, software, and/or firmware resources in the ADM system 600 of FIG. 6, the edge network environment 800 of FIG. 8, etc. In some examples, the resource manager orchestration circuitry 1304 can effectuate identification and mitigation of an ethical divergence in a data stream by reducing utilization of resource(s) and/or deployment of resource(s) to accomplish data usage monitoring objectives while complying with the policy 1322 and/or otherwise satisfying the policy 1322. An example process that may be executed and/or instantiated by processor circuitry to implement block 1704 is described below in connection with FIG. 19.

At block 1706, the data usage monitoring circuitry 1300 executes a machine learning (ML) model based on the data to generate outputs, the outputs including at least one of a first value representative of a characteristic of a data stream or a second value representative of a characteristic of an AI application node (consuming or attempting to the consume the data stream). For example, the ML circuitry 1306 (FIG. 13) can execute the ML model A 1328 (FIG. 13) and/or the ML model B 1330 (FIG. 13) using the ingested data from the data sources 604 as inputs to generate outputs, which can include values (e.g., data values, numerical values, alphanumerical values, etc.) indicative of and/or otherwise representative of characteristics of a data stream and/or characteristics of an AI application node associated with the ingested data. In some examples, the ML circuitry 1306 can facilitate identification and mitigation of an ethical divergence in a data stream by generating outputs that may indirectly cause a node to identify and mitigate a potential ethical divergence of a target data stream (e.g., the node may utilize the outputs as inputs to another model, algorithm, routine, process, etc., that may lead to such an identification and mitigation). In some examples, the ML circuitry 1306 can effectuate identification and mitigation of an ethical divergence in a data stream by generating outputs that may directly cause a node to identify and mitigate an ethical divergence in a data stream (e.g., the node may take action(s) based on the outputs). An example process that may be executed and/or instantiated by processor circuitry to implement block 1706 is described below in connection with FIGS. 20-26.

At block 1708, the data usage monitoring circuitry 1300 determines the outputs of a machine learning model satisfy a threshold value to effectuate identification and mitigation of an ethical divergence in a data stream. In some examples, the threshold value is compared against a delta value associated with a difference between a value that represents the target data stream (at least corresponding to one or more data points within the target data stream) and a value that represents at least one nominal/baseline data stream. In some examples, data usage monitoring circuitry 1300 normalizes the delta value (representing a value between 0 and 1). In some examples, if the delta value is above a delta score/value difference, then the threshold value is satisfied (e.g., exceeded).

At block 1710, the data usage monitoring circuitry 1300 causes operation(s) at node(s) of the edge environment based on at least one of the data or the outputs, the node(s) associated with the data. In some examples, the data usage monitoring circuitry 1300 causes operation(s) in response to the threshold value being satisfied.

For example, the operation execution circuitry 1310 (FIG. 13) can cause and/or otherwise invoke a node of the edge network environment 800 to perform one or more operations based on the ingested data. In some examples, the operations can include controlling an action or motion of an autonomous vehicle, equipment, etc., based on ingested sensor data associated with the autonomous vehicle, equipment, etc. An example process that may be executed and/or instantiated by processor circuitry to implement block 1710 is described below in connection with FIG. 27. In response to causing operation(s) at the node(s) of the edge environment based on the data at block 1710, the example machine readable instructions and/or the example operations 1700 of FIG. 17 conclude.

FIG. 18 is a flowchart representative of example machine readable instructions and/or example operations 1800 that may be executed and/or instantiated by processor circuitry to ingest data from a data source. In some examples, the machine readable instructions and/or the operations 1800 of FIG. 18 can be executed and/or instantiated by processor circuitry to implement block 1702 of the machine readable instructions and/or the operations 1700 of FIG. 17. The example machine readable instructions and/or the example operations 1800 of FIG. 18 begin at block 1802, at which the data usage monitoring circuitry 1300 ingests data from a data source at a node. For example, the interface circuitry 1302 (FIG. 13) of the edge cloud 810 can ingest data, such as video data, from the first monitoring sensor 828.

At block 1804, the data usage monitoring circuitry 1300 tags portion(s) of the data with metadata. For example, the interface circuitry 1302 can tag, assign, and/or otherwise embed metadata into portion(s) of the video data. In some examples, the interface circuitry 1302 (e.g., the interface circuitry 1310 of the edge cloud 810) can generate metadata including an IP address, a MAC address, a device type of the first monitoring sensor 828, etc., and/or any combination(s) thereof, and associate the metadata with the portion(s) of the video data.

At block 1806, the data usage monitoring circuitry 1300 queries an orchestrator to identify a machine learning (ML) model as associated with the metadata. For example, the interface circuitry 1302 (e.g., the interface circuitry 1302 of the edge cloud 810) can query the resource manager/orchestration agent 642 of FIG. 6, the edge cloud 110, the cloud data center 130, etc., for one(s) of the algorithms 638, the ML model 1368, etc., that is/are associated with the metadata.

At block 1808, the data usage monitoring circuitry 1300 executes ML model(s) at the node to determine at least one of the first value representative of a characteristic of a data stream or a second value representative of a characteristic of an AI application node. For example, the ML circuitry 1306 (FIG. 13) (e.g., the ML circuitry 1306 of the edge cloud 810) can execute the one(s) of the algorithms 638, the ML model A 1328 and/or the ML model B 1330, etc., that correspond(s) to the metadata. In some examples, the ML circuitry 1306 can execute the algorithms 638, the ML model A 1328 and/or the ML model B 1330, etc., using the metadata and/or the portion(s) of the video data as inputs to generate outputs, which can include the first value representative of a characteristic of a data stream or a second value representative of a characteristic of an AI application node.

In response to executing ML model(s) at the node to determine at least one of the first value representative of a characteristic of a data stream or a second value representative of a characteristic of an AI application node at block 1808, the example machine readable instructions and/or the example operations 1800 of FIG. 18 conclude. For example, the machine readable instructions and/or the operations 1800 of FIG. 18 can return to block 1704 of the machine readable instructions and/or the operations 1700 of FIG. 17 to orchestrate resources in an edge environment based on the data.

FIG. 19 is a flowchart representative of example machine readable instructions and/or example operations 1900 that may be executed and/or instantiated by processor circuitry to orchestrate resources in an edge environment based on data ingested from a data source. In some examples, the machine readable instructions and/or the operations 1900 of FIG. 19 can be executed and/or instantiated by processor circuitry to implement block 1704 of the machine readable instructions and/or the operations 1700 of FIG. 17.

The example machine readable instructions and/or the example operations 1900 of FIG. 19 begin at block 1902, at which the data usage monitoring circuitry 1300 obtains an orchestration policy indicative of at least one of a quantity or a type of workload(s) to be executed in the edge environment. For example, the resource manager orchestration circuitry 1304 (FIG. 13) can obtain the policy 1322 (FIG. 13) associated with the edge network environment 800 of FIG. 8. In some examples, the policy 1362 associated with the edge network environment 800 can include a quantity of workloads expected to be executed during a time period (e.g., every minute, hour, day, week, month, etc.) and/or a type of workload (e.g., types of workloads including acceleration, compute, memory, storage, network, security, etc., workloads) to be executed with resources of the edge network environment 800. In some examples, the resource manager orchestration circuitry 1304 can identify an intention of the policy 1322 to satisfy parameters of an SLA, QoS policy, etc., while achieving overarching goals of reducing environment impact.

At block 1904, the data usage monitoring circuitry 1300 instantiates resources in the edge environment to execute workload(s) based on the orchestration policy. For example, the resource manager orchestration circuitry 1304 can allocate, deploy, and/or launch hardware, software, and/or firmware resources at a node, such as the edge cloud 810 of FIG. 8. In some examples, the edge cloud 810 can execute the workloads with the instantiated resources.

At block 1906, the data usage monitoring circuitry 1300 generates a topology associated with the resources to at least one of execute a workload or route data in the edge environment with the resources. For example, the resource manager orchestration circuitry 1304 can generate a topology (e.g., a resource topology, a network topology, etc.) associated with resources of the edge network environment 800. In some examples, the topology can be a network topology including network connections (e.g., connections using IP addresses and ports, MAC addresses, logical addresses, etc.) to one(s) of the first monitoring sensor 828, the second monitoring sensor 830, the first industrial machine 816, the edge cloud 810, etc., of FIG. 8.

At block 1908, the data usage monitoring circuitry 1300 identifies node(s) as preferred node(s) in the edge environment based on the topology, the preferred node(s) to generate local determinations associated with the data. For example, the resource manager orchestration circuitry 1304 can identify the edge cloud 810 as one of the preferred node(s) in the preferred nodes table 624 of FIG. 6. In some examples, the resource manager orchestration circuitry 1304 can identify the edge cloud 810 to generate local determinations, such as outputs from the ML model A 1328 and/or the ML model B 1330, based on locally generated data, such as data ingested from nodes local to the edge network environment 800, which can include the first monitoring sensor 828, the second monitoring sensor 830, the first industrial machine 816, etc.

At block 1910, the data usage monitoring circuitry 1300 deploys machine learning (ML) model(s) to the node(s) in response to identification(s) of the node(s) as the preferred node(s). For example, the resource manager orchestration circuitry 1304 can provide, transmit, and/or otherwise deliver the ML model A 1328 and/or the ML model B 1330 to the edge cloud 810 in response to an identification of the edge cloud 810 as a preferred node in the edge network environment 800.

In response to deploying ML model(s) to the node(s) in response to identification(s) of the node(s) as the preferred node(s) at block 1910, the example machine readable instructions and/or the example operations 1900 of FIG. 19 conclude. For example, the machine readable instructions and/or the operations 1900 of FIG. 19 can return to block 1706 of the machine readable instructions and/or the operations 1700 of FIG. 17 to execute an ML model based on the data to generate outputs including at least one of a first value representative of a characteristic of a data stream or a second value representative of a characteristic of an AI application node.

FIG. 20 is a flowchart representative of example machine readable instructions and/or example operations 2000 that may be executed and/or instantiated by processor circuitry to execute a machine learning (ML) model based on data to generate outputs. In some examples, the machine readable instructions and/or the operations 2000 of FIG. 20 can be executed and/or instantiated by processor circuitry to implement block 1706 of the machine readable instructions and/or the operations 1700 of FIG. 17.

The example machine readable instructions and/or the example operations 2000 of FIG. 20 begin at block 2002, at which the data usage monitoring circuitry 1300 determines whether to execute the ML model in a training phase or an inference phase. For example, the ML circuitry 1306 (FIG. 13) can determine whether to execute the ML model A 1328 and/or the ML model B 1330 in a training phase (e.g., to train the ML model A 1328 and/or the ML model B 1330 based on training data) or an inference phase (e.g., to execute the ML model A 1328 and/or the ML model B 1330 using live or real-world data).

If, at block 2002, the data usage monitoring circuitry 1300 determines to execute the ML model in a training phase, control proceeds to block 2004. At block 2004, the data usage monitoring circuitry 1300 obtains training data for an observation period. For example, the ML circuitry 1306 can obtain training data, such as labeled sensor data, for an observation period (e.g., labeled sensor data captured for a particular hour, week, day, etc.). In some examples, the ML circuitry 1306 can obtain training data, such as labeled intentions or desires of the identification and mitigation of the ethical divergence of a data stream, to be used to train the ML model 1368.

At block 2006, the data usage monitoring circuitry 1300 executes the ML model using the training data to generate outputs representative of baseline data for characteristics of a data stream and characteristics of an AI application node. For example, the ML circuitry 1306 can execute the ML model A 1328 and/or the ML model B 1330 to determine at least one of a first baseline value of a characteristic of a data stream or a second baseline value of a characteristic of an AI application node associated with the training data that can be used for the inference phase. An example process that may be executed and/or instantiated by processor circuitry to implement block 2006 is described below in connection with FIG. 20. In response to executing the ML model using the training data to generate outputs representative of baseline data for characteristics of a data stream and characteristics of an AI application node at block 2006, control proceeds to block 2010.

If, at block 2002, the data usage monitoring circuitry 1300 determines to execute the ML model in an inference phase, control proceeds to block 2008. At block 2008, the data usage monitoring circuitry 1300 executes the ML model using the ingested data to generate outputs representative of characteristics of a data stream and characteristics of an AI application node of the ingested data. For example, the ML circuitry 1306 can execute the ML model A 1328 and/or the ML model B 1330, based on the training data, to determine at least one of a first value representative of a characteristic of a data stream ingested at a node or a second value representative of a characteristic of an AI application node (consuming or attempting to consume the data stream). An example process that may be executed and/or instantiated by processor circuitry to implement block 2008 is described below in connection with FIG. 21. In response to executing the ML model using the ingested data to generate outputs representative of characteristics of a data stream and characteristics of an AI application node of the ingested data at block 2008, control proceeds to block 2010.

At block 2010, the data usage monitoring circuitry 1300 determines whether to update the baseline data based on the outputs. For example, the ML circuitry 1306 can determine that ingested data is target data stream data and thereby the ML circuitry 1306 can determine to update stored baseline data.

If, at block 2010, the data usage monitoring circuitry 1300 determines to update the baseline data based on the outputs, control proceeds to block 2012. At block 2012, the data usage monitoring circuitry 1300 updates the baseline data in a datastore based on the outputs. In response to updating the baseline data in a datastore based on the outputs at block 2012, the example machine readable instructions and/or the example operations 2000 of FIG. 20 conclude. For example, the machine readable instructions and/or the operations 2000 of FIG. 20 can return to block 1708 of the machine readable instructions and/or the operations 1700 of FIG. 17 to reduce resource requirements associated with the resources of the edge environment to effectuate identification and mitigation of ethical divergence of a data stream based on the outputs.

If, at block 2010, the data usage monitoring circuitry 1300 determines not to update the baseline data based on the outputs, control proceeds to block 2014. At block 2014, the data usage monitoring circuitry 1300 tags the ingested data for identification and mitigation of ethical divergence operations. For example, the ML circuitry 1306 can append metadata to the ingested data to cause the ingested data to undergo identification and mitigation of ethical divergence operation(s), such as being compressed, discarded, and/or otherwise modified. In response to tagging the ingested data for identification and mitigation of ethical divergence operations at block 2014, the example machine readable instructions and/or the example operations 2000 conclude. For example, the machine readable instructions and/or the operations 2000 of FIG. 20 can return to block 1708 of the machine readable instructions and/or the operations 1700 of FIG. 17 to reduce resource requirements associated with the resources of the edge environment to effectuate identification and mitigation of ethical divergence of a data stream based on the outputs.

FIG. 21 is a flowchart representative of example machine readable instructions and/or example operations 2100 that may be executed and/or instantiated by processor circuitry to execute an ML model to generate outputs representative of characteristics of a data stream and characteristics of an AI application node. In some examples, the machine readable instructions and/or the operations 2100 of FIG. 21 can be executed and/or instantiated by processor circuitry to implement block 2006 and/or block 2008 of the machine readable instructions and/or the operations 2000 of FIG. 20.

The example machine readable instructions and/or the example operations 2100 of FIG. 21 begin at block 2102, at which the data usage monitoring circuitry 1300 determines a value of a data stream characteristic based on at least one of training data or ingested data. For example, the ML circuitry 1306 can execute the ML model 1368 (FIG. 13) to determine a first value representative of a characteristic of a data stream associated with training data during a training phase, a second value representative of a characteristic of the data stream associated with data ingested at the logical entity 601 of FIG. 1 during an inference phase, etc. An example process that may be executed and/or instantiated by processor circuitry to implement block 2102 is described below in connection with FIG. 21.

At block 2104, the data usage monitoring circuitry 1300 determines a value of data quality based on the at least one of the training data or the ingested data. For example, the ML circuitry 1306 can execute the ML model 1368 to determine a first value representative of a characteristic of an AI application node associated with training data during a training phase, a second value of the AI application node associated with data ingested at the logical entity 601 of FIG. 1 during an inference phase, etc. An example process that may be executed and/or instantiated by processor circuitry to implement block 2104 is described below in connection with FIG. 22.

At block 2106, the data usage monitoring circuitry 1300 determines whether at least one of the values of the data stream characteristic or the AI application node characteristic satisfy a threshold. For example, the ML circuitry 1306 can execute the ML model 1368 to determine that the first value of the data stream characteristic satisfies a first threshold (e.g., a data stream characteristic threshold, a data stream characteristic threshold associated with a training phase, etc.), the second value of the data stream characteristic satisfies a second threshold (e.g., a data stream characteristic threshold, a data stream characteristic threshold associated with an inference phase, etc.), etc., and/or any combination(s) thereof. In some examples, the ML circuitry 1306 can execute the ML model 1368 to determine that the first value of the AI application node characteristic satisfies a third threshold (e.g., an AI application node characteristic threshold, a, AI application node characteristic threshold associated with a training phase, etc.), the second value of the data quality satisfies a fourth threshold (e.g., an AI application node characteristic, an AI application node characteristic threshold associated with an inference phase, etc.), etc., and/or any combination(s) thereof.

If, at block 2106, the data usage monitoring circuitry 1300 determines that at least one of the values of the data stream characteristic or the AI application node characteristic do not satisfy a threshold, the example machine readable instructions and/or the example operations 2100 conclude. For example, the machine readable instructions and/or the example operations 2100 can return to block 2010 of the machine readable instructions and/or the operations 2000 of FIG. 20 to determine whether to update the baseline data based on the outputs.

If, at block 2106, the data usage monitoring circuitry 1300 determines that at least one of the values of the data stream characteristic or the AI application node characteristic satisfy a threshold, control proceeds to block 2108. At block 2108, the data usage monitoring circuitry 1300 stores the data in a datastore. For example, the ML circuitry 1306 can store the training data, or portion(s) thereof, as the normal data 1418, the critical data 1424, etc., of FIG. 14, and/or any combination(s) thereof. In some examples, the ML circuitry 1306 can store the ingested data, or portion(s) thereof, as the normal data 1418, the critical data 1424, etc., of FIG. 14, and/or any combination(s) thereof.

At block 2110, the data usage monitoring circuitry 1300 generates an alert. For example, the ML circuitry 1306 can generate an alert indicative of the training data, the ingested data, etc., being representative of data to be surfaced to a user, another electronic system in the edge network environment 800 of FIG. 8, etc. In some examples, the ML circuitry 1306 can generate the alert to be indicative of a detection of an unauthorized person in the edge network environment 800, and the alert can be provided to a user (e.g., an HR, IT, security, etc., personnel), an electronic device associated with the user, an electronic device of the edge network environment 800 (e.g., an electronic security system, an alarm, etc.). In some examples, the ML circuitry 1306 can generate the alert to be indicative of a detection of a hazard on a roadway, and the alert can be provided to a vehicle (e.g., a user-operated vehicle, an autonomous vehicle, etc.) to cause the vehicle to avoid the hazard in substantially real time.

In response to generating an alert at block 2110, the example machine readable instructions and/or the example operations 2100 conclude. For example, the machine readable instructions and/or the example operations 2100 can return to block 2010 of the machine readable instructions and/or the operations 2000 of FIG. 20 to determine whether to update the baseline data based on the outputs.

FIG. 22 is a flowchart representative of example machine readable instructions and/or example operations 2200 that may be executed and/or instantiated by processor circuitry to determine a value of a data stream characteristic based on at least one of training data or ingested data. In some examples, the machine readable instructions and/or the operations 2200 of FIG. 22 can be executed and/or instantiated by processor circuitry to implement block 2102 of the machine readable instructions and/or the operations 2100 of FIG. 21.

The example machine readable instructions and/or the example operations 2100 of FIG. 21 begin at block 2102, at which the data usage monitoring circuitry 1300 identifies a potential consequence if the data is not processed or stored. For example, the ML circuitry 1306 (FIG. 13) can generate an output (e.g., an output based on ingested data from the data sources 604), which can be an identification of a concern or consequence, such as violating a regulatory requirement, if data from the data sources 604 is not processed and/or stored. For example, the consequence can be violating a government regulation regarding citizen privacy if biometric data associated with a person is not processed to randomize and/or otherwise obfuscate the biometric data. In some examples, the ML circuitry 1306 can identify a consequence, such as an adverse or undesirable event occurring, if data from the data sources 604 is not processed and/or stored. For example, the undesirable event can be a collision in a warehouse between a hazard (e.g., a person, an object, etc.) and the first industrial machine 816 of FIG. 8 if video data from the first monitoring sensor 828 is not processed and/or stored. In some examples, a consequence can be an increase in environment impacts, such as an increase in resources (e.g., compute, storage, security, acceleration, etc., resources) required to process ingested data. Additionally and/or alternatively, the ML circuitry 1306 can determine that the ingested data includes metadata that corresponds to and/or otherwise identifies the concern or consequence.

The example machine readable instructions and/or the example operations 2200 of FIG. 22 begin at block 2202, at which the data usage monitoring circuitry 1300 determines/identifies a content type of the target data stream. For example, a content type of the data stream may include classifying the content within the data stream as image data, audio data, textual data, telemetry data (e.g., sensor data, etc.), or any other type of data, or a combination of two or more types of such data.

At block 2204, the data usage monitoring circuitry 1300 determines/identifies a sensitive attribute of the target data stream. For example, a sensitive attribute of the data stream may include data that provides/describes sensitive topics, such as identification information of individuals (e.g., human resources data), classification of individuals based on race, gender, or other or any other type of classification, information about financial statements, governmental records, confidential records, or one or more other sensitive attributes.

At block 2206, the data usage monitoring circuitry 1300 determines/identifies a security level of the target data stream. For example, a security level of a data stream may include a level of known confidentiality based on content type, sensitive attributes, etc. such as confidential data, restricted data, top secret data, etc.

At block 2208, the data usage monitoring circuitry 1300 determines/identifies a source location of the source node that is sourcing the target data stream. For example, a source location of a source node that is sourcing the target data stream may be described by an Internet Protocol (IP) address, a physical address, or another type of address that corresponds to a virtual or geographic location that may change how data is viewed. For example, a virtual source location IP address from a bank may cause heightened ethical scrutiny for a data stream with financial data. For example, a geographical source location at a military facility may cause heighted ethical scrutiny for image data. For example, a geographical source location in a recording studio may cause heighted ethical scrutiny for audio data.

At block 2210, the data usage monitoring circuitry 1300 determines a value representative of a characteristic of the target data stream based on the determinations. For example, the ML circuitry 1306 can generate an output (e.g., an output based on ingested data from the data sources 604) by executing the data graph model 1326, the ML model A 1328 and/or the ML model B 1330, etc., and/or any combination(s) thereof, to determine a value of a characteristic of the data streambased on at least one of a content type of the target data stream, a sensitive attribute of the target data stream, a security level of the target data stream, and/or a source location of the source node that is sourcing target data stream. In some examples, the ML circuitry 1306 can determine the value of the characteristic of the data stream to be a numerical value (e.g., a value in a range of 0 to 1). In some examples, the ML circuitry 1306 can determine the value of the data stream characteristic to be a numerical identifier. In some examples, the ML circuitry 1306 can determine the value of data stream characteristic to be a label or text identifier.

In response to determining the value of the data stream characteristic based on the determinations at block 2210, the example machine readable instructions and/or the example operations 2200 conclude. For example, the machine readable instructions and/or the example operations 2200 can return to block 2104 of the machine readable instructions and/or the operations 2100 of FIG. 21 to determine a value of data quality based on the at least one of the training data or the ingested data.

FIG. 23 is a flowchart representative of example machine readable instructions and/or example operations 2200 that may be executed and/or instantiated by processor circuitry to determine a value representative of an AI application node characteristic based on the at least one of the training data or the ingested data. In some examples, the machine readable instructions and/or the operations 2300 of FIG. 23 can be executed and/or instantiated by processor circuitry to implement block 2104 of the machine readable instructions and/or the operations 2100 of FIG. 21.

The example machine readable instructions and/or the example operations 2300 of FIG. 23 begin at block 2302, at which the data usage monitoring circuitry 1300 determines a service type attribute of the data. For example, the ML circuitry 1306 (FIG. 10) can execute one(s) of the algorithms 638 of FIG. 6, the ML model A 1328 and/or the ML model B 1330, etc., to generate outputs. In some examples, the ML circuitry 1306 can execute the ML model A 1328 and/or the ML model B 1330 using the ingested data as inputs to generate outputs, which can include an accuracy of the ingested data. For example, the ML circuitry 1306 can determine whether the data has an accuracy above a threshold (e.g., an accuracy threshold), and thereby satisfies the threshold, based on a comparison of the data, or portion(s) thereof, to data associated with the data graph model 1326 (FIG. 13), baseline data as described herein, etc., and/or any combination(s) thereof. For example, the ML circuitry 1306 can determine that the outputs include a decision, a determination, an insight, etc., that comports and/or otherwise aligns with labeled training data or another source of ground truth data.

At block 2304, the data usage monitoring circuitry 1300 determines a usage context of the AI application node.

At block 2306, the data usage monitoring circuitry 1300 determines a value representative of an AI application node characteristic based on the determinations. For example, the ML circuitry 1306 can determine a value of an AI application node characteristic associated with the ingested data, or portion(s) thereof, based on at least one of the service type attribute or the usage context of the data.

In response to determining the value of the data quality based on the determinations at block 2306, the example machine readable instructions and/or the example operations 2300 conclude. For example, the machine readable instructions and/or the example operations 2300 can return to block 2106 of the machine readable instructions and/or the operations 2100 of FIG. 21 to determine whether at least one of the values representative of the data stream characteristic or the AI application node characteristic satisfy a threshold.

FIG. 24 is a flowchart representative of example machine readable instructions and/or example operations 2400 that may be executed and/or instantiated by processor circuitry to implement a super node in the edge environment. The example machine readable instructions and/or the example operations 2400 of FIG. 24 begin at block 2402, at which the data usage monitoring circuitry 1300 instantiates a first super node.

At block 2404 the data usage monitoring circuitry 1300 ingests network traffic at the first super node. In some examples, a target data stream is present within the ingested network traffic. In some examples, a target AI application node is consuming the target data stream or attempting to consume the target data stream.

At block 2406 the data usage monitoring circuitry 1300 executes the machine learning model at the first super node based on the ingested network traffic at the first super node to generate outputs representative of a data stream characteristic or an AI application node characteristic.

At block 2408, the data usage monitoring circuitry 1300 shares the outputs from the first super node with one or more additional super nodes (e.g., a second super node).

At block 2410, the data usage monitoring circuitry 1300 obtains outputs from one or more additional super nodes (e.g., the second super node). In some examples, the first super node may share first outputs with a first set of additional super nodes and may obtain second outputs from a second set of additional super nodes where the first and second set of additional super nodes may not be the same. For example, the first and second set of super nodes may be the same set of super nodes, may be a completely different set of super nodes with no overlapping super nodes, or may be a combination of some super nodes from the first set of super nodes and other super nodes not in the first set of super nodes. In some examples, the one or more additional super nodes that share the second outputs with the first super node may also share the data streams (or portions of the data streams) that created the second outputs when input into the machine learning model(s). In some examples, the portions of the data streams may include individual data points and/or features within the data streams.

At block 2412, the data usage monitoring circuitry 1300 tracks metadata in the target data stream and tracks metadata that is present in the target data stream and tracks behavior of the target AI application node that is consuming or attempting to consume the target data stream. In some examples, the data consumption tracker circuitry 1316 accesses the target data stream at the super node location and keep a count of each metadata tagging an amount of data in the target data stream. For example, the data consumption tracker circuitry 1316 may count metadata tags that indicate a specific content type in one or more portions of data in the target data stream. In some examples, the data consumption tracker circuitry 1316 counts occurrences of behavior from the target AI application node. For example, the data consumption tracker circuitry 1316 may determine a usage context of the target AI application node when the target AI application node attempts to consume a portion of data in the data stream that is tagged with metadata. For example, a patient intake target AI application node in a healthcare facility retrieves a portion of data in the data stream that is tagged with a sensitive attribute regarding personal identification information, but in the usage context, this particular target AI application node does not indicate an ethical divergence anomaly. Thus, in some examples, the data consumption tracker circuitry 1316 tracks both metadata in the target data stream and tracks metadata that is present in the target data stream and tracks behavior of the target AI application node in order to accurately provide detection of anomalies.

At block 2414, the data usage monitoring circuitry 1300 builds at least one consensus nominal data stream pattern based on the first outputs (e.g., generated by the first super node) and the second outputs (e.g., generated by one or more additional super nodes). In some examples, the ML circuitry 1306 executes the machine learning model with a combination of the first outputs and second outputs in a training phase to converge to a nominal data stream pattern for use in comparison against target data streams.

In response to building the consensus nominal data stream pattern at block 2414, the example machine readable instructions and/or the example operations 2400 conclude.

FIG. 25 is a flowchart representative of example machine readable instructions and/or example operations 2500 that may be executed and/or instantiated by processor circuitry to implement a deep data inspection (DDI) node in the edge environment. The example machine readable instructions and/or the example operations 2500 of FIG. 25 begin at block 2502, at which the data usage monitoring circuitry 1300 instantiates/deploys a DDI node. For example, the resource manager orchestration circuitry 1304 can orchestrate resources in the edge environment to instantiate a DDI node. In some examples, the DDI node is a physical node that provides DDI functionality to the edge environment. In some examples, the DDI node is a virtual node that is instantiated within another physical node. For example, the resource manager orchestration circuitry 1304 can deploy a virtual DDI node into the AI application node that is to be tracked.

At block 2504, the data usage monitoring circuitry 1300 ingests network traffic (e.g., ingests data in data streams in the network traffic) at the DDI node. For example, the interface circuitry 1302 (FIG. 13) can ingest data, such as video data or audio data.

At block 2506, the data usage monitoring circuitry 1300 tracks metadata in the target data stream and tracks the behavior of the target AI application node at the DDI node. For example, the data consumption tracking circuitry 1316 can perform tasks that are described in FIG. 28.

At block 2508, the data usage monitoring circuitry 1300 selects a machine learning model for use at the DDI node. For example, the algorithm manager circuitry 1312 selects an AI algorithm used to train a machine learning model to be executed on the DDI node. In some examples, the machine learning model selected (e.g., selected either directly or through the selection of the AI algorithm used to train the machine learning model) has matching features to features/characteristics present in the target data stream. For example, the interface circuitry 1302 may tag data within a target data stream with metadata that corresponds to one or more features (e.g., characteristics such as data stream characteristics and/or AI application node characteristics) representative of the data. In some examples, the algorithm manager circuitry 1312 may have access to a plurality of AI algorithms (e.g., algorithms 1-3 (638)), each of which may focus on a subset of an overall set of features available. Thus, in some examples, the algorithm manager circuitry 1312 may select a second algorithm used to train ML model B 1330 (as opposed to a first algorithm used to train ML model A 1328). In some examples, ML model A 1328 and ML model B 1330 are trained to different feature sets. In some examples, algorithm manager circuitry 1312 will select an ML model trained with either an algorithm that utilizes the set of features in the target data stream, or, alternatively, an algorithm that utilizes a closest feature set available to the set of features in the target data stream.

At block 2510, the data usage monitoring circuitry 1300 executes, over a period of time, a machine learning model (e.g., selected at block 2508) based on the target data stream, to generate second outputs representative of a data stream characteristic and/or a target AI application node characteristic. For example, the ML circuitry 1306 can execute the machine learning model to focus on a most closely related set of features to the features/characteristics in the target data stream to more accurately allow a comparison of the target data stream to one or more nominal data streams.

At block 2512, the data usage monitoring circuitry 1300 synchronizes data with other DDI nodes deployed in the edge environment over a period of time. For example, the DDI circuitry 1314 can synchronize data patterns with other one(s) of the nodes 1408, 1410, and/or 1412 (FIG. 14) and/or with the central node or cloud 1438 (FIG. 14). For example, the data query manager 610 may synchronize the data patterns with stored data patterns.

At block 2514, the data usage monitoring circuitry 1300 determines whether the second outputs, generated from block 2510, satisfy a threshold value at any point within the period of time. For example, the DDI circuitry 1314 may not continue to process data related to a target AI application node if no anomaly is found when the target data stream is compared against one or more nominal data streams over a period of time. If a threshold value (e.g., an anomaly threshold ethical divergence value) is not found, then the process completes. Otherwise, the process continues at block 2516.

At block 2516, the data usage monitoring circuitry 1300 determines whether an ethical divergence of a target data stream occurs at the DDI node greater than a frequency threshold. For example, the DDI circuitry 1314 may observe a divergence from the baseline/nominal data stream happens at least at a minimum rate over time. In some examples, if the ethical divergence rate does not exceed the minimum threshold, then the process completes. Otherwise, the process continues at block 2518.

At block 2516, the data usage monitoring circuitry 1300 determines whether an ethical divergence of a target data stream occurs at other DDI nodes greater than a frequency threshold. For example, the DDI circuitry 1314 may observe a divergence from the baseline/nominal data stream happens at least at a minimum rate over time at other DDI nodes (e.g., the divergence is potentially commonplace at more than just the target AI application node). In some examples, if the ethical divergence rate at other DDI nodes does not exceed the minimum threshold, then the process completes. Otherwise, the process continues at block 2520.

At block 2520, the data usage monitoring circuitry 1300 determines whether an ethical divergence of a target data stream meets a constraint. For example, the DDI circuitry 1314 may observe that while a divergence is present and the divergence exceeds the minimum threshold, the divergence is within a known constraint. For example, a regulatory requirement associated with a geographical region is known and therefore divergences happen, but are not mitigated for a reason. In some examples, if the ethical divergence of the target data stream meets the constraint, then the process completes. Otherwise, the process continues at block 2520

At block 2522, the data usage monitoring circuitry 1300 causes modification of the target data stream or a response to the attempt to consume the target data stream. For example, the DDI circuitry 1314 may trigger an action and/or control of data usage as defined in a service level agreement that governs the ingested data request. In some examples, a modification to the target data stream or a response to the attempt to consume the target data stream may include one or more processes described in FIG. 29. In response to triggering the action and/or control of data usage at block 2522, the example machine readable instructions and/or the example operations 2500 concludes.

FIG. 26 is a flowchart representative of example machine readable instructions and/or example operations 2600 that may be executed and/or instantiated by processor circuitry to generate graph nodes for use in an edge environment. In some examples, the machine readable instructions and/or the operations 2600 of FIG. 26 can be executed and/or instantiated by processor circuitry to implement machine readable instructions.

The example machine readable instructions and/or the example operations 2600 of FIG. 26 begin at block 2602, at which the data usage monitoring circuitry 1300 aggregates data at node(s) based on at least one of filtering, correlation, or delta scores using data graph model(s). For example, the metadata manager circuitry 1308 (FIG. 13) can collect and/or otherwise aggregate data at one(s) of the first nodes 1206 (FIG. 12). In some examples, the metadata manager circuitry 1308 can filter the data using one or more filter parameters (e.g., a type of data, a type of device that produced the data, a timestamp or range of timestamps associated with the data, metadata associated with the data, etc.) to identify a subset of the data. In some examples, the metadata manager circuitry 1308 can correlate the data at one(s) of the first nodes 1206 by analyzing vector lengths, angles, etc., of the data graph model 1326 of FIG. 13 and/or of the graph models 1202, 1204 of FIG. 12, etc.

In some examples, the metadata manager circuitry 1308 may filter data using multiple layer classifications. In some examples, data from a target data stream that is represented by a group of data points in a graph model (e.g., graph models 1202 and/or 1204 in FIG. 12) may include multiple layers of data. For example, a first (highest) layer of data ingested from a target data stream may include a type of data (e.g., image data, audio data, telemetry data, textual data, etc.) and filtering may include or exclude ingested data when data does not meet a particular type. For example, when a node (such as a super node) monitors data ingested in a data stream for a deviation (e.g., ethical divergence), comparing the target data stream to one or more nominal data streams, the super node may be monitoring for a particular type of data. For example, a super node may be monitoring for anomalies, deviations, ethical divergences of financial data to generate an alert. The example super node that is monitoring for financial data may disregard (e.g., filter) data that does not meet a textual data type (e.g., financial figures in the form of numbers/text). Thus, at a first filtering layer, the example super node may allow numbers and text to pass through for examination by a second filtering layer. At the second filtering layer, the example super node may compare the textual/number data that passed through the first filtering layer to specific financial information (e.g., bank account numbers and passwords, financial earnings data during a blackout period prior to public earnings announcements, or other vulnerable financial data). Thus, the ingested data may make it through the first layer filter because it is textual/number data, but it may not make it past the second filter to get to an alert generation stage because the textual/number data may be innocuous rather than constituting vulnerable financial data. In some examples, each filtering layer may be the job of a different super node (or other type of node, such as a standard node or DDI node) and the set of nodes that make up a set of filter layers may be in a nested exposure to the data. For example, the node implementing the first/broadest filtering layer may see all/most of the data from an ingested data stream but only data that passes through the first layer may be seen (e.g., operated upon) by a next filter layer. In some examples, the nesting of filtering layers across nodes may be beneficial from a data throughput/performance standpoint because a highest/most basic filtering layer may be less computationally intense than a lowest/most complex filtering layer.

At block 2604, the data usage monitoring circuitry 1300 removes duplicate data based on the data graph model(s). For example, the metadata manager circuitry 1308 can determine that data at and/or otherwise ingested at one(s) of the first nodes 1206 is/are duplicative based on comparing first metadata of the data with second metadata of the data graph model 1326, the graph models 1102, 1104. For example, the green data management circuitry 1040 can determine that the data has previously been analyzed and/or stored based on a presence of the first metadata, or portion(s) thereof, in the data graph model 1066, the graph models 1202, 1204, etc., as part of the second metadata.

In response to removing duplicate data based on the data graph model(s) at block 2604, the example machine readable instructions and/or the example operations 2600 conclude.

FIG. 27 is a flowchart representative of example machine readable instructions and/or example operations 2700 that may be executed and/or instantiated by processor circuitry to cause operation(s) at node(s) of the edge environment based on at least one of the data or the outputs, the node(s) associated with the data. In some examples, the machine readable instructions and/or the operations 2700 of FIG. 27 can be executed and/or instantiated by processor circuitry to implement block 1710 of the machine readable instructions and/or the operations 1700 of FIG. 17.

The example machine readable instructions and/or the example operations 2700 of FIG. 27 begin at block 2702, at which the data usage monitoring circuitry 1300 identifies operation(s) to be executed at node(s) based on the outputs. For example, the operation execution circuitry 1310 (FIG. 13) can identify an operation to be executed at a node, such as the first industrial machine 816, the edge cloud 810, etc., and/or any combination(s) thereof. In some examples, the operation execution circuitry 1310 can identify an operation based on data ingested at one(s) of the data sources 604 to cause the one(s) of the data sources 604 to achieve reduced environment impact. For example, the one(s) of the data sources 604 can generate sensor data; the operation execution circuitry 1310 can identify an operation to reduce resources associated with the one(s) of the data sources 604 based on the sensor data; and the one(s) of the data sources 604 can operate with reduced environment impact in response to execution of the operation.

At block 2704, the data usage monitoring circuitry 1300 alerts the node(s) of the operation(s) to cause the operation(s) to be executed. For example, the operation execution circuitry 1310 can generate an alert indicative of an operation associated with super node 1512 (FIG. 15), such as initiate data monitoring on data streams 1516 (FIG. 15). For example, the first industrial machine 816, such as performing a lift task or operation. In some examples, the operation execution circuitry 1050 can generate the alert to include a command, an instruction, etc., to cause the first industrial machine 816 to perform the operation. For example, the operation execution circuitry 1050 can transmit the alert to a controller, processor circuitry, etc., of the first industrial machine 816 to move to a different position in the edge network environment 800 (e.g., a different aisle, section, etc.) at a specified velocity to perform a lifting task or operation based on a determination that the first industrial machine 816 completed a previous lifting task or operation.

In response to alerting the node(s) of the operation(s) to cause the operation(s) to be executed at block 2704, the example machine readable instructions and/or the example operations 2700 conclude. For example, the machine readable instructions and/or the example operations 2700 can return to the machine readable instructions and/or the operations 1700 of FIG. 17.

FIG. 28 is a flowchart representative of example machine readable instructions and/or example operations 2800 that may be executed and/or instantiated by processor circuitry to track data consumption in an edge environment. In some examples, the machine readable instructions and/or the operations 2800 of FIG. 28 can be executed and/or instantiated by processor circuitry to implement block 2412 of the machine readable instructions and/or the operations 2400 of FIG. 24 or to implement block 2506 of the machine readable instructions and/or the operations 2500 of FIG. 25.

The example machine readable instructions and/or the example operations 2800 of FIG. 28 begin at block 2802, at which the data usage monitoring circuitry 1300 filters data in a target data stream to obtain metadata. For example, the data consumption tracking circuitry 1316 can monitor data within a target data stream for metadata tags that are attached to the data. In some examples, the target data stream is being consumed or attempted to be consumed by a target AI application node. In some examples, the data being monitored in the target data stream is within data payload portions of data packets in the target data stream and the metadata tags are in header portions of the data packets. In some examples, the metadata being filtered is in the data payload portion of one or more data packets, such as hashed into the raw data or concatenated to the raw data in a data packet. Once the data consumption tracking circuitry 1316 observes a metadata tag, it creates a copy of the observed metadata tag for future use.

At block 2804, the data usage monitoring circuitry 1300 analyzes the target AI application node behavior corresponding to the obtained metadata. For example, a filtered metadata tag may correspond to a characteristic (e.g., feature) of the data in the target data stream, such as a sensitive attribute (e.g., financial information), and the data consumption tracking circuitry 1316 may associate an observed/analyzed characteristic of the target AI application node (e.g., a service type attribute or a usage context) to the metadata tag from the data stream to provide greater context to the consumption (or consumption attempt).

At block 2806, the data usage monitoring circuitry 1300 adds to a metadata count and/or a target AI application behavior count for obtained metadata in the data stream and the corresponding behavior. For example, the data consumption tracking circuitry 1316 may keep a running count of each metadata tag for the target data stream and, additionally, may keep a running count of each target AI application node characteristic accompanying the consumption (or consumption attempts).

At block 2808, the data usage monitoring circuitry 1300 determines whether any metadata count or target AI application behavior count (e.g., updated during the operation of block 2806) has satisfied a threshold value. For example, the data consumption tracking circuitry 1316 may compare a count of a metadata tag to a threshold count and if the metadata tag count equals or exceeds the threshold count, then the threshold value has been satisfied. In some examples, the monitored counts of metadata tags and/or target AI application node behavior counts may be counted over a period of time. For example, there may be a window of a period of time that a count has to exceed within the period of time (e.g., a rolling window of the most recent one second of monitoring). In some examples, a threshold value may be a rate of occurrences per second instead of an actual count value of occurrences.

At block 2810, in response to the metadata count or the target AI application behavior count satisfying the threshold value, the data usage monitoring circuitry 1300 may generate an alert. For example, the data consumption tracking circuitry 1316 can send an alert to a designated alert handler node. In some examples, the data consumption tracking circuitry 1316 may send the alert to a cloud or to a central node. In some examples, the data consumption tracking circuitry 1316 may send the alert to the data rights management (DRM) circuitry 1318 for further handling. In some examples, the data consumption tracking circuitry 1316 may send the alert to the operation execution circuitry for further handling.

In response to alerting the node(s) of the operation(s) to cause the operation(s) to be executed at block 2810, the example machine readable instructions and/or the example operations 2800 conclude. For example, the machine readable instructions and/or the example operations 2800 can return to the machine readable instructions and/or the operations 2400 of FIG. 24 or 2500 of FIG. 25.

FIG. 29 is a flowchart representative of example machine readable instructions and/or example operations 2900 that may be executed and/or instantiated by processor circuitry to track data consumption in an edge environment. In some examples, the machine readable instructions and/or the operations 2900 of FIG. 29 can be executed and/or instantiated by processor circuitry to implement block 2522 of the machine readable instructions and/or the operations 2500 of FIG. 25.

In some examples, the machine readable instructions and/or the example operations 2900 of FIG. 29 begin at block 2902, at which the data usage monitoring circuitry 1300 determines a digital rights management (DRM) policy (e.g., one or more policies, such as policy 1322, stored in the datastore 1320) for the data in a target data stream. In some examples, the policy determines, based on a consumption attempt of the data in the target data stream, if one or more of the processes described by blocks 2904, 2906, 2908, and/or 2910 are performed by the DRM circuitry 1318.

At block 2904, the data usage monitoring circuitry 1300, based on policy 1322, may tag data in the target data stream with metadata for tracking. For example, the DRM circuitry 1318 may tag data in the target data stream with metadata tags corresponding to tracking information about the time, date, source location, ownership of the data, identification of target AI application node attempting the consumption, and/or other types of tags.

At block 2906, the data usage monitoring circuitry 1300, based on policy 1322, may hash tracking information into the data in the target data stream. For example, the DRM circuitry 1318 may conceal tags/tracking information in the data stream by hashing such information into raw data in the target data stream.

At block 2908, the data usage monitoring circuitry 1300, based on policy 1322, may implement blockchaining for data in the target data stream. For example, the DRM circuitry 1318 may manage a blockchain ledger for the data in the target data stream to keep track of every owner of the data since the inception of the data.

At block 2910, the data usage monitoring circuitry 1300, based on policy 1322, may prohibit consumption of the data stream by the target AI application node. For example, the DRM circuitry 1318 may divert or stop the target data stream from reaching the node in any one of a number of ways, such as directing the source node to discontinue streaming the data in question in the target data stream.

FIG. 30 is a block diagram of an example processor platform 3000 structured to execute and/or instantiate the machine readable instructions and/or the operations of FIGS. 7 and/or 17-29 to implement the logical entity 601 of FIG. 6, and/or, more generally, the ADM system 600 of FIG. 6. The processor platform 3000 can be, for example, a server, a personal computer, a workstation, a self-learning machine (e.g., a neural network), a mobile device (e.g., a cell phone, a smart phone, a tablet such as an iPad™), a personal digital assistant (PDA), an Internet appliance, a DVD player, a CD player, a digital video recorder, a Blu-ray player, a gaming console, a personal video recorder, a set top box, a headset (e.g., an augmented reality (AR) headset, a virtual reality (VR) headset, etc.) or other wearable device, or any other type of computing device.

The processor platform 3000 of the illustrated example includes processor circuitry 3012. The processor circuitry 3012 of the illustrated example is hardware. For example, the processor circuitry 3012 can be implemented by one or more integrated circuits, logic circuits, FPGAs, microprocessors, CPUs, GPUs, XPUs, DSPs, and/or microcontrollers from any desired family or manufacturer. The processor circuitry 3012 may be implemented by one or more semiconductor based (e.g., silicon based) devices. In this example, the processor circuitry 3012 implements the data ingestion manager 606, the data query manager 610, the data publishing manager 618, the node manager 622, the data security manager 632, the algorithm manager/recommender 634, and the analytics manager 636 of FIG. 6.

The processor circuitry 3012 of the illustrated example includes a local memory 3013 (e.g., a cache, registers, etc.). The processor circuitry 3012 of the illustrated example is in communication with a main memory including a volatile memory 3014 and a non-volatile memory 3016 by a bus 3018. The volatile memory 3014 may be implemented by Synchronous Dynamic Random Access Memory (SDRAM), Dynamic Random Access Memory (DRAM), RAMBUS® Dynamic Random Access Memory (RDRAM®), and/or any other type of RAM device. The non-volatile memory 3016 may be implemented by flash memory and/or any other desired type of memory device. Access to the main memory 3014, 3016 of the illustrated example is controlled by a memory controller 3017.

The processor platform 3000 of the illustrated example also includes interface circuitry 3020. The interface circuitry 3020 may be implemented by hardware in accordance with any type of interface standard, such as an Ethernet interface, a universal serial bus (USB) interface, a Bluetooth® interface, a near field communication (NFC) interface, a Peripheral Component Interconnect (PCI) interface, and/or a Peripheral Component Interconnect Express (PCIe) interface.

In the illustrated example, one or more input devices 3022 are connected to the interface circuitry 3020. The input device(s) 3022 permit(s) a user to enter data and/or commands into the processor circuitry 3012. The input device(s) 3022 can be implemented by, for example, a sensor (e.g., a light sensor, a humidity sensor, a motion sensor, a temperature sensor, etc.), an audio sensor, a microphone, a camera (still or video), a keyboard, a button, a mouse, a touchscreen, a track-pad, a trackball, an isopoint device, and/or a voice recognition system.

One or more output devices 3024 are also connected to the interface circuitry 3020 of the illustrated example. The output device(s) 3024 can be implemented, for example, by display devices (e.g., a light emitting diode (LED), an organic light emitting diode (OLED), a liquid crystal display (LCD), a cathode ray tube (CRT) display, an in-place switching (IPS) display, a touchscreen, etc.), a tactile output device, a printer, and/or speaker. The interface circuitry 3020 of the illustrated example, thus, typically includes a graphics driver card, a graphics driver chip, and/or graphics processor circuitry such as a GPU.

The interface circuitry 3020 of the illustrated example also includes a communication device such as a transmitter, a receiver, a transceiver, a modem, a residential gateway, a wireless access point, and/or a network interface to facilitate exchange of data with external machines (e.g., computing devices of any kind) by a network 3026. The communication can be by, for example, an Ethernet connection, a digital subscriber line (DSL) connection, a telephone line connection, a coaxial cable system, a satellite system, a line-of-site wireless system, a cellular telephone system, an optical connection, etc.

The processor platform 3000 of the illustrated example also includes one or more mass storage devices 3028 to store software and/or data. Examples of such mass storage devices 3028 include magnetic storage devices, optical storage devices, floppy disk drives, HDDs, CDs, Blu-ray disk drives, redundant array of independent disks (RAID) systems, solid state storage devices such as flash memory devices and/or SSDs, and DVD drives. In this example, the one or more mass storage devices 3028 implement the distributed datastore 644, the metadata datastore 646, and the raw datastore 648 of FIG. 6.

The machine executable instructions 3032, which may be implemented by the machine readable instructions of FIGS. 7 and/or 17-29, may be stored in the mass storage device 3028, in the volatile memory 3014, in the non-volatile memory 3016, and/or on a removable non-transitory computer readable storage medium such as a CD or DVD.

The processor platform 3000 of the illustrated example of FIG. 30 includes example acceleration circuitry 3040, which includes an example graphics processing unit (GPU) 3042, an example vision processing unit (VPU) 3044, and an example neural network processor 3046. In this example, the GPU 3042, the VPU 3044, and the neural network processor 3046 are in communication with different hardware of the processor platform 3000, such as the volatile memory 3014, the non-volatile memory 3016, etc., via the bus 3018. In this example, the neural network processor 3046 may be implemented by one or more integrated circuits, logic circuits, microprocessors, GPUs, DSPs, or controllers from any desired family or manufacturer that can be used to execute an AI model, such as a neural network, which may be implemented by the algorithms 638 of FIG. 6. In some examples, one or more of the data ingestion manager 606, the data query manager 610, the data publishing manager 618, the node manager 622, the data security manager 632, the algorithm manager/recommender 634, and/or the analytics manager 636 of FIG. 6 can be implemented in or with at least one of the GPU 3042, the VPU 3044, or the neural network processor 3046 instead of or in addition to the processor circuitry 3012.

FIG. 31 is a block diagram of an example processor platform 3100 structured to execute and/or instantiate the machine readable instructions and/or the operations of FIGS. 7 and/or 17-29 to implement the data usage monitoring circuitry 1300 of FIG. 10. The processor platform 3100 can be, for example, a server, a personal computer, a workstation, a self-learning machine (e.g., a neural network), a mobile device (e.g., a cell phone, a smart phone, a tablet such as an iPad™), a PDA, an Internet appliance, a DVD player, a CD player, a digital video recorder, a Blu-ray player, a gaming console, a personal video recorder, a set top box, a headset (e.g., an AR headset, a VR headset, etc.) or other wearable device, or any other type of computing device.

The processor platform 3100 of the illustrated example includes processor circuitry 3112. The processor circuitry 3112 of the illustrated example is hardware. For example, the processor circuitry 3112 can be implemented by one or more integrated circuits, logic circuits, FPGAs, microprocessors, CPUs, GPUs, XPUs, DSPs, and/or microcontrollers from any desired family or manufacturer. The processor circuitry 3112 may be implemented by one or more semiconductor based (e.g., silicon based) devices. In this example, the processor circuitry 3112 implements the resource manager orchestration circuitry 1304, the ML circuitry 1306, the metadata manager circuitry 1308, the operation execution circuitry 1310, the algorithm manager circuitry 1312, and the DDI circuitry 1314 of FIG. 13.

The processor circuitry 3112 of the illustrated example includes a local memory 3113 (e.g., a cache, registers, etc.). The processor circuitry 3112 of the illustrated example is in communication with a main memory including a volatile memory 3114 and a non-volatile memory 3116 by a bus 3118. The volatile memory 3114 may be implemented by SDRAM, DRAM, RDRAM®, and/or any other type of RAM device. The non-volatile memory 3116 may be implemented by flash memory and/or any other desired type of memory device. Access to the main memory 3114, 3116 of the illustrated example is controlled by a memory controller 3117.

The processor platform 3100 of the illustrated example also includes interface circuitry 3120. The interface circuitry 3120 may be implemented by hardware in accordance with any type of interface standard, such as an Ethernet interface, a USB interface, a Bluetooth® interface, an NFC interface, a PCI interface, and/or a PCIe interface. In this example, the interface circuitry 3120 implements the interface circuitry 1302 of FIG. 13.

In the illustrated example, one or more input devices 3122 are connected to the interface circuitry 3120. The input device(s) 3122 permit(s) a user to enter data and/or commands into the processor circuitry 3112. The input device(s) 3122 can be implemented by, for example, a sensor (e.g., a light sensor, a humidity sensor, a motion sensor, a temperature sensor, etc.), an audio sensor, a microphone, a camera (still or video), a keyboard, a button, a mouse, a touchscreen, a track-pad, a trackball, an isopoint device, and/or a voice recognition system.

One or more output devices 3124 are also connected to the interface circuitry 3120 of the illustrated example. The output device(s) 3124 can be implemented, for example, by display devices (e.g., an LED, an OLED, an LCD, a CRT display, an IPS display, a touchscreen, etc.), a tactile output device, a printer, and/or speaker. The interface circuitry 3120 of the illustrated example, thus, typically includes a graphics driver card, a graphics driver chip, and/or graphics processor circuitry such as a GPU.

The interface circuitry 3120 of the illustrated example also includes a communication device such as a transmitter, a receiver, a transceiver, a modem, a residential gateway, a wireless access point, and/or a network interface to facilitate exchange of data with external machines (e.g., computing devices of any kind) by a network 3126. The communication can be by, for example, an Ethernet connection, a DSL connection, a telephone line connection, a coaxial cable system, a satellite system, a line-of-site wireless system, a cellular telephone system, an optical connection, etc.

The processor platform 3100 of the illustrated example also includes one or more mass storage devices 3128 to store software and/or data. Examples of such mass storage devices 3128 include magnetic storage devices, optical storage devices, floppy disk drives, HDDs, CDs, Blu-ray disk drives, RAID systems, solid state storage devices such as flash memory devices and/or SSDs, and DVD drives. In this example, the one or more mass storage devices 3128 implement the datastore 1320, the policy 1322, the metadata 1324, the data graph model 1326, and the ML models AB 1328/1330 of FIG. 13.

The machine executable instructions 3132, which may be implemented by the machine readable instructions of FIGS. 7 and/or 17-29, may be stored in the mass storage device 3128, in the volatile memory 3114, in the non-volatile memory 3116, and/or on a removable non-transitory computer readable storage medium such as a CD or DVD.

The processor platform 3100 of the illustrated example of FIG. 31 includes example acceleration circuitry 3140, which includes an example graphics processing unit (GPU) 3142, an example vision processing unit (VPU) 3144, and an example neural network processor 3146. In this example, the GPU 3142, the VPU 3144, and the neural network processor 3146 are in communication with different hardware of the processor platform 3100, such as the volatile memory 3114, the non-volatile memory 3116, etc., via the bus 3118. In this example, the neural network processor 3146 may be implemented by one or more integrated circuits, logic circuits, microprocessors, GPUs, DSPs, or controllers from any desired family or manufacturer that can be used to execute an AI model, such as a neural network, which may be implemented by the ML model 1068 of FIG. 10. In some examples, one or more of the resource manager orchestration circuitry 1020, the ML circuitry 1306, and/or the operation execution circuitry 1310 can be implemented in or with at least one of the GPU 3142, the VPU 3144, or the neural network processor 3146 instead of or in addition to the processor circuitry 3112.

FIG. 32 is a block diagram of an example implementation of the processor circuitry 3012 of FIG. 30 and/or the processor circuitry 3112 of FIG. 31. In this example, the processor circuitry 3012 of FIG. 30 and/or the processor circuitry 3112 of FIG. 31 is implemented by a general purpose microprocessor 3200. The general purpose microprocessor circuitry 3200 executes some or all of the machine readable instructions of the flowcharts of FIGS. 7 and/or 17-29 to effectively instantiate the data usage monitoring circuitry 1300 of FIG. 13 as logic circuits to perform the operations corresponding to those machine readable instructions. In some such examples, the data usage monitoring circuitry 1300 of FIG. 13 is instantiated by the hardware circuits of the microprocessor 3200 in combination with the instructions. For example, the microprocessor 3200 may implement multi-core hardware circuitry such as a CPU, a DSP, a GPU, an XPU, an ASIC, etc. Although it may include any number of example cores 3202 (e.g., 1 core), the microprocessor 3200 of this example is a multi-core semiconductor device including N cores. The cores 3202 of the microprocessor 3200 may operate independently or may cooperate to execute machine readable instructions. For example, machine code corresponding to a firmware program, an embedded software program, or a software program may be executed by one of the cores 3202 or may be executed by multiple ones of the cores 3202 at the same or different times. In some examples, the machine code corresponding to the firmware program, the embedded software program, or the software program is split into threads and executed in parallel by two or more of the cores 3202. The software program may correspond to a portion or all of the machine readable instructions and/or operations represented by the FIGS. 1 and/or 17-29.

The cores 3202 may communicate by a first example bus 3204. In some examples, the first bus 3204 may implement a communication bus to effectuate communication associated with one(s) of the cores 3202. For example, the first bus 3204 may implement at least one of an Inter-Integrated Circuit (I2C) bus, a Serial Peripheral Interface (SPI) bus, a PCI bus, or a PCIe bus. Additionally or alternatively, the first bus 3204 may implement any other type of computing or electrical bus. The cores 3202 may obtain data, instructions, and/or signals from one or more external devices by example interface circuitry 3206. The cores 3202 may output data, instructions, and/or signals to the one or more external devices by the interface circuitry 3206. Although the cores 3202 of this example include example local memory 3220 (e.g., Level 1 (L1) cache that may be split into an L1 data cache and an L1 instruction cache), the microprocessor 3200 also includes example shared memory 3210 that may be shared by the cores (e.g., Level 2 (L2_cache)) for high-speed access to data and/or instructions. Data and/or instructions may be transferred (e.g., shared) by writing to and/or reading from the shared memory 3210. The local memory 3220 of each of the cores 3202 and the shared memory 3210 may be part of a hierarchy of storage devices including multiple levels of cache memory and the main memory (e.g., the main memory 3014, 3016 of FIG. 30, the main memory 3114, 3116 of FIG. 31, etc.). Typically, higher levels of memory in the hierarchy exhibit lower access time and have smaller storage capacity than lower levels of memory. Changes in the various levels of the cache hierarchy are managed (e.g., coordinated) by a cache coherency policy.

Each core 3202 may be referred to as a CPU, DSP, GPU, etc., or any other type of hardware circuitry. Each core 3202 includes control unit circuitry 3214, arithmetic and logic (AL) circuitry (sometimes referred to as an ALU) 3216, a plurality of registers 3218, the L1 cache 3220, and a second example bus 3222. Other structures may be present. For example, each core 3202 may include vector unit circuitry, single instruction multiple data (SIMD) unit circuitry, load/store unit (LSU) circuitry, branch/jump unit circuitry, floating-point unit (FPU) circuitry, etc. The control unit circuitry 3214 includes semiconductor-based circuits structured to control (e.g., coordinate) data movement within the corresponding core 3202. The AL circuitry 3216 includes semiconductor-based circuits structured to perform one or more mathematic and/or logic operations on the data within the corresponding core 3202. The AL circuitry 3216 of some examples performs integer based operations. In other examples, the AL circuitry 3216 also performs floating point operations. In yet other examples, the AL circuitry 3216 may include first AL circuitry that performs integer based operations and second AL circuitry that performs floating point operations. In some examples, the AL circuitry 3216 may be referred to as an Arithmetic Logic Unit (ALU). The registers 3218 are semiconductor-based structures to store data and/or instructions such as results of one or more of the operations performed by the AL circuitry 3216 of the corresponding core 3202. For example, the registers 3218 may include vector register(s), SIMD register(s), general purpose register(s), flag register(s), segment register(s), machine specific register(s), instruction pointer register(s), control register(s), debug register(s), memory management register(s), machine check register(s), etc. The registers 3218 may be arranged in a bank as shown in FIG. 32. Alternatively, the registers 3218 may be organized in any other arrangement, format, or structure including distributed throughout the core 3202 to shorten access time. The second bus 3222 may implement at least one of an I2C bus, a SPI bus, a PCI bus, or a PCIe bus.

Each core 3202 and/or, more generally, the microprocessor 3200 may include additional and/or alternate structures to those shown and described above. For example, one or more clock circuits, one or more power supplies, one or more power gates, one or more cache home agents (CHAs), one or more converged/common mesh stops (CMSs), one or more shifters (e.g., barrel shifter(s)) and/or other circuitry may be present. The microprocessor 3200 is a semiconductor device fabricated to include many transistors interconnected to implement the structures described above in one or more integrated circuits (ICs) contained in one or more packages. The processor circuitry may include and/or cooperate with one or more accelerators. In some examples, accelerators are implemented by logic circuitry to perform certain tasks more quickly and/or efficiently than can be done by a general purpose processor. Examples of accelerators include ASICs and FPGAs such as those discussed herein. A GPU or other programmable device can also be an accelerator. Accelerators may be on-board the processor circuitry, in the same chip package as the processor circuitry and/or in one or more separate packages from the processor circuitry.

FIG. 33 is a block diagram of another example implementation of the processor circuitry 3312 of FIG. 33 and/or the processor circuitry 3112 of FIG. 31. In this example, the processor circuitry 3312 and/or the processor circuitry 3112 is implemented by FPGA circuitry 3300. The FPGA circuitry 3300 can be used, for example, to perform operations that could otherwise be performed by the example microprocessor 3000 of FIG. 30 executing corresponding machine readable instructions. However, once configured, the FPGA circuitry 3300 instantiates the machine readable instructions in hardware and, thus, can often execute the operations faster than they could be performed by a general purpose microprocessor executing the corresponding software.

More specifically, in contrast to the microprocessor 3200 of FIG. 32 described above (which is a general purpose device that may be programmed to execute some or all of the machine readable instructions represented by the flowcharts of FIGS. 7, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, and/or 27 but whose interconnections and logic circuitry are fixed once fabricated), the FPGA circuitry 3300 of the example of FIG. 33 includes interconnections and logic circuitry that may be configured and/or interconnected in different ways after fabrication to instantiate, for example, some or all of the machine readable instructions represented by the flowcharts of FIGS. 7, and/or 17-29. In particular, the FPGA 3300 may be thought of as an array of logic gates, interconnections, and switches. The switches can be programmed to change how the logic gates are interconnected by the interconnections, effectively forming one or more dedicated logic circuits (unless and until the FPGA circuitry 3300 is reprogrammed). The configured logic circuits enable the logic gates to cooperate in different ways to perform different operations on data received by input circuitry. Those operations may correspond to some or all of the software represented by the flowcharts of FIGS. 7, and/or 17-29. As such, the FPGA circuitry 3300 may be structured to effectively instantiate some or all of the machine readable instructions of the flowcharts of FIGS. 7, and/or 17-29 as dedicated logic circuits to perform the operations corresponding to those software instructions in a dedicated manner analogous to an ASIC. Therefore, the FPGA circuitry 3300 may perform the operations corresponding to the some or all of the machine readable instructions of FIGS. 7, and/or 17-29 faster than the general purpose microprocessor can execute the same.

In the example of FIG. 33, the FPGA circuitry 3300 is structured to be programmed (and/or reprogrammed one or more times) by an end user by a hardware description language (HDL) such as Verilog. The FPGA circuitry 3300 of FIG. 33, includes example input/output (I/O) circuitry 3302 to obtain and/or output data to/from example configuration circuitry 3304 and/or external hardware (e.g., external hardware circuitry) 3306. For example, the configuration circuitry 3304 may implement interface circuitry that may obtain machine readable instructions to configure the FPGA circuitry 3300, or portion(s) thereof. In some such examples, the configuration circuitry 3304 may obtain the machine readable instructions from a user, a machine (e.g., hardware circuitry (e.g., programmed or dedicated circuitry) that may implement an Artificial Intelligence/Machine Learning (AI/ML) model to generate the instructions), etc. In some examples, the external hardware 3306 may implement the microprocessor 3200 of FIG. 32. The FPGA circuitry 3300 also includes an array of example logic gate circuitry 3308, a plurality of example configurable interconnections 3310, and example storage circuitry 3312. The logic gate circuitry 3308 and interconnections 3310 are configurable to instantiate one or more operations that may correspond to at least some of the machine readable instructions of FIGS. 7 and/or 17-29 and/or other desired operations. The logic gate circuitry 3308 shown in FIG. 33 is fabricated in groups or blocks. Each block includes semiconductor-based electrical structures that may be configured into logic circuits. In some examples, the electrical structures include logic gates (e.g., And gates, Or gates, Nor gates, etc.) that provide basic building blocks for logic circuits. Electrically controllable switches (e.g., transistors) are present within each of the logic gate circuitry 3308 to enable configuration of the electrical structures and/or the logic gates to form circuits to perform desired operations. The logic gate circuitry 3308 may include other electrical structures such as look-up tables (LUTs), registers (e.g., flip-flops or latches), multiplexers, etc.

The interconnections 3310 of the illustrated example are conductive pathways, traces, vias, or the like that may include electrically controllable switches (e.g., transistors) whose state can be changed by programming (e.g., using an HDL instruction language) to activate or deactivate one or more connections between one or more of the logic gate circuitry 3308 to program desired logic circuits.

The storage circuitry 3312 of the illustrated example is structured to store result(s) of the one or more of the operations performed by corresponding logic gates. The storage circuitry 3312 may be implemented by registers or the like. In the illustrated example, the storage circuitry 3312 is distributed amongst the logic gate circuitry 3308 to facilitate access and increase execution speed.

The example FPGA circuitry 3300 of FIG. 33 also includes example Dedicated Operations Circuitry 3314. In this example, the Dedicated Operations Circuitry 3314 includes special purpose circuitry 3316 that may be invoked to implement commonly used functions to avoid the need to program those functions in the field. Examples of such special purpose circuitry 3316 include memory (e.g., DRAM) controller circuitry, PCIe controller circuitry, clock circuitry, transceiver circuitry, memory, and multiplier-accumulator circuitry. Other types of special purpose circuitry may be present. In some examples, the FPGA circuitry 3300 may also include example general purpose programmable circuitry 3318 such as an example CPU 3320 and/or an example DSP 3322. Other general purpose programmable circuitry 3318 may additionally or alternatively be present such as a GPU, an XPU, etc., that can be programmed to perform other operations.

Although FIGS. 32 and 33 illustrate two example implementations of the processor circuitry 3012 of FIG. 30 and/or the processor circuitry 3112 of FIG. 31, many other approaches are contemplated. For example, as mentioned above, modern FPGA circuitry may include an on-board CPU, such as one or more of the example CPU 3320 of FIG. 33. Therefore, the processor circuitry 3012 of FIG. 30 and/or the processor circuitry 3112 of FIG. 31 may additionally be implemented by combining the example microprocessor 3200 of FIG. 32 and the example FPGA circuitry 3300 of FIG. 33. In some such hybrid examples, a first portion of the machine readable instructions represented by the flowcharts of FIGS. 7 and/or 17-29 may be executed by one or more of the cores 3202 of FIG. 32, a second portion of the machine readable instructions represented by the flowcharts of FIGS. 7 and/or 17-29 may be executed by the FPGA circuitry 3300 of FIG. 33, and/or a third portion of the machine readable instructions represented by the flowcharts of FIGS. 7 and/or 17-29 may be executed by an ASIC. It should be understood that some or all of the data usage monitoring circuitry 1300 of FIG. 13 may, thus, be instantiated at the same or different times. Some or all of the circuitry may be instantiated, for example, in one or more threads executing concurrently and/or in series. Moreover, in some examples, some or all of the data usage monitoring circuitry 1300 of FIG. 13 may be implemented within one or more virtual machines and/or containers executing on the microprocessor.

In some examples, the processor circuitry 3312 of FIG. 33 and/or the processor circuitry 3112 of FIG. 31 may be in one or more packages. For example, the processor circuitry 3200 of FIG. 32 and/or the FPGA circuitry 3300 of FIG. 33 may be in one or more packages. In some examples, an XPU may be implemented by the processor circuitry 3312 of FIG. 33 and/or the processor circuitry 3112 of FIG. 31, which may be in one or more packages. For example, the XPU may include a CPU in one package, a DSP in another package, a GPU in yet another package, and an FPGA in still yet another package.

A block diagram illustrating an example software distribution platform 3405 to distribute software such as the example machine readable instructions 3032 of FIG. 30 and/or the example machine readable instructions 3132 of FIG. 31 to hardware devices owned and/or operated by third parties is illustrated in FIG. 34. The example software distribution platform 3405 may be implemented by any computer server, data facility, cloud service, etc., capable of storing and transmitting software to other computing devices. The third parties may be customers of the entity owning and/or operating the software distribution platform 3405. For example, the entity that owns and/or operates the software distribution platform 3405 may be a developer, a seller, and/or a licensor of software such as the example machine readable instructions 3032 of FIG. 30 and/or the example machine readable instructions 3132 of FIG. 31. The third parties may be consumers, users, retailers, OEMs, etc., who purchase and/or license the software for use and/or re-sale and/or sub-licensing. In the illustrated example, the software distribution platform 3405 includes one or more servers and one or more storage devices. The storage devices store the machine readable instructions 3032, which may correspond to the example machine readable instructions 700, 1700, 1800, 1900, 2000, 2100, 2200, 2300, 2400, 2500, 2600, 2700, 2800, 2900 of FIGS. 7 and/or 17-29, as described above. The one or more servers of the example software distribution platform 3405 are in communication with a network 3410, which may correspond to any one or more of the Internet and/or any of the example networks 110, 806, 808, 3029, 3126 described above. In some examples, the one or more servers are responsive to requests to transmit the software to a requesting party as part of a commercial transaction. Payment for the delivery, sale, and/or license of the software may be handled by the one or more servers of the software distribution platform and/or by a third party payment entity. The servers enable purchasers and/or licensors to download the machine readable instructions 3032, 3132 from the software distribution platform 3405. For example, the software, which may correspond to the example machine readable instructions 700, 1700, 1800, 1900, 2000, 2100, 2200, 2300, 2400, 2500, 2600, 2700, 2800, 2900 of FIGS. 7 and/or 17-29, may be downloaded to (i) the example processor platform 3000, which is to execute the machine readable instructions 3032 to implement the logical entity 601, and/or, more generally, the ADM system 600 of FIG. 6, and/or (ii) the example processor platform 3100, which is to execute the machine readable instructions 3132 to implement the data usage monitoring circuitry 1300 of FIG. 13. In some examples, one or more servers of the software distribution platform 3405 periodically offer, transmit, and/or force updates to the software (e.g., the example machine readable instructions 3032 of FIG. 30, the example machine readable instructions 3132 of FIG. 31, etc.) to ensure improvements, patches, updates, etc., are distributed and applied to the software at the end user devices.

From the foregoing, it will be appreciated that example systems, methods, apparatus, and articles of manufacture have been disclosed for proactive data management and analytics. Disclosed systems, methods, apparatus, and articles of manufacture achieve and/or otherwise implement improved data ingestion, bus, analytics, storage, data publishing, privacy, security, and trust techniques over conventional data management systems. Disclosed systems, methods, apparatus, and articles of manufacture improve the efficiency of using a computing device by implementing examples disclosed herein. Disclosed systems, methods, apparatus, and articles of manufacture are accordingly directed to one or more improvement(s) in the operation of a machine such as a computer or other electronic and/or mechanical device.

Further examples and combinations thereof include the following:

Example 1 includes an apparatus to monitor data usage, comprising interface circuitry to communicatively couple a network to processor circuitry, and the processor circuitry including one or more of at least one of a central processor unit, a graphics processor unit, or a digital signal processor, the at least one of the central processor unit, the graphics processor unit, or the digital signal processor having control circuitry to control data movement within the processor circuitry, arithmetic and logic circuitry to perform one or more first operations corresponding to instructions, and one or more registers to store a result of the one or more first operations, the instructions in the apparatus, a Field Programmable Gate Array (FPGA), the FPGA including logic gate circuitry, a plurality of configurable interconnections, and storage circuitry, the logic gate circuitry and the plurality of the configurable interconnections to perform one or more second operations, the storage circuitry to store a result of the one or more second operations, or Application Specific Integrated Circuitry (ASIC) including logic gate circuitry to perform one or more third operations, the processor circuitry to perform at least one of the first operations, the second operations, or the third operations to instantiate resource manager orchestration circuitry to orchestrate resources in an edge environment based on ingested network traffic on an edge network, at least some of the ingested network traffic associated with at least one source node that is to source a target data stream and at least one target artificial intelligence (AI) application node that is to consume at least a portion of the target data stream, and machine learning circuitry to execute a machine learning model based on the ingested network traffic to generate one or more outputs, the one or more outputs including at least one of a first value representative of a data stream characteristic or a second value representative of an AI application node characteristic, determine the one or more outputs satisfy a threshold value, and generate an alert in response to the outputs satisfying the threshold value.

Example 2 includes the apparatus of example 1, wherein the interface circuitry is to ingest at least a portion of the target data stream from the data source, tag the at least portion of the target data stream with metadata, and query an orchestrator to identify the machine learning model as associated with the metadata, and the machine learning circuitry is to execute the machine learning model to determine the at least one of the first value representative of the data stream characteristic or the second value representative of the AI application node characteristic.

Example 3 includes the apparatus of example 1, wherein the machine learning circuitry is to determine at least one of a content type of the target data stream, a sensitive attribute of the target data stream, a security level of the target data stream, or a source location of the target data stream, and execute the machine learning model to determine the first value of the data stream characteristic based on the at least one of the content type, the sensitive attribute, the security level, or the source location.

Example 4 includes the apparatus of example 3, wherein the target data stream includes one or more target data points associated with the at least one of the content type, the sensitive attribute, the security level, or the source location, and the processor circuitry is to perform the at least one of the first operations, the second operations, or the third operations to instantiate metadata manager circuitry to generate at least one target graph node representation of the target data stream based on at least one of the one or more target data points, and the machine learning circuitry is to execute the machine learning model to compare the at least one target graph node representation to one or more baseline graph node representations, the one or more baseline graph node representations including respective one or more nominal data points from nominal data streams, the one or more nominal data points including at least one of a content type of nominal data stream, a sensitive attribute of the nominal data stream, a security level of the nominal data stream, or a source location of the nominal data stream.

Example 5 includes the apparatus of example 1, wherein the machine learning circuitry is to determine at least one of a service type attribute of the AI application node or a usage context of the target data stream for the AI application node, and execute the machine learning model to determine the second value of the AI application node characteristic based on at least one of the service type attribute or the usage context.

Example 6 includes the apparatus of example 1, wherein the machine learning circuitry is to train the machine learning model with at least one of nominal traffic or nominal node behavior, the nominal traffic indicative of one or more nominal data streams with one or more expected data points, the nominal node behavior indicative of one or more expected data consumption patterns by one or more nominal nodes.

Example 7 includes the apparatus of example 1, wherein the one or more outputs are one or more first outputs, and the resource manager orchestration circuitry is to instantiate a first super node in the edge environment, and deploy a second instantiation of the machine learning model to the first super node, and the machine learning circuitry is to execute the second instantiation of the machine learning model based on a first plurality of data streams within network traffic ingested at the first super node to generate one or more second outputs, the one or more second outputs including values representative of data stream characteristics and values representative of AI application node characteristics, share the one or more second outputs with a second super node in the edge environment, obtain one or more third outputs from the second super node, the one or more third outputs generated from a third instantiation of the machine learning model executed at the second super node based on a second plurality of data streams within network traffic ingested at the second super node, the one or more third outputs including values representative of data stream characteristics and values representative of AI application node characteristics, and train the machine learning model using at least one of the one or more second outputs or the one or more third outputs to build a consensus nominal data stream pattern.

Example 8 includes the apparatus of example 1, wherein the machine learning model is a first machine learning model, the one or more outputs are one or more first outputs, the threshold value is a first threshold value, and the resource manager orchestration circuitry is to instantiate a deep data inspection node in the edge environment, the deep data inspection node to have access to the target data stream, the interface circuitry to verify at least one feature in the target data stream corresponds to at least one of a target data stream characteristic or a target AI application node characteristic, the processor circuitry is to perform the at least one of the first operations, the second operations, or the third operations to instantiate algorithm manager circuitry to select a second machine learning model trained on a feature set representative of at least one of at least one of the target data stream characteristic or the target AI application node characteristic, the machine learning circuitry is to execute the second machine learning model over a period of time at the deep data inspection node based on the target data stream to generate one or more second outputs, the one or more second outputs including at least a third value representative of a deviation condition of the target data stream, and the processor circuitry is to perform the at least one of the first operations, the second operations, or the third operations to instantiate deep data inspection circuitry to determine the one or more second outputs satisfy a second threshold value at least once over the period of time.

Example 9 includes the apparatus of example 8, wherein the resource manager orchestration circuitry is to deploy the trained second machine learning model across the deep data inspection node and one or more additional deep data inspection nodes in the edge environment, and the machine learning circuitry is to train the second machine learning model with at least one nominal data stream pattern shared across the deep data inspection node and the one or more additional deep data inspection nodes.

Example 10 includes the apparatus of example 9, wherein the deep data inspection circuitry is to at least one of determine the deviation condition of the target data stream occurs less than or equal to a frequency threshold, determine the deviation condition of the target data stream occurs less than or equal to a frequency threshold at the one or more additional deep data inspection nodes, or determine the deviation condition of the target data stream does not meet at least one constraint, and the processor circuitry to perform the at least one of the first operations, the second operations, or the third operations to instantiate operation execution circuitry to cause at least one of a modification to the target data stream or a response to an attempt to consume the target data stream.

Example 11 includes the apparatus of example 1, wherein the processor circuitry to perform the at least one of the first operations, the second operations, or the third operations to instantiate operation execution circuitry to, in response to the alert being generated, at least one of tag a portion of the target data stream with metadata, implement a blockchain for at least one data point in the target data stream, or prohibit consumption of the target data stream by the target AI application node.

Example 12 includes a non-transitory machine readable storage medium comprising instructions that, when executed, cause processor circuitry to at least orchestrate resources in an edge environment based on ingested network traffic on an edge network, at least some of the ingested network traffic associated with at least one source node that is to source a target data stream and at least one target artificial intelligence (AI) application node that is to consume at least a portion of the target data stream, execute a machine learning model based on the ingested network traffic to generate one or more outputs, the one or more outputs including at least one a first value representative of a data stream characteristic or a second value representative of an AI application node characteristic, determine the one or more outputs satisfy a threshold value, and generate an alert in response to the outputs satisfying the threshold value.

Example 13 includes the non-transitory machine readable storage medium 12, wherein the instructions, when executed, further cause the processor circuitry to ingest at least a portion of the target data stream from the data source, tag the at least portion of the target data stream with metadata, query an orchestrator to identify the machine learning model as associated with the metadata, and execute the machine learning model to determine the at least one of the first value representative of the data stream characteristic or the second value representative of the AI application node characteristic.

Example 14 includes the non-transitory machine readable storage medium 12, wherein the instructions, when executed, further cause the processor circuitry to determine at least one of a content type of the target data stream, a sensitive attribute of the target data stream, a security level of the target data stream, or a source location of the target data stream, and execute the machine learning model to determine the first value of the data stream characteristic based on the at least one of the content type, the sensitive attribute, the security level, or the source location.

Example 15 includes the non-transitory machine readable storage medium of example 12, wherein the target data stream includes one or more target data points associated with at least one of a content type, a sensitive attribute, a security level, or a source location, and wherein the instructions, when executed, further cause the processor circuitry to generate at least one target graph node representation of the target data stream based on at least one of the one or more target data points, and execute the machine learning model to compare the at least one target graph node representation to one or more baseline graph node representations, the one or more baseline graph node representations including respective one or more nominal data points from nominal data streams, the one or more nominal data points including at least one of a content type of nominal data stream, a sensitive attribute of the nominal data stream, a security level of the nominal data stream, or a source location of the nominal data stream.

Example 16 includes the non-transitory machine readable storage medium of example 12, wherein the instructions, when executed, further cause the processor circuitry to determine at least one of a service type attribute of the AI application node or a usage context of the target data stream for the AI application node, and execute the machine learning model to determine the second value of the AI application node characteristic based on at least one of the service type attribute or the usage context.

Example 17 includes the non-transitory machine readable storage medium of example 12, wherein the instructions, when executed, further cause the processor circuitry to select at least one policy for the ingested network traffic used to initiate the machine learning model, and train the machine learning model with at least one of nominal traffic or nominal node behavior, the nominal traffic indicative of one or more nominal data streams with one or more expected data points, the nominal node behavior indicative of one or more expected data consumption patterns by one or more nominal nodes.

Example 18 includes the non-transitory machine readable storage medium of example 12, wherein the instructions, when executed, further cause the processor circuitry to instantiate a first super node in the edge environment, and deploy a second instantiation of the machine learning model to the first super node, execute the second instantiation of the machine learning model based on a first plurality of data streams within network traffic ingested at the first super node to generate one or more second outputs, the one or more second outputs including values representative of data stream characteristics and values representative of AI application node characteristics, share the one or more second outputs with a second super node in the edge environment, obtain one or more third outputs from the second super node, the one or more third outputs generated from a third instantiation of the machine learning model executed at the second super node based on a second plurality of data streams within network traffic ingested at the second super node, the one or more third outputs including values representative of data stream characteristics and values representative of AI application node characteristics, and train the machine learning model using at least one of the one or more second outputs or the one or more third outputs to build a consensus nominal data stream pattern.

Example 19 includes the non-transitory machine readable storage medium of example 12, wherein the machine learning model is a first machine learning model, wherein the outputs are first outputs, wherein the threshold value is a first threshold value, and wherein the instructions, when executed, further cause the processor circuitry to instantiate a deep data inspection node in the edge environment, the deep data inspection node to have access to the target data stream, verify at least one feature in the target data stream corresponds to at least one of a target data stream characteristic or a target AI application node characteristic, select a second machine learning model trained on a feature set representative of at least one of at least one of the target data stream characteristic or the target AI application node characteristic, execute the second machine learning model over a period of time at the deep data inspection node based on the target data stream to generate one or more second outputs, the one or more second outputs including at least a third value representative of a deviation condition of the target data stream, and determine the one or more second outputs satisfy a second threshold value at least once over the period of time.

Example 20 includes the non-transitory machine readable storage medium of example 19, wherein the instructions, when executed, further cause the processor circuitry to deploy the trained second machine learning model across the deep data inspection node and one or more additional deep data inspection nodes in the edge environment, and train the second machine learning model with at least one nominal data stream pattern shared across the deep data inspection node and the one or more additional deep data inspection nodes.

Example 21 includes the non-transitory machine readable storage medium of example 19, wherein the instructions, when executed, further cause the processor circuitry to at least one of determine the deviation condition of the target data stream occurs less than or equal to a frequency threshold, determine the deviation condition of the target data stream occurs less than or equal to a frequency threshold at the one or more additional deep data inspection nodes, or determine the deviation condition of the target data stream does not meet at least one constraint, and cause at least one of a modification to the target data stream or a response to an attempt to consume the target data stream.

Example 22 includes the non-transitory machine readable storage medium of example 12, wherein the instructions, when executed, further cause the processor circuitry to in response to the alert being generated, at least one of tag a portion of the target data stream with metadata, implement a blockchain for at least one data point in the target data stream, or prohibit consumption of the target data stream by the target AI application node.

Example 23 includes a method, comprising orchestrating resources in an edge environment based on ingested network traffic on an edge network, at least some of the ingested network traffic associated with at least one source node that is to source a target data stream and at least one target artificial intelligence (AI) application node that is to consume at least a portion of the target data stream, and executing a machine learning model based on the ingested network traffic to generate one or more outputs, the one or more outputs including at least one a first value representative of a data stream characteristic or a second value representative of an AI application node characteristic, determining the one or more outputs satisfy a threshold value, and generating an alert in response to the outputs satisfying the threshold value.

Example 24 includes the method of example 23, including determining at least one of a content type of the target data stream, a sensitive attribute of the target data stream, a security level of the target data stream, or a source location of the target data stream, and executing the machine learning model to determine the first value of the data stream characteristic based on the at least one of the content type, the sensitive attribute, the security level, or the source location.

Example 25 includes the method of example 23, including determining at least one of a service type attribute of the AI application node or a usage context of the target data stream for the AI application node, and executing the machine learning model to determine the second value of the AI application node characteristic based on at least one of the service type attribute or the usage context.

Example 26 includes the method of example 23, including ingesting at least a portion of the target data stream from the data source, tagging the at least portion of the target data stream with metadata, querying an orchestrator to identify the machine learning model as associated with the metadata, and executing the machine learning model to determine the at least one of the first value representative of the data stream characteristic or the second value representative of the AI application node characteristic.

Example 27 includes the method of example 23, wherein the target data stream includes one or more target data points associated with at least one of a content type, a sensitive attribute, a security level, or a source location, including generating at least one target graph node representation of the target data stream based on at least one of the one or more target data points, and executing the machine learning model to compare the at least one target graph node representation to one or more baseline graph node representations, the one or more baseline graph node representations including respective one or more nominal data points from nominal data streams, the one or more nominal data points including at least one of a content type of nominal data stream, a sensitive attribute of the nominal data stream, a security level of the nominal data stream, or a source location of the nominal data stream.

Example 28 includes the method of example 23, including selecting at least one policy for the ingested network traffic used to initiate the machine learning model, and training the machine learning model with at least one of nominal traffic or nominal node behavior, the nominal traffic indicative of one or more nominal data streams with one or more expected data points, the nominal node behavior indicative of one or more expected data consumption patterns by one or more nominal nodes.

Example 29 includes the method of example 23, including instantiating a first super node in the edge environment, and deploying a second instantiation of the machine learning model to the first super node, executing the second instantiation of the machine learning model based on a first plurality of data streams within network traffic ingested at the first super node to generate one or more second outputs, the one or more second outputs including values representative of data stream characteristics and values representative of AI application node characteristics, sharing the one or more second outputs with a second super node in the edge environment, obtaining one or more third outputs from the second super node, the one or more third outputs generated from a third instantiation of the machine learning model executed at the second super node based on a second plurality of data streams within network traffic ingested at the second super node, the one or more third outputs including values representative of data stream characteristics and values representative of AI application node characteristics, and training the machine learning model using at least one of the one or more second outputs or the one or more third outputs to build a consensus nominal data stream pattern.

Example 30 includes the method of example 23, wherein the machine learning model is a first machine learning model, wherein the outputs are first outputs, wherein the threshold value is a first threshold value, including instantiating a deep data inspection node in the edge environment, the deep data inspection node to have access to the target data stream, verifying at least one feature in the target data stream corresponds to at least one of a target data stream characteristic or a target AI application node characteristic, selecting a second machine learning model trained on a feature set representative of at least one of at least one of the target data stream characteristic or the target AI application node characteristic, executing the second machine learning model over a period of time at the deep data inspection node based on the target data stream to generate one or more second outputs, the one or more second outputs including at least a third value representative of a deviation condition of the target data stream, and determining the one or more second outputs satisfy a second threshold value at least once over the period of time.

Example 31 includes the method of example 30, including deploying the trained second machine learning model across the deep data inspection node and one or more additional deep data inspection nodes in the edge environment, and training the second machine learning model with at least one nominal data stream pattern shared across the deep data inspection node and the one or more additional deep data inspection nodes.

Example 32 includes the method of example 30, including at least one of determining the deviation condition of the target data stream occurs less than or equal to a frequency threshold, determining the deviation condition of the target data stream occurs less than or equal to a frequency threshold at the one or more additional deep data inspection nodes, or determining the deviation condition of the target data stream does not meet at least one constraint, and causing at least one of a modification to the target data stream or a response to an attempt to consume the target data stream.

Example 33 includes the method of example 23, including in response to the alert being generated, at least one of tagging a portion of the target data stream with metadata, implementing a blockchain for at least one data point in the target data stream, or prohibiting consumption of the target data stream by the target AI application node.

Example 34 is edge server processor circuitry to perform the method of any of Examples 23-33.

Example 35 is an edge cloud processor circuitry to perform the method of any of Examples 23-33.

Example 31 is edge node processor circuitry to perform the method of any of Examples 23-33.

Example 32 is dedicated private network circuitry to perform the method of any of Examples 23-33.

Example 33 is a programmable location data collector to perform the method of any of Examples 23-33.

Example 34 is an apparatus comprising processor circuitry to perform the method of any of Examples 23-33.

Example 35 is an apparatus comprising one or more edge gateways to perform the method of any of Examples 23-33.

Example 36 is an apparatus comprising one or more edge switches to perform the method of any of Examples 23-33.

Example 37 is an apparatus comprising at least one of one or more edge gateways or one or more edge switches to perform the method of any of Examples 23-33.

Example 38 is an apparatus comprising accelerator circuitry to perform the method of any of Examples 23-33.

Example 39 is an apparatus comprising one or more graphics processor units to perform the method of any of Examples 23-33.

Example 40 is an apparatus comprising one or more Artificial Intelligence processors to perform the method of any of Examples 23-33.

Example 41 is an apparatus comprising one or more machine learning processors to perform the method of any of Examples 23-33.

Example 42 is an apparatus comprising one or more neural network processors to perform the method of any of Examples 23-33.

Example 43 is an apparatus comprising one or more digital signal processors to perform the method of any of Examples 23-33.

Example 44 is an apparatus comprising one or more general purpose processors to perform the method of any of Examples 23-33.

Example 45 is an apparatus comprising network interface circuitry to perform the method of any of Examples 23-33.

Example 46 is an Infrastructure Processor Unit to perform the method of any of Examples 23-33.

Example 47 is hardware queue management circuitry to perform the method of any of Examples 23-33.

Example 48 is at least one of remote radio unit circuitry or radio access network circuitry to perform the method of any of Examples 23-33.

Example 49 is base station circuitry to perform the method of any of Examples 23-33.

Example 50 is user equipment circuitry to perform the method of any of Examples 23-33.

Example 51 is an Internet of Things device to perform the method of any of Examples 23-33.

Example 52 is a software distribution platform to distribute machine-readable instructions that, when executed by processor circuitry, cause the processor circuitry to perform the method of any of Examples 23-33.

Example 53 is edge cloud circuitry to perform the method of any of Examples 23-33.

Example 54 is distributed unit circuitry to perform the method of any of Examples 23-33.

Example 55 is control unit circuitry to perform the method of any of Examples 23-33.

Example 56 is core server circuitry to perform the method of any of Examples 23-33.

Example 57 is satellite circuitry to perform the method of any of Examples 23-33.

The following claims are hereby incorporated into this Detailed Description by this reference. Although certain example systems, methods, apparatus, and articles of manufacture have been disclosed herein, the scope of coverage of this patent is not limited thereto. On the contrary, this patent covers all systems, methods, apparatus, and articles of manufacture fairly falling within the scope of the claims of this patent.

SYSTEMS, APPARATUS, ARTICLES OF MANUFACTURE, AND METHODS FOR DATA USAGE MONITORING TO IDENTIFY AND MITIGATE ETHICAL DIVERGENCE

Information

Publication Number

Date Filed

Date Published

Inventors

Original Assignees

CPC

International Classifications

Abstract

Description

Claims

RELATED APPLICATION

Provisional Applications (1)