Technologies for annotating process and user information for network flows

Abstract

Systems, methods, and computer-readable media for annotating process and user information for network flows. In some embodiments, a capturing agent, executing on a first device in a network, can monitor a network flow associated with the first device. The first device can be, for example, a virtual machine, a hypervisor, a server, or a network device. Next, the capturing agent can generate a control flow based on the network flow. The control flow may include metadata that describes the network flow. The capturing agent can then determine which process executing on the first device is associated with the network flow and label the control flow with this information. Finally, the capturing agent can transmit the labeled control flow to a second device, such as a collector, in the network.

Description

TECHNICAL FIELD

The present technology pertains to network analytics, and more specifically to annotating process and user information in a network environment.

BACKGROUND

In a network environment, capturing agents or sensors can be placed at various devices or elements in the network to collect flow data and network statistics from different locations. The collected data from the capturing agents can be analyzed to monitor and troubleshoot the network. The data collected from the capturing agents can provide valuable details about the status, security, or performance of the network, as well as any network elements. Information about the capturing agents can also help interpret the data from the capturing agents, in order to infer or ascertain additional details from the collected data. For example, understanding the placement of a capturing agent relative to other capturing agents in the network can provide a context to the data reported by the capturing agents, which can further help identify specific patterns or conditions in the network. Unfortunately, however, information gathered from the capturing agents distributed throughout the network is often limited and may not include certain types of useful information. Moreover, as the network grows and changes, the information can quickly become outdated.

As data centers grow in size and complexity, the tools that manage them must be able to effectively identify inefficiencies while implementing appropriate security policies. Traditionally, network administrators have to manually implement security policies, manage access control lists (ACLs), configure firewalls, identify misconfigured or infected machines, etc. These tasks can become exponentially more complicated as a network grows in size and require an intimate knowledge of a large number of data center components. Furthermore, malicious attacks or misconfigured machines can shut down a data center within minutes while it could take a network administrator hours or days to determine the root problem and provide a solution. What is needed is a broad and deep network monitoring system that can automatically determine the network topology, map application dependencies, monitor traffic flow, dynamically analyze network performance, identify problems, implement policies, and present a network administrator with an interface reflecting the current state of the data center. The traffic monitoring system herein disclosed can provide such functionality.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to describe the manner in which the above-recited and other advantages and features of the disclosure can be obtained, a more particular description of the principles briefly described above will be rendered by reference to specific embodiments thereof which are illustrated in the appended drawings. Understanding that these drawings depict only exemplary embodiments of the disclosure and are not therefore to be considered to be limiting of its scope, the principles herein are described and explained with additional specificity and detail through the use of the accompanying drawings in which:

FIG. 1A illustrates a diagram of an example network environment;

FIG. 1B illustrates an example traffic monitoring system;

FIG. 2A illustrates a schematic diagram of an example capturing agent deployment in a virtualized environment;

FIG. 2B illustrates a schematic diagram of an example capturing agent deployment in an example network device;

FIG. 2C illustrates a schematic diagram of an example reporting system in an example capturing agent topology;

FIGS. 3A through 3F illustrate schematic diagrams of example configurations for reporting flows captured by capturing agents in an example capturing agent topology;

FIG. 4 illustrates a schematic diagram of an example configuration for collecting capturing agent reports;

FIG. 5 illustrates a diagram of an example capturing agent reporting process;

FIG. 6 illustrates a table of an example mapping of flow reports to capturing agents;

FIG. 7 illustrates a listing of example fields on a capturing agent report;

FIG. 8 illustrates an example method embodiment related to process information;

FIG. 9 illustrates an example method embodiment related to user information;

FIG. 10 illustrates an example network device; and

FIGS. 11A and 11B illustrate example system embodiments.

DESCRIPTION OF EXAMPLE EMBODIMENTS

Various embodiments of the disclosure are discussed in detail below. While specific implementations are discussed, it should be understood that this is done for illustration purposes only. A person skilled in the relevant art will recognize that other components and configurations may be used without parting from the spirit and scope of the disclosure.

Overview

Additional features and advantages of the disclosure will be set forth in the description which follows, and in part will be obvious from the description, or can be learned by practice of the herein disclosed principles. The features and advantages of the disclosure can be realized and obtained by means of the instruments and combinations particularly pointed out in the appended claims. These and other features of the disclosure will become more fully apparent from the following description and appended claims, or can be learned by the practice of the principles set forth herein.

The approaches set forth herein can be used to annotate process and user information related to network flows captured by various capturing agents or sensors deployed throughout a virtualized compute environment. The capturing agents can be packet inspection sensors configured to monitor, capture, and/or report network traffic information at the various locations. The capturing agents can be deployed on virtual machines, hypervisors, servers, and network devices (e.g., physical switches) on the network. The various capturing agents can capture traffic from their respective locations (e.g., traffic processed by their hosts), and report captured data to one or more devices, such as a collector system or a processing engine. The captured data can include any traffic and/or process information captured by the capturing agents including reports or control flows generated by other capturing agents.

The data reported from the various capturing agents can be used to determine the particular process or user involved with a given flow being reported. For example, capturing agents deployed throughout the network can be configured to identify the process or operating system user account that is responsible for generating or processing a network flow and report such findings to a collector in the form of a control flow. The reported process and user information can be used to understand the relationships of the flows and the corresponding processes and users, and may drive further analytics on the network.

A flow is conventionally represented as a 5-tuple comprising a source address, destination address, source port, destination port, and protocol. Thus, if a user desired to search flow data, the user could only search based on these attributes.

NetFlow exposes other attributes of flows but none of the additional attributes of this invention nor does NetFlow enable users to customize the attributes of flows. A flow can be tagged with metadata to provide additional information about the flow such that the flows are searchable based on tags, or flows having common tags can be aggregated to visualize flow data. Users can also define custom tags and rules by which flows should be tagged.

Advantages include: capable of searching flows based on tags; enable improved visualization of flows. Industry use: can be by public cloud competitors (of Nimbus/CCS) (e.g., Amazon, Google, Microsoft, Rackspace, Oracle, etc.). Product documentation, UI, claims that a product allows a user to search on flows based on non-conventional attributes or visualize flows according to non-conventional attributes.

Disclosed are systems, methods, and computer-readable storage media for annotating process and user information in a network. A system may include a virtual machine, a hypervisor hosting the virtual machine, and a network device such as a switch communicatively connected to the hypervisor. The virtual machine can have a first capturing agent or sensor that is configured to monitor a first network flow associated with the virtual machine. The first capturing agent can generate a first control flow based on the first network flow. The first control flow can include first metadata that describes the first network flow. The first capturing agent can label the first control flow with a first identifier of a first process executing on the virtual machine, thus yielding a first labeled control flow. The first process can be associated with the first network flow. The first capturing agent can then transmit the labeled control flow to a collector via the network.

The hypervisor may also have a second capturing agent. The second capturing agent can be configured to monitor a second network flow associated with the hypervisor, and the second network flow can include at least the first labeled control flow. The second capturing agent can generate a second control flow based on the second network flow. The second control flow can include second metadata that describes the second network flow. The second control flow can then label the second control flow with a second identifier of a second process executing on the hypervisor, thus yielding a second labeled control flow. The second process can be associated with the second network flow. Next, the second capturing agent can transmit the second labeled control flow to the collector via the network.

In addition, the network device can have a third capturing agent that is configured to monitor a third network flow associated with the network device. The third network flow can include the first labeled control flow and/or the second labeled control flow. The third capturing agent can generate a third control flow based on the third network flow, and the third control flow may include third metadata describing the third network flow. The third capturing agent can then label the third control flow with a third identifier of a third process that is executing on the network device and associated with the third network flow, thus yielding a third labeled control flow. Finally, the third capturing agent can transmit the third labeled control flow to the collector via the network.

DESCRIPTION

The disclosed technology addresses the need in the art for understanding data reported from capturing agents on a virtualized network. Disclosed are systems, methods, and computer-readable storage media for determining relative placement and topology of capturing agents deployed throughout a network. A description of an example network environment, as illustrated in FIG. 1, is first disclosed herein. A discussion of capturing agents and capturing agent topologies in virtualized environments, as illustrated in FIGS. 2A-C, will then follow. The discussion follows with a discussion of mechanisms for determining relative placement and topology information for capturing agents in a network environment, as illustrated in FIGS. 3-7. Then, example methods practiced according to the various embodiments disclosed herein will be discussed, as illustrated in FIGS. 9-10. The discussion then concludes with a description of example devices, as illustrated in FIGS. 10 and 11A-B. These variations shall be described herein as the various embodiments are set forth. The disclosure now turns to FIG. 1.

FIG. 1 illustrates a diagram of example network environment 100. Fabric 112 can represent the underlay (i.e., physical network) of network environment 100. Fabric 112 can include spine routers 1-N (102_{A_N}) (collectively “102”) and leaf routers 1-N (104A-N) (collectively “104”). Leaf routers 104 can reside at the edge of fabric 112, and can thus represent the physical network edges. Leaf routers 104 can be, for example, top-of-rack (“ToR”) switches, aggregation switches, gateways, ingress and/or egress switches, provider edge devices, and/or any other type of routing or switching device.

Leaf routers 104 can be responsible for routing and/or bridging tenant or endpoint packets and applying network policies. Spine routers 102 can perform switching and routing within fabric 112. Thus, network connectivity in fabric 112 can flow from spine routers 102 to leaf routers 104, and vice versa.

Leaf routers 104 can provide servers 1-5 (106_A-E) (collectively “106”), hypervisors 1-4 (108_A-108_D) (collectively “108”), and virtual machines (VMs) 1-5 (110_A-110_E) (collectively “110”) access to fabric 112. For example, leaf routers 104 can encapsulate and decapsulate packets to and from servers 106 in order to enable communications throughout environment 100. Leaf routers 104 can also connect other devices, such as device 114, with fabric 112. Device 114 can be any network-capable device(s) or network(s), such as a firewall, a database, a server, a collector 118 (further described below), an engine 120 (further described below), etc. Leaf routers 104 can also provide any other servers, resources, endpoints, external networks, VMs, services, tenants, or workloads with access to fabric 112.

VMs 110 can be virtual machines hosted by hypervisors 108 running on servers 106. VMs 110 can include workloads running on a guest operating system on a respective server. Hypervisors 108 can provide a layer of software, firmware, and/or hardware that creates and runs the VMs 110. Hypervisors 108 can allow VMs 110 to share hardware resources on servers 106, and the hardware resources on servers 106 to appear as multiple, separate hardware platforms. Moreover, hypervisors 108 and servers 106 can host one or more VMs 110. For example, server 106_A and hypervisor 108_A can host VMs 11 O_A-B.

In some cases, VMs 110 and/or hypervisors 108 can be migrated to other servers 106. For example, VM 110_A can be migrated to server 106c and hypervisor 108B. Servers 106 can similarly be migrated to other locations in network environment 100. For example, a server connected to a specific leaf router can be changed to connect to a different or additional leaf router. In some cases, some or all of servers 106, hypervisors 108, and/or VMs 110 can represent tenant space. Tenant space can include workloads, services, applications, devices, and/or resources that are associated with one or more clients or subscribers. Accordingly, traffic in network environment 100 can be routed based on specific tenant policies, spaces, agreements, configurations, etc. Moreover, addressing can vary between one or more tenants. In some configurations, tenant spaces can be divided into logical segments and/or networks and separated from logical segments and/or networks associated with other tenants.

Any of leaf routers 104, servers 106, hypervisors 108, and VMs 110 can include capturing agent 116 (also referred to as a “sensor”) configured to capture network data, and report any portion of the captured data to collector 118. Capturing agents 116 can be processes, agents, modules, drivers, or components deployed on a respective system (e.g., a server, VM, hypervisor, leaf router, etc.), configured to capture network data for the respective system (e.g., data received or transmitted by the respective system), and report some or all of the captured data to collector 118.

For example, a VM capturing agent can run as a process, kernel module, or kernel driver on the guest operating system installed in a VM and configured to capture data (e.g., network and/or system data) processed (e.g., sent, received, generated, etc.) by the VM. Additionally, a hypervisor capturing agent can run as a process, kernel module, or kernel driver on the host operating system installed at the hypervisor layer and configured to capture data (e.g., network and/or system data) processed (e.g., sent, received, generated, etc.) by the hypervisor. A server capturing agent can run as a process, kernel module, or kernel driver on the host operating system of a server and configured to capture data (e.g., network and/or system data) processed (e.g., sent, received, generated, etc.) by the server. And a network device capturing agent can run as a process or component in a network device, such as leaf routers 104, and configured to capture data (e.g., network and/or system data) processed (e.g., sent, received, generated, etc.) by the network device.

Capturing agents 116 or sensors can be configured to report the observed data and/or metadata about one or more packets, flows, communications, processes, events, and/or activities to collector 118. For example, capturing agents 116 can capture network data as well as information about the system or host of the capturing agents 116 (e.g., where the capturing agents 116 are deployed). Such information can also include, for example, data or metadata of active or previously active processes of the system, operating system user identifiers, metadata of files on the system, system alerts, networking information, etc. Capturing agents 116 may also analyze all the processes running on the respective VMs, hypervisors, servers, or network devices to determine specifically which process is responsible for a particular flow of network traffic. Similarly, capturing agents 116 may determine which operating system user(s) is responsible for a given flow. Reported data from capturing agents 116 can provide details or statistics particular to one or more tenants. For example, reported data from a subset of capturing agents 116 deployed throughout devices or elements in a tenant space can provide information about the performance, use, quality, events, processes, security status, characteristics, statistics, patterns, conditions, configurations, topology, and/or any other information for the particular tenant space.

Collectors 118 can be one or more devices, modules, workloads and/or processes capable of receiving data from capturing agents 116. Collectors 118 can thus collect reports and data from capturing agents 116. Collectors 118 can be deployed anywhere in network environment 100 and/or even on remote networks capable of communicating with network environment 100. For example, one or more collectors can be deployed within fabric 112 or on one or more of the servers 106. One or more collectors can be deployed outside of fabric 112 but connected to one or more leaf routers 104. Collectors 118 can be part of servers 106 and/or separate servers or devices (e.g., device 114). Collectors 118 can also be implemented in a cluster of servers.

Collectors 118 can be configured to collect data from capturing agents 116. In addition, collectors 118 can be implemented in one or more servers in a distributed fashion. As previously noted, collectors 118 can include one or more collectors. Moreover, each collector can be configured to receive reported data from all capturing agents 116 or a subset of capturing agents 116. For example, a collector can be assigned to a subset of capturing agents 116 so the data received by that specific collector is limited to data from the subset of capturing agents.

Collectors 118 can be configured to aggregate data from all capturing agents 116 and/or a subset of capturing agents 116. Moreover, collectors 118 can be configured to analyze some or all of the data reported by capturing agents 116. For example, collectors 118 can include analytics engines (e.g., engines 120) for analyzing collected data. Environment 100 can also include separate analytics engines 120 configured to analyze the data reported to collectors 118. For example, engines 120 can be configured to receive collected data from collectors 118 and aggregate the data, analyze the data (individually and/or aggregated), generate reports, identify conditions, compute statistics, visualize reported data, troubleshoot conditions, visualize the network and/or portions of the network (e.g., a tenant space), generate alerts, identify patterns, calculate misconfigurations, identify errors, generate suggestions, generate testing, and/or perform any other analytics functions. Analytics engines can determine dependencies of components within the network. For example, if component A routinely sends data to component B but component B never sends data to component A, then analytics engines can determine that component B is dependent on component A, but A is likely not dependent on component B. If, however, component B also sends data to component A, then they are likely interdependent. These components can be processes, virtual machines, hypervisors, VLANs, etc. Once an engine has determined component dependencies, it can then form a component (“application”) dependency map. This map can be instructive when analytics engines attempts to determine the root cause of a failure (because failure of one component can cascade and cause failure of its dependent components) or when analytics engines attempts to predict what will happen if a component is taken offline. Additionally, engines can associate edges of an application dependency map with expected latency, bandwidth, etc. for that individual edge. Analytics engines can establish patterns and norms for component behavior. For example, it can determine that certain processes (when functioning normally) will only send a certain amount of traffic to a certain VM using a small set of ports. Engines can establish these norms by analyzing individual components or by analyzing data coming from similar components (e.g., VMs with similar configurations). Similarly, engines can determine expectations for network operations. For example, it can determine the expected latency between two components, the expected throughput of a component, response times of a component, typical packet sizes, traffic flow signatures, etc. In some embodiments, engines can combine its dependency map with pattern analysis to create reaction expectations. For example, if traffic increases with one component, other components may predictably increase traffic in response (or latency, compute time, etc.

While collectors 118 and engines 120 are shown as separate entities, this is for illustration purposes as other configurations are also contemplated herein. For example, any of collectors 118 and engines 120 can be part of a same or separate entity. Moreover, any of the collector, aggregation, and analytics functions can be implemented by one entity (e.g., collectors 118) or separately implemented by multiple entities (e.g., engine 120 and/or collectors 118).

Each of the capturing agents 116 can use a respective address (e.g., internet protocol (IP) address, port number, etc.) of their host to send information to collectors 118 and/or any other destination. Collectors 118 may also be associated with their respective addresses such as IP addresses. Moreover, capturing agents 116 can periodically send information about flows they observe to collectors 118. Capturing agents 116 can be configured to report each and every flow they observe. Capturing agents 116 can report a list of flows that were active during a period of time (e.g., between the current time and the time of the last report). The consecutive periods of time of observance can be represented as pre-defined or adjustable time series. The series can be adjusted to a specific level of granularity. Thus, the time periods can be adjusted to control the level of details in statistics and can be customized based on specific requirements, such as security, scalability, storage, etc. The time series information can also be implemented to focus on more important flows or components (e.g., VMs) by varying the time intervals. The communication channel between a capturing agent and collector 118 can also create a flow in every reporting interval. Thus, the information transmitted or reported by capturing agents 116 can also include information about the flow created by the communication channel.

FIG. 1B depicts traffic monitoring system. This traffic monitoring system can comprise sensors 116, collectors 118, an analytics engine, and a presentation module.

Policy engine can configure analytics engine to establish what network policies exist or should be maintained. For example, policy engine may specify that certain machines should not intercommunicate or that certain ports are restricted. Network and security policy controller can set the parameters of policy engine. In some embodiments, policy engine is accessible via the presentation module.

Presentation module can comprise serving layer, authentication module, web front end, and public alert module connected to third party tools. As analytics engine processes the data and generates reports, they may not be in a human-readable form or they may be too large for an administrator to navigate. Presentation module can take the reports generated by analytics module and further summarize, filter, and organize the reports as well as create intuitive presentations of the reports.

Serving layer can be the interface between presentation module and analytics engine. As analytics module generates reports, predictions, and conclusions, serving layer can summarize, filter, and organize the information that comes from analytics module. In some embodiments, serving layer can request raw data from a sensor, collector, or analytics module.

Web frontend can connect with serving layer to present the data from serving layer in a page for human presentation. For example, web frontend can present the data in bar charts, core charts, tree maps, acyclic dependency maps, line graphs, tables, etc. Web frontend can be configured to allow a user to “drill down” on information sets to get a filtered data representation specific to the item the user wishes to “drill down” to. For example, individual traffic flows, components, etc. Web frontend can also be configured to allow a user to filter by search. This search filter can use natural language processing to determine analyze the network administrator's input. There can be options to view data relative to the current second, minute, hour, day, etc. Web frontend can allow a network administrator to view traffic flows, application dependency maps, network topology, etc.

In some embodiments, web frontend is solely configured to present information. In some embodiments, web frontend can receive inputs from a network administrator to configure traffic monitoring system or components of the datacenter. These instructions can be passed through serving layer, sent to configuration and image manager, or sent to policy engine. Authentication module can verify the identity and privileges of the network administrator. In some embodiments, authentication module can grant network administrators different rights according to established policies. Public alert module can identify network conditions that satisfy specified criteria and push alerts to third party tools. Public alert module can use reports generated or accessible through analytics engine. One example of third party tools is a security information and event management system. Third party tools may retrieve information from serving layer through an API.

FIG. 2A illustrates a schematic diagram of an example capturing agent deployment 200 in a virtualized environment. Server 106_A can execute and host one or more VMs 202_{A_C} (collectively “202”). VMs 202_{A_C} can be similar to VMs 110_{A_E} of FIG. 1. For example, VM 1 (202_A) of FIG. 2A can be VM 1 (110_A) of FIG. 1, and so forth. VMs 202 can be configured to run workloads (e.g., applications, services, processes, functions, etc.) based on hardware resources 212 on server 106_A. VMs 202 can run on guest operating systems 206_A-C (collectively “206”) on a virtual operating platform provided by hypervisor 208. Each VM 202 can run a respective guest operating system 206 which can be the same or different as other guest operating systems 206 associated with other VMs 202 on server 106_A. Each of guest operating systems 206 can execute one or more processes, which may in turn be programs, applications, modules, drivers, services, widgets, etc. Each of guest operating systems 206 may also be associated with one or more user accounts. For example, many popular operating systems such as LINUX, UNIX, WINDOWS, MAC OS, etc., offer multi-user environments where one or more users can use the system concurrently and share software/hardware resources. One or more users can sign in or log in to their user accounts associated with the operating system and run various workloads. Moreover, each VM 202 can have one or more network addresses, such as an internet protocol (IP) address. VMs 202 can thus communicate with hypervisor 208, server 106_A, and/or any remote devices or networks using the one or more network addresses.

Hypervisor 208 (otherwise known as a virtual machine monitor) can be a layer of software, firmware, and/or hardware that creates and runs VMs 202. Guest operating systems 206 running on VMs 202 can share virtualized hardware resources created by hypervisor 208. The virtualized hardware resources can provide the illusion of separate hardware components. Moreover, the virtualized hardware resources can perform as physical hardware components (e.g., memory, storage, processor, network interface, etc.), and can be driven by hardware resources 212 on server 106_A. Hypervisor 208 can have one or more network addresses, such as an internet protocol (IP) address, to communicate with other devices, components, or networks. For example, hypervisor 208 can have a dedicated IP address which it can use to communicate with VMs 202, server 106_A, and/or any remote devices or networks.

Hardware resources 212 of server 106_A can provide the underlying physical hardware that drive operations and functionalities provided by server 106_A, hypervisor 208, and VMs 202. Hardware resources 212 can include, for example, one or more memory resources, one or more storage resources, one or more communication interfaces, one or more processors, one or more circuit boards, one or more buses, one or more extension cards, one or more power supplies, one or more antennas, one or more peripheral components, etc. Additional examples of hardware resources are described below with reference to FIGS. 10 and 11A-B.

Server 106_A can also include one or more host operating systems (not shown). The number of host operating system can vary by configuration. For example, some configurations can include a dual boot configuration that allows server 106_A to boot into one of multiple host operating systems. In other configurations, server 106_A may run a single host operating system. Host operating systems can run on hardware resources 212. In some cases, hypervisor 208 can run on, or utilize, a host operating system on server 106_A. Each of the host operating systems can execute one or more processes, which may be programs, applications, modules, drivers, services, widgets, etc. Each of the host operating systems may also be associated with one or more OS user accounts.

Server 106_A can also have one or more network addresses, such as an internet protocol (IP) address, to communicate with other devices, components, or networks. For example, server 106_A can have an IP address assigned to a communications interface from hardware resources 212, which it can use to communicate with VNIs 202, hypervisor 208, leaf router 104_A in FIG. 1, collectors 118 in FIG. 1, and/or any remote devices or networks.

VM capturing agents 204_{A_C} (collectively “204”) can be deployed on one or more of VMs 202. V1\4 capturing agents 204 can be data and packet inspection agents or sensors deployed on VMs 202 to capture packets, flows, processes, events, traffic, and/or any data flowing into, out of, or through VMs 202. VM capturing agents 204 can be configured to export or report any data collected or captured by the capturing agents 204 to a remote entity, such as collectors 118, for example. VM capturing agents 204 can communicate or report such data using a network address of the respective VMs 202 (e.g., VM IP address).

VM capturing agents 204 can capture and report any traffic (e.g., packets, flows, etc.) sent, received, generated, and/or processed by VMs 202. For example, capturing agents 204 can report every packet or flow of communication sent and received by VMs 202. Such communication channel between capturing agents 204 and collectors 108 creates a flow in every monitoring period or interval and the flow generated by capturing agents 204 may be denoted as a control flow. Moreover, any communication sent or received by VMs 202, including data reported from capturing agents 204, can create a network flow. VM capturing agents 204 can report such flows in the form of a control flow to a remote device, such as collectors 118 illustrated in FIG. 1. VM capturing agents 204 can report each flow separately or aggregated with other flows. When reporting a flow via a control flow, VM capturing agents 204 can include a capturing agent identifier that identifies capturing agents 204 as reporting the associated flow. VM capturing agents 204 can also include in the control flow a flow identifier, an IP address, a timestamp, metadata, a process ID, an OS username associated with the process ID, and any other information, as further described below. In addition, capturing agents 204 can append the process and user information (i.e., which process and/or user is associated with a particular flow) to the control flow. The additional information as identified above can be applied to the control flow as labels. Alternatively, the additional information can be included as part of a header, a trailer, or a payload.

VM capturing agents 204 can also report multiple flows as a set of flows. When reporting a set of flows, VM capturing agents 204 can include a flow identifier for the set of flows and/or a flow identifier for each flow in the set of flows. VM capturing agents 204 can also include one or more timestamps and other information as previously explained.

VM capturing agents 204 can run as a process, kernel module, or kernel driver on guest operating systems 206 of VMs 202. VM capturing agents 204 can thus monitor any traffic sent, received, or processed by VMs 202, any processes running on guest operating systems 206, any users and user activities on guest operating system 206, any workloads on VMs 202, etc.

Hypervisor capturing agent 210 can be deployed on hypervisor 208. Hypervisor capturing agent 210 can be a data inspection agent or a sensor deployed on hypervisor 208 to capture traffic (e.g., packets, flows, etc.) and/or data flowing through hypervisor 208. Hypervisor capturing agent 210 can be configured to export or report any data collected or captured by hypervisor capturing agent 210 to a remote entity, such as collectors 118, for example. Hypervisor capturing agent 210 can communicate or report such data using a network address of hypervisor 208, such as an IP address of hypervisor 208.

Because hypervisor 208 can see traffic and data originating from VMs 202, hypervisor capturing agent 210 can also capture and report any data (e.g., traffic data) associated with VMs 202. For example, hypervisor capturing agent 210 can report every packet or flow of communication sent or received by VMs 202 and/or VM capturing agents 204. Moreover, any communication sent or received by hypervisor 208, including data reported from hypervisor capturing agent 210, can create a network flow. Hypervisor capturing agent 210 can report such flows in the form of a control flow to a remote device, such as collectors 118 illustrated in FIG. 1. Hypervisor capturing agent 210 can report each flow separately and/or in combination with other flows or data. When reporting a flow, hypervisor capturing agent 210 can include a capturing agent identifier that identifies hypervisor capturing agent 210 as reporting the flow. Hypervisor capturing agent 210 can also include in the control flow a flow identifier, an IP address, a timestamp, metadata, a process ID, and any other information, as explained below. In addition, capturing agents 210 can append the process and user information (i.e., which process and/or user is associated with a particular flow) to the control flow. The additional information as identified above can be applied to the control flow as labels. Alternatively, the additional information can be included as part of a header, a trailer, or a payload.

Hypervisor capturing agent 210 can also report multiple flows as a set of flows. When reporting a set of flows, hypervisor capturing agent 210 can include a flow identifier for the set of flows and/or a flow identifier for each flow in the set of flows. Hypervisor capturing agent 210 can also include one or more timestamps and other information as previously explained, such as process and user information.

As previously explained, any communication captured or reported by VM capturing agents 204 can flow through hypervisor 208. Thus, hypervisor capturing agent 210 can observe and capture any flows or packets reported by VM capturing agents 204, including any control flows. Accordingly, hypervisor capturing agent 210 can also report any packets or flows reported by VM capturing agents 204 and any control flows generated by VM capturing agents 204. For example, VM capturing agent 204_A on VM 1 (202_A) captures flow 1 (“F1”) and reports F1 to collector 118 on FIG. 1. Hypervisor capturing agent 210 on hypervisor 208 can also see and capture F1, as F1 would traverse hypervisor 208 when being sent or received by VM 1 (202_A). Accordingly, hypervisor capturing agent 210 on hypervisor 208 can also report F1 to collector 118. Thus, collector 118 can receive a report of F1 from VM capturing agent 204_A on VM 1 (202_A) and another report of F1 from hypervisor capturing agent 210 on hypervisor 208.

When reporting F1, hypervisor capturing agent 210 can report F1 as a message or report that is separate from the message or report of F1 transmitted by VM capturing agent 204A on VNI 1 (202_A). However, hypervisor capturing agent 210 can also, or otherwise, report F1 as a message or report that includes or appends the message or report of F1 transmitted by VM capturing agent 204_A on VM 1 (202_A). In other words, hypervisor capturing agent 210 can report F1 as a separate message or report from VM capturing agent 204_A's message or report of F1, and/or a same message or report that includes both a report of F1 by hypervisor capturing agent 210 and the report of F1 by VM capturing agent 204_A at VM 1 (202_A). In this way, VM capturing agents 204 at VMs 202 can report packets or flows received or sent by VMs 202, and hypervisor capturing agent 210 at hypervisor 208 can report packets or flows received or sent by hypervisor 208, including any flows or packets received or sent by VMs 202 and/or reported by VM capturing agents 204.

Hypervisor capturing agent 210 can run as a process, kernel module, or kernel driver on the host operating system associated with hypervisor 208. Hypervisor capturing agent 210 can thus monitor any traffic sent and received by hypervisor 208, any processes associated with hypervisor 208, etc.

Server 106_A can also have server capturing agent 214 running on it. Server capturing agent 214 can be a data inspection agent or sensor deployed on server 106_A to capture data (e.g., packets, flows, traffic data, etc.) on server 106_A. Server capturing agent 214 can be configured to export or report any data collected or captured by server capturing agent 214 to a remote entity, such as collector 118, for example. Server capturing agent 214 can communicate or report such data using a network address of server 106_A, such as an IP address of server 106_A.

Server capturing agent 214 can capture and report any packet or flow of communication associated with server 106_A. For example, capturing agent 216 can report every packet or flow of communication sent or received by one or more communication interfaces of server 106_A. Moreover, any communication sent or received by server 106_A, including data reported from capturing agents 204 and 210, can create a network flow associated with server 106_A. Server capturing agent 214 can report such flows in the form of a control flow to a remote device, such as collector 118 illustrated in FIG. 1. Server capturing agent 214 can report each flow separately or in combination. When reporting a flow, server capturing agent 214 can include a capturing agent identifier that identifies server capturing agent 214 as reporting the associated flow. Server capturing agent 214 can also include in the control flow a flow identifier, an IP address, a timestamp, metadata, a process ID, and any other information. In addition, capturing agent 214 can append the process and user information (i.e., which process and/or user is associated with a particular flow) to the control flow. The additional information as identified above can be applied to the control flow as labels. Alternatively, the additional information can be included as part of a header, a trailer, or a payload.

Server capturing agent 214 can also report multiple flows as a set of flows. When reporting a set of flows, server capturing agent 214 can include a flow identifier for the set of flows and/or a flow identifier for each flow in the set of flows. Server capturing agent 214 can also include one or more timestamps and other information as previously explained.

Any communications captured or reported by capturing agents 204 and 210 can flow through server 106_A. Thus, server capturing agent 214 can observe or capture any flows or packets reported by capturing agents 204 and 210. In other words, network data observed by capturing agents 204 and 210 inside VMs 202 and hypervisor 208 can be a subset of the data observed by server capturing agent 214 on server 106_A. Accordingly, server capturing agent 214 can report any packets or flows reported by capturing agents 204 and 210 and any control flows generated by capturing agents 204 and 210. For example, capturing agent 204_A on VM 1 (202_A) captures flow 1 (F1) and reports F1 to collector 118 as illustrated on FIG. 1. Capturing agent 210 on hypervisor 208 can also observe and capture F1, as F1 would traverse hypervisor 208 when being sent or received by VM 1 (202_A). In addition, capturing agent 214 on server 106_A can also see and capture F1, as F1 would traverse server 106_A when being sent or received by VM 1 (202_A) and hypervisor 208. Accordingly, capturing agent 214 can also report F1 to collector 118. Thus, collector 118 can receive a report (i.e., control flow) regarding F1 from capturing agent 204_A on VM 1 (202_A), capturing agent 210 on hypervisor 208, and capturing agent 214 on server 106_A.

When reporting F1, server capturing agent 214 can report F1 as a message or report that is separate from any messages or reports of F 1 transmitted by capturing agent 204_A on VM 1 (202_A) or capturing agent 210 on hypervisor 208. However, server capturing agent 214 can also, or otherwise, report F1 as a message or report that includes or appends the messages or reports or metadata of F1 transmitted by capturing agent 204_A on VM 1 (202_A) and capturing agent 210 on hypervisor 208. In other words, server capturing agent 214 can report F1 as a separate message or report from the messages or reports of F1 from capturing agent 204_A and capturing agent 210, and/or a same message or report that includes a report of F1 by capturing agent 204_A, capturing agent 210, and capturing agent 214. In this way, capturing agents 204 at VMs 202 can report packets or flows received or sent by VMs 202, capturing agent 210 at hypervisor 208 can report packets or flows received or sent by hypervisor 208, including any flows or packets received or sent by VMs 202 and reported by capturing agents 204, and capturing agent 214 at server 106_A can report packets or flows received or sent by server 106_A, including any flows or packets received or sent by VMs 202 and reported by capturing agents 204, and any flows or packets received or sent by hypervisor 208 and reported by capturing agent 210.

Server capturing agent 214 can run as a process, kernel module, or kernel driver on the host operating system or a hardware component of server 106_A. Server capturing agent 214 can thus monitor any traffic sent and received by server 106_A, any processes associated with server 106_A, etc.

In addition to network data, capturing agents 204, 210, and 214 can capture additional information about the system or environment in which they reside. For example, capturing agents 204, 210, and 214 can capture data or metadata of active or previously active processes of their respective system or environment, operating system user identifiers, metadata of files on their respective system or environment, timestamps, network addressing information, flow identifiers, capturing agent identifiers, etc. Moreover, capturing agents 204, 210, 214 are not specific to any operating system environment, hypervisor environment, network environment, or hardware environment. Thus, capturing agents 204, 210, and 214 can operate in any environment.

As previously explained, capturing agents 204, 210, and 214 can send information about the network traffic they observe. This information can be sent to one or more remote devices, such as one or more servers, collectors, engines, etc. Each capturing agent can be configured to send respective information using a network address, such as an IP address, and any other communication details, such as port number, to one or more destination addresses or locations. Capturing agents 204, 210, and 214 can send metadata about one or more flows, packets, communications, processes, events, etc.

Capturing agents 204, 210, and 214 can periodically report information about each flow or packet they observe. The information reported can contain a list of flows or packets that were active during a period of time (e.g., between the current time and the time at which the last information was reported). The communication channel between the capturing agent and the destination can create a flow in every interval. For example, the communication channel between capturing agent 214 and collector 118 can create a control flow. Thus, the information reported by a capturing agent can also contain information about this control flow. For example, the information reported by capturing agent 214 to collector 118 can include a list of flows or packets that were active at hypervisor 208 during a period of time, as well as information about the communication channel between capturing agent 210 and collector 118 used to report the information by capturing agent 210.

FIG. 2B illustrates a schematic diagram of example capturing agent deployment 220 in an example network device. The network device is described as leaf router 104A, as illustrated in FIG. 1. However, this is for explanation purposes. The network device can be any other network device, such as any other switch, router, etc.

In this example, leaf router 104_A can include network resources 222, such as memory, storage, communication, processing, input, output, and other types of resources. Leaf router 104_A can also include operating system environment 224. The operating system environment 224 can include any operating system, such as a network operating system, embedded operating system, etc. Operating system environment 224 can include processes, functions, and applications for performing networking, routing, switching, forwarding, policy implementation, messaging, monitoring, and other types of operations.

Leaf router 104_A can also include capturing agent 226. Capturing agent 226 can be an agent or sensor configured to capture network data, such as flows or packets, sent received, or processed by leaf router 104_A. Capturing agent 226 can also be configured to capture other information, such as processes, statistics, users, alerts, status information, device information, etc. Moreover, capturing agent 226 can be configured to report captured data to a remote device or network, such as collector 118 shown in FIG. 1, for example. Capturing agent 226 can report information using one or more network addresses associated with leaf router 104_A or collector 118. For example, capturing agent 226 can be configured to report information using an IP assigned to an active communications interface on leaf router 104A.

Leaf router 104_A can be configured to route traffic to and from other devices or networks, such as server 106_A. Accordingly, capturing agent 226 can also report data reported by other capturing agents on other devices. For example, leaf router 104_A can be configured to route traffic sent and received by server 106_A to other devices. Thus, data reported from capturing agents deployed on server 106_A, such as VM and hypervisor capturing agents on server 106_A, would also be observed by capturing agent 226 and can thus be reported by capturing agent 226 as data observed at leaf router 104_A. Such report can be a control flow generated by capturing agent 226. Data reported by the VM and hypervisor capturing agents on server 106_A can therefore be a subset of the data reported by capturing agent 226.

Capturing agent 226 can run as a process or component (e.g., firmware, module, hardware device, etc.) in leaf router 104_A. Moreover, capturing agent 226 can be installed on leaf router 104_A as a software or firmware agent. In some configurations, leaf router 104_A itself can act as capturing agent 226. Moreover, capturing agent 226 can run within operating system 224 and/or separate from operating system 224.

FIG. 2C illustrates a schematic diagram of example reporting system 240 in an example capturing agent topology. Leaf router 104_A can route packets or traffic 242 between fabric 112 and server 106_A, hypervisor 108_A, and VM 110_A. Packets or traffic 242 between VM 110_A and leaf router 104_A can flow through hypervisor 108_A and server 106_A. Packets or traffic 242 between hypervisor 108_A and leaf router 104_A can flow through server 106_A. Finally, packets or traffic 242 between server 106_A and leaf router 104_A can flow directly to leaf router 104_A. However, in some cases, packets or traffic 242 between server 106_A and leaf router 104_A can flow through one or more intervening devices or networks, such as a switch or a firewall.

Moreover, VM capturing agent 204_A at VM 110_A, hypervisor capturing agent 210 at hypervisor 108_A, network device capturing agent 226 at leaf router 104_A, and any server capturing agent at server 106_A (e.g., capturing agent running on host environment of server 106_A) can send reports 244 (also referred to as control flows) to collector 118 based on the packets or traffic 242 captured at each respective capturing agent. Reports 244 from VM capturing agent 204_A to collector 118 can flow through VM¹¹⁰A, hypervisor 108_A, server 106_A, and leaf router 104_A. Reports 244 from hypervisor capturing agent 210 to collector 118 can flow through hypervisor 108_A, server 106_A, and leaf router 104_A. Reports 244 from any other server capturing agent at server 106_A to collector 118 can flow through server 106_A and leaf router 104_A. Finally, reports 244 from network device capturing agent 226 to collector 118 can flow through leaf router 104_A. Although reports 244 are depicted as being routed separately from traffic 242 in FIG. 2C, one of ordinary skill in the art will understand that reports 244 and traffic 242 can be transmitted through the same communication channel(s).

Reports 244 can include any portion of packets or traffic 242 captured at the respective capturing agents. Reports 244 can also include other information, such as timestamps, process information, capturing agent identifiers, flow identifiers, flow statistics, notifications, logs, user information, system information, etc. Some or all of this information can be appended to reports 244 as one or more labels, metadata, or as part of the packet(s)′ header, trailer, or payload. For example, if a user opens a browser on VM 110_A and navigates to examplewebsite.com, VNI capturing agent 204_A of VM 110_A can determine which user (i.e., operating system user) of VM 110_A (e.g., username “johndoe85⁻) and which process being executed on the operating system of VM 110_A (e.g., “chrome.exe”) were responsible for the particular network flow to and from examplewebsite.com. Once such information is determined, the information can be included in report 244 as labels for example, and report 244 can be transmitted from VM capturing agent 204_A to collector 118. Such additional information can help system 240 to gain insight into flow information at the process and user level, for instance. This information can be used for security, optimization, and determining structures and dependencies within system 240. Moreover, reports 244 can be transmitted to collector 118 periodically as new packets or traffic 242 are captured by a capturing agent. Further, each capturing agent can send a single report or multiple reports to collector 118. For example, each of the capturing agents 116 can be configured to send a report to collector 118 for every flow, packet, message, communication, or network data received, transmitted, and/or generated by its respective host (e.g., VM 110_A, hypervisor 108_A, server 106_A, and leaf router 104_A). As such, collector 118 can receive a report of a same packet from multiple capturing agents.

For example, a packet received by VM 110_A from fabric 112 can be captured and reported by VM capturing agent 204_A. Since the packet received by VM 110_A will also flow through leaf router 104_A and hypervisor 108_A, it can also be captured and reported by hypervisor capturing agent 210 and network device capturing agent 226. Thus, for a packet received by VM 110_A from fabric 112, collector 118 can receive a report of the packet from VM capturing agent 204_A, hypervisor capturing agent 210, and network device capturing agent 226.

Similarly, a packet sent by VM 110_A to fabric 112 can be captured and reported by VM capturing agent 204_A. Since the packet sent by VM 110_A will also flow through leaf router 104_A and hypervisor 108_A, it can also be captured and reported by hypervisor capturing agent 210 and network device capturing agent 226. Thus, for a packet sent by VM 110_A to fabric 112, collector 118 can receive a report of the packet from VM capturing agent 204_A, hypervisor capturing agent 210, and network device capturing agent 226.

On the other hand, a packet originating at, or destined to, hypervisor 108_A, can be captured and reported by hypervisor capturing agent 210 and network device capturing agent 226, but not VM capturing agent 204_A, as such packet may not flow through VM 110_A. Moreover, a packet originating at, or destined to, leaf router 104_A, will be captured and reported by network device capturing agent 226, but not VM capturing agent 204_A, hypervisor capturing agent 210, or any other capturing agent on server 106_A, as such packet may not flow through VM 110_A, hypervisor 108_A, or server 106A.

Each of the capturing agents 204_A, 210, 226 can include a respective unique capturing agent identifier on each of reports 244 it sends to collector 118, to allow collector 118 to determine which capturing agent sent the report. Reports 244 can be used to analyze network and/or system data and conditions for troubleshooting, security, visualization, configuration, planning, and management. Capturing agent identifiers in reports 244 can also be used to determine which capturing agents reported what flows. This information can then be used to determine capturing agent placement and topology, as further described below, as well as mapping individual flows to processes and users. Such additional insights gained can be useful for analyzing the data in reports 244, as well as troubleshooting, security, visualization, configuration, planning, and management.

FIGS. 3A-F illustrate schematic diagrams of example configurations for reporting flows captured by capturing agents in an example capturing agent topology. Referring to FIG. 3A, leaf router 104_A can receive flow 302 from fabric 112. In this example, flow 302 is destined to VM 110_A. Leaf router 104_A can thus forward flow 302 received from fabric 112 to server 106_A and hypervisor 108_A. Network device capturing agent 226 at leaf router 104_A can thus capture flow 302, and send a new control flow 304, reporting the received flow 302, to collector 118. Network device capturing agent 226 may include in control flow 304 any additional information such as process information and user information related to leaf router 104_A and flow 302.

Server 106_A and hypervisor 108_A can receive flow 302 from leaf router 104_A. Hypervisor 108_A can then forward the received flow 302 to VM 110_A. Hypervisor capturing agent 210 can also capture the received flow 302 and send a new control flow 306, reporting the received flow 302, to collector 118. Hypervisor capturing agent 210 may include in control flow 306 any additional information such as process information and user information related to hypervisor 108_A and flow 302. Leaf router 104_A can receive control flow 306, reporting flow 302, originating from hypervisor capturing agent 210, and forward flow 306 to collector 118. Network device capturing agent 226 can also capture control flow 306 received from hypervisor capturing agent 210, and send a new control flow 308, reporting flow 306, to collector 118. Again, network device capturing agent 226 may include in control flow 308 any additional information such as process information and user information related to network device 104_A and flow 306.

Moreover, VM 110_A can receive flow 302 from hypervisor 108_A. At this point, flow 302 has reached its intended destination: VM 110_A. Accordingly, VIVI 110_A can then process flow 302. Once flow 302 is received by VM 110_A, VM capturing agent 204_A can capture received flow 302 and send a new control flow 310, reporting the receipt of flow 302, to collector 118. VM capturing agent 204_A can include in control flow 310 any additional information such as process information and user information related to VM 110_A and flow 302.

Hypervisor 108_A can receive control flow 310 from VM capturing agent 204_A, and forward it to leaf router 104_A. Hypervisor capturing agent 210 can also capture flow 310, received from VM capturing agent 204_A and reporting the receipt of flow 302, and send a new control flow 312, reporting flow 310, to collector 118. Hypervisor capturing agent 210 may include in control flow 312 any additional information such as process information and user information related to hypervisor 108_A and flow 310.

Leaf router 104_A can receive flow 310 forwarded from hypervisor 108_A, and forward it to collector 118. Network device capturing agent 226 can also capture flow 310, forwarded from hypervisor capturing agent 210 and reporting the receipt of flow 302 at VM 110_A, and send a new control flow 314, reporting flow 310, to collector 118. Network device capturing agent 226 may include in control flow 314 any additional information such as process information and user information related to network device 104_A and flow 310.

Leaf router 104_A can receive packet 312 from hypervisor capturing agent 210 and forward it to collector 118. Network device capturing agent 226 can also capture flow 312 and send a new control flow 316, reporting flow 312, to collector 118. Network device capturing agent 226 may include in control flow 316 any additional information such as process information and user information related to network device 104_A and flow 312.

As described above, in this example, flow 302 destined from fabric 112 to VM 110_A, can be reported by network device capturing agent 226, hypervisor capturing agent 210, and VM capturing agent 204_A to collector 118. In addition, hypervisor capturing agent 210 and network device capturing agent 226 can each report the communication from VM 110_A to collector 118, reporting flow 302 to collector 118. Moreover, network device capturing agent 226 can report any communications from hypervisor capturing agent 210 reporting flows or communications captured by hypervisor capturing agent 210. As one of skill in the art will understand, the order in which control flows 304, 306, 308, 310, 312, 314, 316 are reported to collector 118 need not occur in the same order that is presented in this disclosure as long as each control flow is transmitted or forwarded to another device after the flow which the control flow is reporting is received. For example, control flow 314, which reports flow 310, may be transmitted to collector 118 either before or after each of control flows 308, 312, 316 is transmitted or forwarded to collector 118 as long as control flow 314 is transmitted sometime after flow 310 is received at leaf router 104_A. This applies to other control flows illustrated throughout disclosure especially those shown in FIGS. 3A-3F. In addition, other capturing agents such as a server capturing agent (not shown) for 106_A may also capture and report any traffic or flows that server 106_A may send, receive, or otherwise process.

Referring to FIG. 3B, leaf router 104_A can receive flow 324 from fabric 112. In this example, flow 324 is destined for hypervisor 108_A. Leaf router 104_A can thus forward the flow 324 received from fabric 112 to server 106_A and hypervisor 108_A. network device capturing agent 226 at leaf router 104_A can also capture the flow 324, and send a new control flow 318, reporting the received flow 324, to collector 118. Network device capturing agent 226 may include in control flow 318 any additional information such as process information and user information related to network device 104_A and flow 324.

Server 106_A and hypervisor 108_A can receive flow 324 from leaf router 104_A. Hypervisor 108_A can process received flow 324. Hypervisor capturing agent 210 can also capture received flow 324 and send a new control flow 320, reporting received flow 324, to collector 118. Hypervisor capturing agent 210 may include in control flow 320 any additional information such as process information and user information related to hypervisor 108_A and flow 324. Leaf router 104_A can receive flow 320, reporting flow 324, from hypervisor capturing agent 210, and forward control flow 320 to collector 118. Network device capturing agent 226 can also capture flow 320 received from hypervisor capturing agent 210, and send a new control flow 322, reporting flow 320, to collector 118. Network device capturing agent 226 may include in control flow 322 any additional information such as process information and user information related to network device 104_A and flow 320.

As described above, in this example, flow 324 destined from fabric 112 to hypervisor 108_A, can be reported by network device capturing agent 226 and hypervisor capturing agent 210 to collector 118. In addition, network device capturing agent 226 can report the communication from hypervisor 108_A to collector 118, reporting flow 324 to collector 118.

Referring to FIG. 3C, leaf router 104_A can receive flow 326 from fabric 112. In this example, flow 326 is destined for leaf router 104_A. Thus, leaf router 104_A can process flow 326, and network device capturing agent 226 can capture flow 326, and send a new control flow 328, reporting the received flow 326, to collector 118. Network device capturing agent 226 may include in control flow 328 any additional information such as process information and user information related to network device 104_A and flow 326.

Referring to FIG. 3D, VM 110_A can send flow 330 to fabric 112. Hypervisor 108_A can receive flow 330 and forward it to leaf router 104_A. Leaf router 104_A can receive flow 330 and forward it to fabric 112.

VM capturing agent 204_A can also capture flow 330 and send a new control flow 332, reporting flow 330, to collector 118. VM capturing agent 204_A may include in control flow 332 any additional information such as process information and user information related to VM 110_A and flow 330. Hypervisor capturing agent 210 can also capture flow 330 and send a new control flow 334, reporting flow 330, to collector 118. Hypervisor capturing agent 210 may include in control flow 334 any additional information such as process information and user information related to hypervisor 108_A and flow 330. Similarly, network device capturing agent 226 can capture flow 330, and send a new control flow 336, reporting flow 330, to collector 118. Network device capturing agent 226 may include in control flow 336 any additional information such as process information and user information related to network device 104_A and flow 330.

Hypervisor capturing agent 210 can also capture flow 332, reporting flow 330 by VM capturing agent 204_A, and send a new control flow 338, reporting flow 332, to collector 118. Hypervisor capturing agent 210 may include in control flow 338 any additional information such as process information and user information related to hypervisor 108_A and flow 332.

Network device capturing agent 226 can similarly capture flow 332, reporting flow 330 by VM capturing agent 204_A, and send a new control flow 340, reporting flow 332, to collector 118. Network device capturing agent 226 may include in control flow 340 any additional information such as process information and user information related to network device 104_A and flow 332. Moreover, network device capturing agent 226 can capture flow 338, reporting flow 332 from hypervisor capturing agent 210, and send a new control flow 342, reporting flow 338, to collector 118. Network device capturing agent 226 may include in control flow 342 any additional information such as process information and user information related to network device 104_A and flow 338.

As described above, in this example, flow 330 destined to fabric 112 from VM 110_A, can be reported by network device capturing agent 226, hypervisor capturing agent 210, and VM capturing agent 204_A to collector 118. In addition, hypervisor capturing agent 210 and network device capturing agent 226 can each report the communication (i.e., control flow) from VIVI 110_A to collector 118, reporting flow 330 to collector 118. Network device capturing agent 226 can also report any communications from hypervisor capturing agent 210 reporting flows or communications captured by hypervisor capturing agent 210.

Referring to FIG. 3E, hypervisor 108_A can send flow 344 to fabric 112. In this example, flow 344 is originated by hypervisor 108_A. Leaf router 104_A can receive flow 344 and forward it to fabric 112.

Hypervisor capturing agent 210 can also capture flow 344 and send a new control flow 346, reporting flow 344, to collector 118. Hypervisor capturing agent 210 may include in control flow 346 any additional information such as process information and user information related to hypervisor 108_A and flow 344. Similarly, network device capturing agent 226 can capture flow 344, and send a new control flow 348, reporting flow 344, to collector 118. Again, network device capturing agent 226 may include in control flow 348 any additional information such as process information and user information related to network device 104_A and flow 344.

Network device capturing agent 226 can also capture flow 346, reporting flow 344 by hypervisor capturing agent 210, and send a new control flow 350, reporting flow 346, to collector 118. Network device capturing agent 226 may include in control flow 350 any additional information such as process information and user information related to network device 104_A and flow 346.

Referring to FIG. 3F, leaf router 104_A can send flow 352 to fabric 112. In this example, flow 352 is originated by leaf router 104_A. Network device capturing agent 226 can capture flow 352, and send a new control flow 354, reporting flow 352, to collector 118. In addition, network device capturing agent 226 may include in control flow 354 any additional information such as process information and user information related to network device 104_A and flow 352. Thus, collector 118 can receive a report of flow 352 from network device capturing agent 226.

FIG. 4 illustrates a schematic diagram of an example configuration 400 for collecting capturing agent reports (i.e., control flows). In configuration 400, traffic between fabric 112 and VM 110_A is configured to flow through hypervisor 108_A. Moreover, traffic between fabric 112 and hypervisor 108_A is configured to flow through leaf router 104_A.

VM capturing agent 204_A can be configured to report to collector 118 traffic sent, received, or processed by VM 110_A. Hypervisor capturing agent 210 can be configured to report to collector 118 traffic sent, received, or processed by hypervisor 108_A. Finally, network device capturing agent 226 can be configured to report to collector 118 traffic sent, received, or processed by leaf router 104_A.

Collector 118 can thus receive flows 402 from VM capturing agent 204_A, flows 404 from hypervisor capturing agent 210, and flows 406 from network device capturing agent 226. Flows 402, 404, and 406 can include control flows. Flows 402 can include flows captured by VM capturing agent 204_A at VM 110_A.

Flows 404 can include flows captured by hypervisor capturing agent 210 at hypervisor 108_A. Flows captured by hypervisor capturing agent 210 can also include flows 402 captured by VM capturing agent 204_A, as traffic sent and received by VM 110_A will be received and observed by hypervisor 108_A and captured by hypervisor capturing agent 210.

Flows 406 can include flows captured by network device capturing agent 226 at leaf router 104_A. Flows captured by network device capturing agent 226 can also include flows 402 captured by VM capturing agent 204_A and flows 404 captured by hypervisor capturing agent 210, as traffic sent and received by VM 110_A and hypervisor 108_A is routed through leaf router 104_A and can thus be captured by network device capturing agent 226.

Collector 118 can collect flows 402, 404, and 406, and store the reported data. Collector 118 can also forward some or all of flows 402, 404, and 406, and/or any respective portion thereof, to engine 120. Engine 120 can process the information, including any process information and user information, received from collector 118 to identify patterns, conditions, statuses, network or device characteristics; log statistics or history details; aggregate and/or process the data; generate reports, timelines, alerts, graphical user interfaces; detect errors, events, inconsistencies; troubleshoot networks or devices; configure networks or devices; deploy services or devices; reconfigure services, applications, devices, or networks; etc. In particular, collector 118 or engine 120 can map individual flows that traverse VM 110_A, hypervisor 108_A, and/or leaf router 104_A to specific processes or users that are associated with VM 110_A, hypervisor 108_A, and/or leaf router 104_A. For example, collector 118 or engine 120 can determine that a particular flow that originated from VM 110_A and destined for fabric 112 was sent by an OS user named X on VM 110_A and via a process named Y on VM 110_A. It may be determined that the same flow was received by a process named Z on hypervisor 108_A and forwarded to a process named Won leaf router 104_A.

While engine 120 is illustrated as a separate entity, other configurations are also contemplated herein. For example, engine 120 can be part of collector 118 and/or a separate entity. Indeed, engine 120 can include one or more devices, applications, modules, databases, processing components, elements, etc. Moreover, collector 118 can represent one or more collectors. For example, in some configurations, collector 118 can include multiple collection systems or entities, which can reside in one or more networks.

FIG. 5 illustrates a sequence diagram of example capturing agent reporting process 500. In this example, flow 1 (502) has been observed (e.g., received, sent, generated, processed) by VIVI 11 O_A, hypervisor 108_A, and leaf router 104_A. Flow 2 (504) has been observed by hypervisor 108_A and leaf router 104_A. Flow 3 (506) has only been observed by leaf router 104_A.

Since flow 1 (502) has been observed by VM 11 O_A, hypervisor 108_A, and leaf router 104_A, it can be captured and reported to collector 118 by VM capturing agent 204_A at VM 110_A, hypervisor capturing agent 210 at hypervisor 108_A, and network device capturing agent 226 at leaf router 104_A. On the other hand, since flow 2 (504) has been observed by hypervisor 108_A and leaf router 104_A but not by VM 110_A, it can be captured and reported to collector 118 by hypervisor capturing agent 210 at hypervisor 108_A and network device capturing agent 226 at leaf router 104_A, but not by VM capturing agent 204_A at VM no_A Finally, since flow 3 (506) has only been observed by leaf router 104_A, it can be captured and reported to collector 118 only by capturing agent 226 at leaf router 104_A.

The reports or control flows received by collector 118 can include information identifying the reporting capturing agent. For example, when transmitting a report to collector 118, each capturing agent can include a unique capturing agent identifier, which the collector 118 and/or any other entity reviewing the reports can use to map a received report with the reporting capturing agent. Furthermore, the reports or control flows received by collector 118 can include information identifying the process or the user responsible for the flow being reported. Collector 118 can use such information to map the flows to corresponding processes or users.

Thus, based on the reports from capturing agents 204_A, 210, and 226, collector 118 and/or a separate entity (e.g., engine 120) can determine that flow 1 (502) was observed and reported by capturing agent 204_A at VM 110_A, capturing agent 210 at hypervisor 108_A, and capturing agent 226 at leaf router 104_A; flow 2 (504) was observed and reported by capturing agent 210 at hypervisor 108_A and capturing agent 226 at leaf router 104_A; and flow 3 (506) was only observed and reported by capturing agent 226 at leaf router 104_A. Based on this information, collector 118 and/or a separate entity, can determine the placement of capturing agents 204_A, 210, 226 within VM 110_A, hypervisor 108_A, and leaf router 104_A, as further described below. In other words, this information can allow a device, such as collector 118, to determine which of capturing agents 204_A, 210, 226 is located at VM 110_A, which is located at hypervisor 108_A, and which is located at leaf router 104_A. If any of VM 110_A, hypervisor 108_A, and leaf router 104_A is moved to a different location (e.g., VM 110_A moved to server 106c and hypervisor 108B), the new flows collected by collector 118 can be used to detect the new placement and topology of VM 110_A, hypervisor 108_A, and leaf router 104_A and/or their respective capturing agents. Furthermore, the process and/or user information included in the control flows received at collector 118 may also assist in determining how VM 110_A, hypervisor 108_A, and/or leaf router 104_A may move to a different location within the network. For example, by recognizing that a new device that just appeared in the network is sending out a flow that matches the process and/or user profiles of a previously known device, such as VM 110_A, collector 118 can determine that the new device is actually VM 110_A that just moved to a different location (e.g., from server 1 (106_A) to server 4 (106_D)) within the network topology.

FIG. 6 illustrates a table of example mapping 600 of flow reports to capturing agents. In this example, flow 602 was sent/received by VM 110_A, flow 604 was sent/received by hypervisor 108_A, and flow 606 was sent/received by leaf router 104_A. Accordingly, flow 602 was reported by VM capturing agent 204_A, hypervisor capturing agent 210, and network device capturing agent 226. Flow 604 was reported by hypervisor capturing agent 210 and network device capturing agent 226, but not by VM capturing agent 204_A. Finally flow 606 was reported by network device capturing agent 226, but not VM capturing agent 204_A or hypervisor capturing agent 210.

FIG. 7 illustrates listing 700 of example fields on a capturing agent report or control flow. The listing 700 can include one or more fields, such as:

Flow identifier (e.g., unique identifier associated with the flow).

Capturing agent identifier (e.g., data uniquely identifying reporting capturing agent).

Timestamp (e.g., time of event, report, etc.).

Interval (e.g., time between current report and previous report, interval between flows or packets, interval between events, etc.).

Duration (e.g., duration of event, duration of communication, duration of flow, duration of report, etc.).

Flow direction (e.g., egress flow, ingress flow, etc.).

Application identifier (e.g., identifier of application associated with flow, process, event, or data).

Port (e.g., source port, destination port, layer 4 port, etc.).

Destination address (e.g., interface address associated with destination, IP address, domain name, network address, hardware address, virtual address, physical address, etc.).

Source address (e.g., interface address associated with source, IP address, domain name, network address, hardware address, virtual address, physical address, etc.).

Interface (e.g., interface address, interface information, etc.).

Protocol (e.g., layer 4 protocol, layer 3 protocol, etc.).

Event (e.g., description of event, event identifier, etc.).

Flag (e.g., layer 3 flag, flag options, etc.).

Tag (e.g., virtual local area network tag, etc.).

Process (e.g., process identifier, etc.).

User (e.g., OS username, etc.).

Bytes (e.g., flow size, packet size, transmission size, etc.).

The listing 700 includes a non-limiting example of fields in a report. Other fields and data items are also contemplated herein, such as handshake information, system information, network address associated with capturing agent or host, operating system environment information, network data or statistics, process statistics, system statistics, etc. The order in which these fields are illustrated is also exemplary and can be rearranged in any other way. One or more of these fields can be part of a header, a trailer, or a payload of in one or more packets. Moreover, one or more of these fields can be applied to the one or more packets as labels. Each of the fields can include data, metadata, and/or any other information relevant to the fields.

Having disclosed some basic system components and concepts, the disclosure now turns to the exemplary method embodiments shown in FIGS. 8-9. For the sake of clarity, the methods are described in terms of capturing agent 116, as shown in FIG. 1, configured to practice the method. However, the example methods can be practiced by any software or hardware components, devices, etc. heretofore disclosed. The steps outlined herein are exemplary and can be implemented in any combination thereof in any order, including combinations that exclude, add, or modify certain steps.

In FIG. 8, capturing agent 116, executing on a first device in a network, can monitor a network flow associated with the first device (802). The first device can be a VM, a hypervisor, a server, a network device, etc. Capturing agent 116 can be a process, a cluster of processes, a kernel module, or a kernel driver. In addition, capturing agent 116 can run on a guest operating system installed in a virtual machine on the device. Capturing agent 116 may also run on a host operating system installed at a hypervisor layer or on a hypervisor. Moreover, capturing agent 116 can be a process or a component in a network device such as a switch. The network flow or stream can be one or more data packets.

At step 804, capturing agent 116 can generate a control flow based on the network flow. The control flow can include metadata describing the network flow. The metadata can relate to network data, an active process of the system, a previously active process of the device, and/or a file that is present on the device. The metadata can also relate to operating system user identifiers, timestamps, network addressing information, flow identifiers, capturing agent identifiers, time interval, interval duration, flow direction, application identifier, port, destination address, source address, interface, protocol, event, flag, tag, user, size, handshake information, statistics, etc. with regards to the network flow being monitored and reported.

At step 806, capturing agent 116 can determine which process executing on the first device is associated with the network flow to yield process information. The process information may include the process identifier of the process. Furthermore, the process information may include information about the OS username associated with the process. The identified process may be responsible for sending, receiving, or otherwise processing the network flow. The process can belong to the operating system environment of the first device. Capturing agent 116 can further determine which OS user of the first device is associated with the network flow to yield user information.

The capturing agent 116 can determine which kernel module has been loaded and/or query the operating system to determine which process is executing on the first device. The capturing agent 116 can also determine process ownership information to identify which user has executed a particular service or process.

At step 808, capturing agent 116 can label the control flow with the process information to yield a labeled control flow. Capturing agent 116 can further label the control flow with user information. The process and/or user information can be applied or added to the control flow as part of a header, a trailer, or a payload.

At step 810, capturing agent can transmit the labeled control flow to a second device in the network. The second device can be a collector that is configured to receive a plurality of control flows from a plurality of devices, particularly from their capturing agents, and analyze the plurality of control flows to determine relationships between network flows and corresponding processes. Those other devices can also be VMs, hypervisors, servers, network devices, etc. equipped with VM capturing agents, hypervisor capturing agents, server capturing agents, network device capturing agents, etc. The second device can map the relationships between the network flows and the corresponding processes within the first device and other devices in the plurality of devices. The second device or another device can utilize this information to identify patterns, conditions, statuses, network or device characteristics; log statistics or history details; aggregate and/or process the data; generate reports, timelines, alerts, graphical user interfaces; detect errors, events, inconsistencies; troubleshoot networks or devices; configure networks or devices; deploy services or devices; reconfigure services, applications, devices, or networks; etc.

In FIG. 9, capturing agent 116, executing on a first device in a network, can monitor a network flow associated with the first device (902). The first device can be a VM, a hypervisor, a server, a network device, etc. Capturing agent 116 can be a process, a cluster of processes, a kernel module, or a kernel driver. In addition, capturing agent 116 can run on a guest operating system installed in a virtual machine on the device. Capturing agent 116 may also run on a host operating system installed at a hypervisor layer or on a hypervisor. Moreover, capturing agent 116 can be a process or a component in a network device such as a switch. The network flow or stream can be one or more data packets.

At step 904, capturing agent 116 can generate a control flow based on the network flow. The control flow can include metadata describing the network flow. The metadata can relate to network data, an active process of the system, a previously active process of the device, and/or a file that is present on the device. The metadata can also relate to processes, timestamps, network addressing information, flow identifiers, capturing agent identifiers, time interval, interval duration, flow direction, application identifier, port, destination address, source address, interface, protocol, event, flag, tag, size, handshake information, statistics, etc. with regards to the network flow being monitored and reported.

At step 906, capturing agent 116 can determine which user of the first device is associated with the network flow to yield user information. The user can be an operating system user account. The user information may include the username or the user identifier associated with the user. The user may be an OS user of the first device's OS environment. The user may be associated with a process that sends, receives, or otherwise processes the network flow. Capturing agent 116 can further determine which process executing on the first device is associated with the network flow to yield process information.

At step 908, capturing agent 116 can label the control flow with the user information to yield a labeled control flow. Capturing agent 116 can further label the control flow with process information. The process and/or user information can be applied or added to the control flow as part of a header, a trailer, or a payload.

At step 910, capturing agent can transmit the labeled control flow to a second device in the network. The second device can be a collector that is configured to receive a plurality of control flows from a plurality of devices, particularly from their capturing agents, and analyze the plurality of control flows to determine relationships between network flows and corresponding processes. Those other devices can also be VMs, hypervisors, servers, network devices, etc. equipped with VM capturing agents, hypervisor capturing agents, server capturing agents, network device capturing agents, etc. The second device can map the relationships between the network flows and the corresponding users associated with the first device or another device in the plurality of devices. The second device or some other device can utilize this information to identify patterns, conditions, statuses, network or device characteristics; log statistics or history details; aggregate and/or process the data; generate reports, timelines, alerts, graphical user interfaces; detect errors, events, inconsistencies; troubleshoot networks or devices; configure networks or devices; deploy services or devices; reconfigure services, applications, devices, or networks; etc.

Annotation

A flow is a collection of packets having a same source address, destination address, source port, destination port, protocol, tenant id, and starting timestamp. But having only this key/signature may not be particularly helpful to users trying to understand this data and we would like to be able tag flows to enable users to search the flow data and to present the flow data more meaningfully to users.

A high-level overview of the pipeline with the key components for flow annotation is provided as the attached figure. Generally, flow data is collected by sensors incorporated at various levels of a data center (e.g., virtual machine, hypervisor, physical switch, etc.) and provided to a Collector. The Collector may perform certain processing on the raw flow data, such as de-duping, and then that data is stored in the RDFS. The Compute Engine processes the flow data in the HDFS, including annotating each flow with certain metadata based on specified rules in order to classify each flow. This enables the UI to present meaningful views of flows or allows users to search flows based on tags.

Each flow is annotated according to certain default tags, such as Attack, Policy, Geo, Bogon, Whitelist, etc. Attack refers to whether a flow has been determined to be a malicious flow. Policy refers to whether a flow is compliant or non-compliant with policy. Geo refers to the geographic location from which the flow originated. This is determined based on IP address. Bogon refers to whether a flow corresponds to an IP address that has not yet been allocated by the IANA. Whitelist refers to a flow that has been determined to be a “good” flow.

Tagging can be hierarchical. For example, in addition to annotating a flow as an Attack flow, the Tetration system can also specify the type of attack, e.g., malware, scan, DDoS, etc. As another example, the Geo tag can classify a flow according to country, state, city, etc.

The Tetration system also enables users to tag flows based on custom tags according to rules that they define. The custom tags and rules can be input by users via the UI coupled to a Rules module. In an embodiment, the Rules module translates the user-defined tags and rules into machine-readable code (e.g., JSON, XML) to integrate the new tags into the HDFS. On the next iteration of the processing by the Compute Engine, the custom tags will be applied to the flows. The rules can be managed according to a Rule Management module that enables users to perform tag-based analytics (e.g., rank custom tags based on usage), share rules and custom tags among different tenants, associate tags to a hierarchy (e.g., classify tags as associated with certain organizations, or classify tags as relating to networking, etc.), alias tags (i.e., same rules w/different names).

Policy Simulation

Determining how changes to the data center (e.g., adding or removing a policy, modifying endpoint group membership, etc.) will affect network traffic.

Policy changes and changes to endpoint group (EPG) membership can be evaluated prior to implementing such changes in a live system. Historical ground truth flows can be used to simulate network traffic based on the policy or EPG membership changes. Real-time flows can also be used to simulate the effects on network traffic based on implementation of an experimental policy set or experimental set of EPGs.

Advantages include:

i) Capable of determining impact on an application due to changes to policies or EPG membership.

ii) Capable of determining impact of future attacks to a data center based on policy or EPG membership changes. Industry use: There does not appear to be any prior art relating to simulation of policies between endpoints or endpoint groups in a data center. However, there appear to be providers in the related space of security policy management for firewalls and network devices (e.g., AlgoSec, FireMon, SolarWinds, Skybox Security, Tufin). The Tetration policy pipeline is composed of four major steps/modules:

(1) Application Dependency Mapping

In this stage, network traffic is analyzed to determine a respective graph for each application operating in a data center (discussed in detail elsewhere). That is, particular patterns of traffic will correspond to an application, and the interconnectivity or dependencies of the application are mapped to generate a graph for the application. In this context, an “application” refers to a set of networking components that provides connectivity for a given set of workloads. For example, in a conventional three-tier architecture for application, the servers and other components of the web tier, application tier, and data tier would make up an application.

(2) Policy Generation

Whitelist rules are then derived for each application graph determined in (1) (discussed in detail elsewhere). As is known in the art, in a blacklist model, all communication is open unless explicitly denied, whereas a whitelist model requires communication to be explicitly defined before being permitted. Conventional systems use a blacklist model. One of the advantages of the Tetration system is implementation of a whitelist model, which may be more secure than a blacklist model. For instance, using a whitelist model is recognized by the Australian Signal Directorate to be the #1 approach for mitigating targeted cyber attacks (http://www.asd.gov.au/infosec/top-mitigations/top-4-strategies-explained.htm).

As an example of whitelist rule generation, suppose there is an edge of an application graph between E1 (e.g., endpoint, endpoint group) and E2. Permissible traffic flows on a set of ports of E1 to one or more ports of E2. A policy can be defined to reflect the permissible traffic from the set of ports of E1 to the one or more ports of E2.

(3) Flow Pre-Processing

After the application dependencies are mapped and the policies are defined, network traffic is pre-processed in the policy pipeline for further analysis. For each flow, the source endpoint of the flow is mapped to a source endpoint group (EPG) and the destination endpoint of the flow is mapped to a destination EPG. Each flow can also be “normalized” by determining which EPG corresponds to the client, and which EPG corresponds to the server.

4) Flow Analysis

Each pre-processed flow is then analyzed to determine various metrics, such as whether a flow is in compliance with security policies, which policies and to what extent those policies are being utilized, etc.

This flow analysis occurs continuously, and the Tetration system allows a user to specify a window of time (e.g., time of day, day of week or month, month(s) in a year, etc.) to determine the number of non-compliant events that occurred during that period.

In addition to evaluating policies actually existing in the data plane, the policy pipeline also enables “what if” analysis, such as analyzing what would happen to network traffic upon adding a new policy, removing an existing policy or changing membership of EPG groups (e.g., adding new endpoints to an EPG, removing endpoints from an EPG, moving an endpoint from one EPG to another).

In one embodiment, historical ground truth flows are utilized for simulating network traffic based on a “what if” experiment. This is referred to as back-testing. In another embodiment, real-time flows can be evaluated against an experimental policy set or experimental set of EPGs to understand how changes to particular policies or EPGs affect network traffic in the data center.

Collapsing and Placement of Applications

To provide visibility of data flows in a multi-tier application and help network teams understand the dataflow of an application and develop the application's dataflow.

The invention is directed to an application dependency map visualized in a collapsible tree flow chart. The tree flow chart is collapsible and displays the policies/relationships between each logical entity that carries a multi-tier application. The collapsible multi-tier application UI displays the data flows of a multi-tier application.

The invention is directed to an application dependency map visualized in a collapsible tree flow chart. The tree flow chart is collapsible and displays the policies/relationships between each logical entity that carries a multi-tier application. The collapsible multi-tier application UI displays the data flows of a multi-tier application. A multitier application can have various aspects of the application running on various hosts. The UI displays the hierarchy and policies or dependencies between each logical entity running the application. The UI is collapsible allowing the user to drill down on any node/logical-entity representing hosts, databases or application tier. By making the UI collapsible, it allows for a more consumable UI.

The UI displays various nodes and interacting with a node will show an exploded view of that node. A node is any logical entity. For example, any application's tier of the multitier application, database tiers, and host tiers. The exploded view of the node will explode new nodes that have edges connecting the new nodes with the exploded node. The edges represent policies between the new nodes and between the new nodes and the exploded node. For example, the original node can be a host running the application. The exploded view displays new nodes. The new nodes represent all neighbors the host communicates with. The new nods are usually exploded right of the exploded node to demonstrate the hierarchy between the logical entities.

The collapsible tree flow chart uses the data gathered from the tetration layer. Data used and made visible in the collapsible tree flow chart are (1) data flows from one logical entity to another logical entity; (2) the policies that govern the data flows from one logical entity to another logical entity; (3) what host the data flow came from; (4) what host group the data flow came from; and (5) what subnet the data flow came from. The UI is customizable. User can select elements to adjust subnet groupings and cluster groupings. Additionally the user can upload side information. Examples of side information are DNS names, host names, etc.

Currently, the problem with tree flow charts is it only shows the flow of information between parent and child. It does not show all the relationships between all the entities. Furthermore if there is a large number of parents and children, the flow chart becomes unmanageable difficult to consume.

Custom Events Processor for Network Event

Malware and other malicious processes can be very harmful on a network. Given the amount of data, flows, and processes running on a network, it can be very difficult to detect malware and malicious events. Some types of malicious events, while very harmful to the network, can be extremely difficult to detect. For example, malicious command-in-control processes can be very difficult to identify particularly when hidden. This can be complicated by the fact that certain commands, while inherently dubious, may be triggered accidentally or by fluke without any necessary malicious intent. Accordingly, it would be valuable to provide a solution that allows to capture events on a network from different perspectives and understand the different patterns to determine if a process is truly malicious or not.

This invention collects sensed data to generate a lineage of every network process. A statistical model can be implemented to then detect patterns based on the lineage of the process and identify any anomalies or malicious events.

Advantages include: This invention can provide a better understanding of processes, particularly with EPGs, and help to detect any anomalies or malicious events when a command or process is executed in the network. This invention can be implemented in a wide variety of contexts using statistical models.

Industry use: Malware and spoofing prior art solutions. However, we are not aware of any solutions that implement a statistical model to generate process lineage mappings and identify anomalies.

This invention is implemented within an architecture for observing and capturing information about network traffic in a datacenter as described below.

Network traffic coming out of a compute environment (whether from a container, VM, hardware switch, hypervisor or physical server) is captured by entities called sensors which can be deployed in or inside different environments as mentioned later. Such capturing agents will be referred to as “Sensors”. Sensors export data or metadata of the observed network activity to collection agents called “Collectors.” Collectors can be a group of processes running on a single machine or a cluster of machines. For sake of simplicity we will treat all collectors as one logical entity and refer to it as one Collector in our discussion. In actual deployment of datacenter scale, there will be more than just one collector, each responsible for handling export data from a group of sensors.

Collectors are capable of doing preprocessing and analysis of the data collected from sensors. It is capable of sending the processed or unprocessed data to a cluster of processes responsible for analysis of network data. The entities which receive the data from Collector can be a cluster of processes, and we will refer to this logical group as Pipeline. Note that sensors and collectors are not limited to observing and processing just network data, but can also capture other system information like currently active processes, active file handles, socket handles, status of I/O devices, memory, etc.

In this context, we can capture data from sensors and use the data to develop a lineage for every process. The lineage can then be used to identify anomalies as further described below.

Every process in a network can have some type of lineage. The current invention performs an analysis of commands and processes in the network to identify a lineage of a process. The lineage can be specifically important and relevant with endpoint groups (EPGs). The lineage can help identify certain types of patterns which may indicate anomalies or malicious events.

For example, the system can identify a process at system Y when command X is executed. Command X may have been observed to be triggered by command Z. We then know that the lineage for the process at system Y is command Z followed by command X. This information can be compared with processes and commands as they are executed and initialized to identify any hidden command-in-control or other anomalies.

To detect anomalies, other factors can also be taken into account. For example, factors which are inherently dubious can be used in the calculus. To illustrate, a process for running a scan on the network is inherently dubious. Thus, we can use the process lineage (i.e., lineage of the process for scanning the network) to determine if the scan was executed by a malicious command or malware. For example, if the scan follows the expected lineage mapped out for that process then we may be able to determine that the scan is legitimate or an accident/fluke. On the other hand, if the scan was triggered by an external command (i.e., command from the outside), then we can infer that this scan is part of an attack or malicious event. Similarly, if the scan does not follow the previously-established lineage (e.g., scan was started by a parent process that is not in the lineage), we can determine that the scan is part of a malicious event.

This invention can use a statistical model, such as markov chains, to study the lineage patterns and detect anomalies. The lineage patterns ascertained through the statistical model can be based on data collected by the sensors on the various devices in the network (VMs, hypervisors, switches, etc.). The statistical models and lineage information can be used in other contexts and may be applied with EPGs for understanding processes and anomalies.

The lineage information can be used to detect a command-in-control for a process and determine if the command is a hidden command or not. For example, if the command is not in the lineage, we can expect the command to be a hidden command. Hidden commands can be inherently dubious and more likely to be malicious. However, based on our statistical model, we can identify whether the hidden command may be a fluke or accident, or whether it is indeed a malicious event.

Discovering Causal Temporal Patterns

Problem to solve: event sequences reveal a temporal structure of various applications running in a computing network. Discovering temporal patterns (sequences) can be an important component of various network-related tasks, such as normalcy modeling and discovering suspect behavior, and building application profiles. There is a need to efficiently determine causal temporal patterns.

The present technology determine causal temporal patterns in a computing network based upon various attributes of network flows, such as server port, packets sent, processes involved in the communications, and timing information when data is exchanged (e.g., flowlets) is recorded (per host).

The present technology determine causal temporal patterns in a computing network based upon various attributes of network flows, such as server port, packets sent, processes involved in the communications, and timing information when data is exchanged (e.g., flowlets) is recorded (per host).

In some embodiments, event co-occurrences can be analyzed within time windows for each host to determine sequential patterns. For example, for requests from host A on a port of host D, host B either becomes a client of host D or host F for 50% of the requests.

In some embodiments, algorithms for determining temporal patterns can also be used to remove noise and co-incidences, be robust to non-deterministic relations, as well as discover and remove periodic events, and be scalable (both memory & time efficiency).

Policy Utilization

The Tetration policy pipeline is composed of four major steps/modules:

(1) Application Dependency Mapping

In this stage, network traffic is analyzed to determine a respective graph for each application operating in a data center (discussed in detail elsewhere). That is, particular patterns of traffic will correspond to an application, and the interconnectivity or dependencies of the application are mapped to generate a graph for the application. In this context, an “application” refers to a set of networking components that provides connectivity for a given set of workloads. For example, in a conventional three-tier architecture for application, the servers and other components of the web tier, application tier, and data tier would make up an application.

(2) Policy Generation Whitelist rules are then derived for each application graph determined in (1) (discussed in detail elsewhere). As is known in the art, in a blacklist model, all communication is open unless explicitly denied, whereas a whitelist model requires communication to be explicitly defined before being permitted. Conventional systems use a blacklist model. One of the advantages of the Tetration system is implementation of a whitelist model, which may be more secure than a blacklist model. For instance, using a whitelist model is recognized by the Australian Signal Directorate to be the #1 approach for mitigating targeted cyber attacks (http://www.asd.gov.au/infosec/top-mitigations/top-4-strategies-explained.htm).

As an example of whitelist rule generation, suppose there is an edge of an application graph between E1 (e.g., endpoint, endpoint group) and E2. Permissible traffic flows on a set of ports of E1 to one or more ports of E2. A policy can be defined to reflect the permissible traffic from the set of ports of E1 to the one or more ports of E2.

(3) Flow Pre-Processing

After the application dependencies are mapped and the policies are defined, network traffic is pre-processed in the policy pipeline for further analysis. For each flow, the source endpoint of the flow is mapped to a source endpoint group (EPG) and the destination endpoint of the flow is mapped to a destination EPG. Each flow can also be “normalized” by determining which EPG corresponds to the client, and which EPG corresponds to the server.

(4) Flow Analysis

Each pre-processed flow is then analyzed to determine which policies are being enforced and the extent (e.g., number of packets, number of flows, number of bytes, etc.) those policies are being enforced within the data center.

This flow analysis occurs continuously, and the Tetration system allows a user to specify a window of time (e.g., time of day, day of week or month, month(s) in a year, etc.) to determine which policies are being implemented (or not being implemented) and how often those policies are being implemented.

An ADM Pipeline to Generate Communication Graph

Problem to solve: a policy determines which nodes (computers/hosts/endpoints) can talk to which others, on which ports in a computing network. Manually building a policy is often too labor intensive and thus prohibitive (and such task needs to be done frequently, due to changes in the network).

Flow data and process information for each node of a computing network is collected. Each node is then represented by one or more vectors using such data. Nodes in the computing network can be grouped into cluster based upon similarities between the nodes. The clusters can be used to generate communication graph.

One advantage of generating communication graph is that it can help discover similar nodes. In order to generate communication graph, information needs to be collected, e.g., communication between nodes in a normal setting. In addition, the communication graph built from the network flow data has other uses: it provides visibility into the network, and makes the task of building application profiles substantially more efficient.

In some embodiments, a policy is built from a clustering as follows: for each observed edge (communication) from a node in cluster A to a node in cluster B, on server port C, a (‘white-list’) policy is introduced such that any node in cluster A can communicated with any node in cluster B on server port C.

The ADM pipeline from a high level can be: network and process data+side information graph and vector construction similarity computation and clustering policy induction and UI presentation/interaction.

User feedback from the UI can repeat this process (ie, re-run the pipeline). User feedback is incorporated into the side information.

Advantages over prior technologies: Flow and process data and a variety of auxiliary/side information are used to build clusters of nodes. The clusters can be used to induce policies and to aid other user tasks (provide visibility into the network and help build application profiles).

Directed Acyclic Graph of Down Services to Prioritize Repair

When multiple services fail, determining priorities in fixing services can be difficult. When multiple services go down, it is useful to determine the root cause of the failure. One way to predict the root cause of failure is to create a service dependency directed acyclic graph (DAG) that represents how services depend from each other. When multiple services fail, the system can try to fix the service that is highest on the hierarchy of down services. In other words, the system can create a new DAG that only represents the services that are down and focus on the root service. If multiple services are down, but there is not a clear root service, the system can focus on fixing the service that is highest on the DAG (meaning it has the most dependents, even if those dependents are currently functioning).

If services appear to be running normally but the system detects cascading anomalies (or events indicative of a problem), the system can use complex analysis to find the root cause of the anomalous behavior.

The system can determine the DAG by monitoring network data and discover what services rely on different services.

Example Devices

FIG. 10 illustrates an example network device 1010 according to some embodiments. Network device 1010 includes a master central processing unit (CPU) 1062, interfaces 1068, and a bus 1015 (e.g., a PCI bus). When acting under the control of appropriate software or firmware, the CPU 1062 is responsible for executing packet management, error detection, and/or routing functions. The CPU 1062 preferably accomplishes all these functions under the control of software including an operating system and any appropriate applications software. CPU 1062 may include one or more processors 1063 such as a processor from the Motorola family of microprocessors or the MIPS family of microprocessors. In an alternative embodiment, processor 1063 is specially designed hardware for controlling the operations of router 1010. In a specific embodiment, a memory 1061 (such as non-volatile RANI and/or ROM) also forms part of CPU 1062. However, there are many different ways in which memory could be coupled to the system.

The interfaces 1068 are typically provided as interface cards (sometimes referred to as “line cards”). Generally, they control the sending and receiving of data packets over the network and sometimes support other peripherals used with the router 1010. Among the interfaces that may be provided are Ethernet interfaces, frame relay interfaces, cable interfaces, DSL interfaces, token ring interfaces, and the like. In addition, various very high-speed interfaces may be provided such as fast token ring interfaces, wireless interfaces, Ethernet interfaces, Gigabit Ethernet interfaces, ATM interfaces, HSSI interfaces, POS interfaces, FDDI interfaces and the like. Generally, these interfaces may include ports appropriate for communication with the appropriate media. In some cases, they may also include an independent processor and, in some instances, volatile RAM. The independent processors may control such communications intensive tasks as packet switching, media control and management. By providing separate processors for the communications intensive tasks, these interfaces allow the master microprocessor 1062 to efficiently perform routing computations, network diagnostics, security functions, etc.

Although the system shown in FIG. 10 is one specific network device of the present invention, it is by no means the only network device architecture on which the present invention can be implemented. For example, an architecture having a single processor that handles communications as well as routing computations, etc. is often used. Further, other types of interfaces and media could also be used with the router.

Regardless of the network device's configuration, it may employ one or more memories or memory modules (including memory 1061) configured to store program instructions for the general-purpose network operations and mechanisms for roaming, route optimization and routing functions described herein. The program instructions may control the operation of an operating system and/or one or more applications, for example. The memory or memories may also be configured to store tables such as mobility binding, registration, and association tables, etc.

FIG. 11A and FIG. 11B illustrate example system embodiments. The more appropriate embodiment will be apparent to those of ordinary skill in the art when practicing the present technology. Persons of ordinary skill in the art will also readily appreciate that other system embodiments are possible.

FIG. 11A illustrates a conventional system bus computing system architecture 1100 wherein the components of the system are in electrical communication with each other using a bus 1105. Exemplary system 1100 includes a processing unit (CPU or processor) 1110 and a system bus 1105 that couples various system components including the system memory 1115, such as read only memory (ROM) 1120 and random access memory (RAM) 1125, to the processor 1110. The system 1100 can include a cache of high-speed memory connected directly with, in close proximity to, or integrated as part of the processor 1110. The system 1100 can copy data from the memory 1115 and/or the storage device 1130 to the cache 1112 for quick access by the processor 1110. In this way, the cache can provide a performance boost that avoids processor 1110 delays while waiting for data. These and other modules can control or be configured to control the processor 1110 to perform various actions. Other system memory 1115 may be available for use as well. The memory 1115 can include multiple different types of memory with different performance characteristics. The processor 1110 can include any general purpose processor and a hardware module or software module, such as module 1 1132, module 2 1134, and module 3 1136 stored in storage device 1130, configured to control the processor 1110 as well as a special-purpose processor where software instructions are incorporated into the actual processor design. The processor 1110 may essentially be a completely self-contained computing system, containing multiple cores or processors, a bus, memory controller, cache, etc. A multi-core processor may be symmetric or asymmetric.

To enable user interaction with the computing device 1100, an input device 1145 can represent any number of input mechanisms, such as a microphone for speech, a touch-sensitive screen for gesture or graphical input, keyboard, mouse, motion input, speech and so forth. An output device 1135 can also be one or more of a number of output mechanisms known to those of skill in the art. In some instances, multimodal systems can enable a user to provide multiple types of input to communicate with the computing device 1100. The communications interface 1140 can generally govern and manage the user input and system output. There is no restriction on operating on any particular hardware arrangement and therefore the basic features here may easily be substituted for improved hardware or firmware arrangements as they are developed.

Storage device 1130 is a non-volatile memory and can be a hard disk or other types of computer readable media which can store data that are accessible by a computer, such as magnetic cassettes, flash memory cards, solid state memory devices, digital versatile disks, cartridges, random access memories (RAMs) 1125, read only memory (ROM) 1120, and hybrids thereof.

The storage device 1130 can include software modules 1132, 1134, 1136 for controlling the processor 1110. Other hardware or software modules are contemplated. The storage device 1130 can be connected to the system bus 1105. In one aspect, a hardware module that performs a particular function can include the software component stored in a computer-readable medium in connection with the necessary hardware components, such as the processor 1110, bus 1105, display 1135, and so forth, to carry out the function.

FIG. 11B illustrates an example computer system 1150 having a chipset architecture that can be used in executing the described method and generating and displaying a graphical user interface (GUI). Computer system 1150 is an example of computer hardware, software, and firmware that can be used to implement the disclosed technology. System 1150 can include a processor 1155, representative of any number of physically and/or logically distinct resources capable of executing software, firmware, and hardware configured to perform identified computations. Processor 1155 can communicate with a chipset 1160 that can control input to and output from processor 1155. In this example, chipset 1160 outputs information to output device 1165, such as a display, and can read and write information to storage device 1170, which can include magnetic media, and solid state media, for example. Chipset 1160 can also read data from and write data to RANI 1175. A bridge 1180 for interfacing with a variety of user interface components 1185 can be provided for interfacing with chipset 1160. Such user interface components 1185 can include a keyboard, a microphone, touch detection and processing circuitry, a pointing device, such as a mouse, and so on. In general, inputs to system 1150 can come from any of a variety of sources, machine generated and/or human generated.

Chipset 1160 can also interface with one or more communication interfaces 1190 that can have different physical interfaces. Such communication interfaces can include interfaces for wired and wireless local area networks, for broadband wireless networks, as well as personal area networks. Some applications of the methods for generating, displaying, and using the GUI disclosed herein can include receiving ordered datasets over the physical interface or be generated by the machine itself by processor 1155 analyzing data stored in storage 1170 or 1175. Further, the machine can receive inputs from a user via user interface components 1185 and execute appropriate functions, such as browsing functions by interpreting these inputs using processor 1155.

It can be appreciated that example systems 1100 and 1150 can have more than one processor 1110 or be part of a group or cluster of computing devices networked together to provide greater processing capability.

For clarity of explanation, in some instances the present technology may be presented as including individual functional blocks including functional blocks comprising devices, device components, steps or routines in a method embodied in software, or combinations of hardware and software.

In some embodiments the computer-readable storage devices, mediums, and memories can include a cable or wireless signal containing a bit stream and the like. However, when mentioned, non-transitory computer-readable storage media expressly exclude media such as energy, carrier signals, electromagnetic waves, and signals per se.

Methods according to the above-described examples can be implemented using computer-executable instructions that are stored or otherwise available from computer readable media. Such instructions can comprise, for example, instructions and data which cause or otherwise configure a general purpose computer, special purpose computer, or special purpose processing device to perform a certain function or group of functions. Portions of computer resources used can be accessible over a network. The computer executable instructions may be, for example, binaries, intermediate format instructions such as assembly language, firmware, or source code. Examples of computer-readable media that may be used to store instructions, information used, and/or information created during methods according to described examples include magnetic or optical disks, flash memory, USB devices provided with non-volatile memory, networked storage devices, and so on.

Devices implementing methods according to these disclosures can comprise hardware, firmware and/or software, and can take any of a variety of form factors. Typical examples of such form factors include laptops, smart phones, small form factor personal computers, personal digital assistants, rackmount devices, standalone devices, and so on. Functionality described herein also can be embodied in peripherals or add-in cards. Such functionality can also be implemented on a circuit board among different chips or different processes executing in a single device, by way of further example.

The instructions, media for conveying such instructions, computing resources for executing them, and other structures for supporting such computing resources are means for providing the functions described in these disclosures.

Although a variety of examples and other information was used to explain aspects within the scope of the appended claims, no limitation of the claims should be implied based on particular features or arrangements in such examples, as one of ordinary skill would be able to use these examples to derive a wide variety of implementations. Further and although some subject matter may have been described in language specific to examples of structural features and/or method steps, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to these described features or acts. For example, such functionality can be distributed differently or performed in components other than those identified herein. Rather, the described features and steps are disclosed as examples of components of systems and methods within the scope of the appended claims. Moreover, claim language reciting “at least one of a set indicates that one member of the set or multiple members of the set satisfy the claim.

It should be understood that features or configurations herein with reference to one embodiment or example can be implemented in, or combined with, other embodiments or examples herein. That is, terms such as “embodiment”, “variation”, “aspect”, “example”, “configuration”, “implementation”, “case”, and any other terms which may connote an embodiment, as used herein to describe specific features or configurations, are not intended to limit any of the associated features or configurations to a specific or separate embodiment or embodiments, and should not be interpreted to suggest that such features or configurations cannot be combined with features or configurations described with reference to other embodiments, variations, aspects, examples, configurations, implementations, cases, and so forth. In other words, features described herein with reference to a specific example (e.g., embodiment, variation, aspect, configuration, implementation, case, etc.) can be combined with features described with reference to another example. Precisely, one of ordinary skill in the art will readily recognize that the various embodiments or examples described herein, and their associated features, can be combined with each other.

A phrase such as an “aspect” does not imply that such aspect is essential to the subject technology or that such aspect applies to all configurations of the subject technology. A disclosure relating to an aspect may apply to all configurations, or one or more configurations. A phrase such as an aspect may refer to one or more aspects and vice versa. A phrase such as a “configuration” does not imply that such configuration is essential to the subject technology or that such configuration applies to all configurations of the subject technology. A disclosure relating to a configuration may apply to all configurations, or one or more configurations. A phrase such as a configuration may refer to one or more configurations and vice versa. The word “exemplary” is used herein to mean “serving as an example or illustration.” Any aspect or design described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other aspects or designs. Moreover, claim language reciting “at least one of a set indicates that one member of the set or multiple members of the set satisfy the claim.

Claims

1. A non-transitory computer-readable media encoding a set of non-transitory computer-readable instructions, which when executed on one or more processors on devices connected to a network, cause one or more devices to: at a first device, receive a stream of network flow data via an attached communications network;evaluate the stream of network flow data to derive a directed control flow graph corresponding to a distributed application, the control flow graph including a plurality of nodes and a plurality of edges between various nodes, wherein: the nodes of the graph correspond to network-addressable application components connected to the communications network, each application component sending and receiving network traffic including one or more packets at a network interface;one or more of the application components includes a workload creating and/or processing a data stream as part of the distributed application; andone or more first edges of the plurality of edges between the nodes of the graph correspond to data streams between source nodes and destination nodes;annotate one or more flows associated with one or more nodes plurality of nodes and/or one or more second edges plurality of edges in the control flow graph with one or more tags, the tags relating to a functioning of the distributed application;identify patterns of normal behavior of the distributed application; anddisplay a representation of the distributed application.

2. The non-transitory computer-readable media of claim 1, further comprising computer-readable instructions that cause a second device to generate network flow data based upon the packets being sent and/or received via a network interface local to the second device, and send generated network flow data to the first device.

3. The non-transitory computer-readable media of claim 2, further comprising computer-readable instructions that, when executed, cause the one or more devices to enable a user to search and query information associated with the directed control flow graph.

4. The non-transitory computer-readable media of claim 2, wherein the one or more first and/or the one or more second edges in the directed control flow graph correspond to dependencies between services in the distributed application, and wherein the computer-readable media further comprise instructions that display to the user the dependencies between services.

5. The non-transitory computer-readable media of claim 2, further comprising computer-readable instructions that, when executed, cause the one or more devices to evaluate newly received information from the stream of network flow data against the patterns and identify a change in conditions, the change in conditions including at least one of network traffic that varies from the normal behavior expected by the patterns and a change in the control flow graph; and respond to the change in conditions, wherein responding to the change in conditions includes updating the displayed representation of the distributed application.

6. The non-transitory computer-readable media of claim 2, wherein the patterns are identified by a machine learning model.

7. The non-transitory computer-readable media of claim 6, further comprising computer-readable instructions that, when executed, cause the one or more devices to use machine learning techniques to identify patterns that are desirable or unwanted.

8. The non-transitory computer-readable media of claim 7, further comprising computer-readable instructions that, when executed, cause the one or more devices to enforce and/or modify a network traffic policy that mitigates an unwanted network traffic pattern.

9. The non-transitory computer-readable media of claim 5, wherein responding to the change in conditions includes sending an alert.

10. The non-transitory computer-readable media of claim 2, further comprising computer-readable instructions that, when executed, cause the one or more devices to persist at least a portion of information from the stream of network flow data via a storage facility accessible to the first device.

11. The non-transitory computer-readable media of claim 2, further comprising computer-readable instructions that, when executed, cause the one or more devices to present a web-based user interface to the user.

12. The non-transitory computer-readable media of claim 1, further comprising computer-readable instructions that, when executed, cause the one or more devices to enable a user to search and query information associated with the control flow graph.

13. The non-transitory computer-readable media of claim 1, wherein the one or more first and/or the one or more second edges in the directed control flow graph correspond to dependencies between services in the distributed application, and wherein the computer-readable media further comprise instructions that display to the user the dependencies between services.

14. The non-transitory computer-readable media of claim 1, further comprising computer-readable instructions that, when executed, cause the one or more devices to evaluate newly received information from the stream of network flow data against the patterns and identify a change in conditions, the change in conditions including at least one of network traffic that varies from the normal behavior expected by the patterns and a change in the control flow graph; and respond to the change in conditions, wherein responding to the change in conditions includes updating the displayed representation of the distributed application.

15. The non-transitory computer-readable media of claim 14, wherein responding to the change in conditions includes sending an alert.

16. The non-transitory computer-readable media of claim 15, further comprising computer-readable instructions that, when executed, cause the one or more devices to use machine learning techniques to identify patterns that are desirable or unwanted.

17. The non-transitory computer-readable media of claim 16, further comprising computer-readable instructions that, when executed, cause the one or more devices to enforce and/or modify a network traffic policy that mitigates an unwanted network traffic pattern.

18. The non-transitory computer-readable media of claim 1, wherein the patterns are identified by a machine learning model.

19. The non-transitory computer-readable media of claim 1, further comprising computer-readable instructions that, when executed, cause the one or more devices to persist at least a portion of information from the stream of network flow data via a storage facility accessible to the first device.

20. The non-transitory computer-readable media of claim 1, further comprising computer-readable instructions that, when executed, cause the one or more devices to present a web-based user interface to the user.

21. A method of monitoring network traffic, the method comprising: at a first device, receiving a stream of network flow data via an attached communications network;evaluating the stream of network flow data to derive a directed control flow graph corresponding to a distributed application, the control flow graph including a plurality of nodes and a plurality of edges between various nodes, wherein: the nodes of the graph correspond to network-addressable application components connected to the communications network, each application component sending and receiving network traffic including one or more packets at a network interface;one or more of the application components includes a container-hosted workload creating and/or processing a data stream as part of the distributed application; andone or more first edges of the plurality of edges between the nodes of the graph correspond to data streams between source nodes and destination nodes;annotating one or more flows associated with one or more nodes of the plurality of nodes and/or one or more second edges of the plurality of edges in the control flow graph with one or more tags, the tags relating to a functioning of the distributed application;identifying patterns of normal behavior of the distributed application; anddisplaying a representation of the distributed application.

22. The method of claim 21, further comprising generating, at a second device, network flow data based upon the packets being sent and/or received via a network interface local to the second device; and sending generated network flow data from the second device to the first device.

23. The method of claim 22, further comprising receiving a user-provided query and searching information associated with the control flow graph based on the query.

24. The method of claim 22, wherein the patterns are identified by a machine learning model.

25. The method of claim 24, further comprising using machine learning techniques to identify traffic patterns that represent normal operation of the distributed application, and traffic patterns which may indicate anomalies in or attacks against the distributed application.

26. The method of claim 21, further comprising enforcing and/or modifying a network traffic policy to mitigate an anomalous or attacking network traffic pattern.

27. The method of claim 21, further comprising receiving a user-provided query and searching information associated with the control flow graph based on the query.

28. The method of claim 21, wherein the patterns are identified by a machine learning model.

29. The method of claim 28, further comprising using machine learning techniques to identify traffic patterns that represent normal operation of the distributed application, and traffic patterns which may indicate anomalies in or attacks against the distributed application.

30. A system comprising: one or more processors; andnon-transitory computer-readable media encoding instructions which, when executed by the one or more processors, cause one or more processors to: receive, at a first device, a stream of network flow data via an attached communications network;evaluate the stream of network flow data to derive a directed control flow graph corresponding to a distributed application, the control flow graph including a plurality of nodes and a plurality of edges between various nodes, wherein: the nodes of the graph correspond to network-addressable application components connected to the communications network, each application component sending and receiving network traffic including one or more packets at a network interface;one or more of the application components includes a workload creating and/or processing a data stream as part of the distributed application; andone or more first edges of the plurality of edges between the nodes of the graph correspond to data streams between source nodes and destination nodes;annotate one or more flows associated with one or more nodes of the plurality of nodes and/or one or more second edges of the plurality of edges in the control flow graph with one or more tags, the tags relating to a functioning of the distributed application;identify patterns of normal behavior of the distributed application; anddisplay a representation of the distributed application.