Computer data system data source having an update propagation graph with feedback cyclicality

Abstract

Described are methods, systems and computer readable media for data source refreshing using an update propagation graph with feedback cyclicality.

Description

Embodiments relate generally to computer data systems, and more particularly, to methods, systems and computer readable media for data source refreshing using an update propagation graph with feedback cyclicality.

Data sources or objects within a computer data system may include static sources and dynamic sources. Some data sources or objects (e.g., tables) may depend on other data sources. As new data is received or obtained for dynamic data sources, those dynamic data sources may be refreshed (or updated). Data sources or objects that are dependent on one or more dynamic sources that have been refreshed may also need to be refreshed. The refreshing of data sources may need to be performed in an order based on dependencies. The dependencies may be defined by an update propagation graph. A need may exist to provide a feedback mechanism within the update propagation graph for purposes such as backtesting of a computer data system or of a data model or technique associated with the computer data system.

Some implementations were conceived in light of the above mentioned needs, problems and/or limitations, among other things.

Some implementations can include a system for updating a data object using an update propagation graph having a cyclicality feedback provider. The system can include one or more hardware processors coupled to a nontransitory computer readable medium having stored thereon software instructions that, when executed by the one or more processors, cause the one or more processors to perform operations. The operations can include constructing a cyclicality feedback provider including a cyclicality feedback provider object including one or more feedback data fields, and obtaining a reference to the cyclicality feedback provider object. The operations can also include constructing a computer data system update propagation graph having one or more update propagation graph data fields that correspond to the one or more feedback data fields. The operations can further include adding a feedback provider listener to the computer data system update propagation graph, wherein the feedback provider listener provides feedback updates to the one or more feedback data fields of the cyclicality feedback provider object when changes to the one or more update propagation graph data fields corresponding to the one or more feedback data fields are detected, and wherein the feedback updates are provided to the one or more feedback data fields of the cyclicality feedback provider object based on a state of a logical clock and on completion of update processing for a given logical clock cycle.

The cyclicality feedback provider object can include a computer data system table object. The update propagation graph can include a hybrid directed acyclic graph having a clock-state controlled cyclicality feedback provided by the cyclicality feedback provider and a state of a logical clock.

The operations can also include determining that a logical clock has transitioned to an update state, and processing events and updates to data sources of the update propagation graph for a current logical clock cycle, wherein processing the events and updates are performed on the hybrid directed acyclic graph as if the cyclicality feedback is not present. The operations can further include after the processing events and updates has completed, providing events and updates from the cyclicality feedback provider object to one or more data objects within the update propagation graph, wherein the events and updates from the cyclicality feedback provider object will be processed through the update propagation graph in a next logical clock cycle.

Processing events and updates to the data sources can include invoking a data source refresh method for a data source for which changes are being processed, determining whether a priority queue for the data source is empty. The operations can also include, when the priority queue is not empty, retrieving a next change notification message from the priority queue and delivering the change notification to a corresponding data source and repeating determining whether the priority is queue is empty. The operations can further include when the priority queue is empty, setting the logical clock to an idle state.

The operations can also include performing a backtesting operation by providing predetermined input data to the update propagation graph as one or more events and updates to one or more data sources, and receiving output results from the update propagation graph for each logical clock cycle. The operations can further include comparing the output results received from the update propagation graph with one or more reference values, and generating an output signal based on the comparing.

Some implementations can include a computer-implemented method for updating a data object using an update propagation graph having a cyclicality feedback provider. The method can include constructing a cyclicality feedback provider including a cyclicality feedback provider object including one or more feedback data fields, and obtaining a reference to the cyclicality feedback provider object. The method can also include constructing a computer data system update propagation graph having one or more update propagation graph data fields that correspond to the one or more feedback data fields, and adding a feedback provider listener to the computer data system update propagation graph, wherein the feedback provider listener provides feedback updates to the one or more feedback data fields of the cyclicality feedback provider object when changes to the one or more update propagation graph data fields corresponding to the one or more feedback data fields are detected, and wherein the feedback updates are provided to the one or more feedback data fields of the cyclicality feedback provider object based on a state of a logical clock and on completion of update processing for a given logical clock cycle.

The cyclicality feedback provider object can include a computer data system table object. The update propagation graph can include a hybrid directed acyclic graph having a clock-state controlled cyclicality feedback provided by the cyclicality feedback provider and a state of a logical clock.

The method can also include determining that a logical clock has transitioned to an update state, and processing events and updates to data sources of the update propagation graph for a current logical clock cycle, wherein processing the events and updates are performed on the hybrid directed acyclic graph as if the cyclicality feedback is not present. The method can further include after the processing events and updates has completed, providing events and updates from the cyclicality feedback provider object to one or more data objects within the update propagation graph, wherein the events and updates from the cyclicality feedback provider object will be processed through the update propagation graph in a next logical clock cycle.

Processing events and updates to the data sources can include invoking a data source refresh method for a data source for which changes are being processed, and determining whether a priority queue for the data source is empty. The method can also include, when the priority queue is not empty, retrieving a next change notification message from the priority queue and delivering the change notification to a corresponding data source and repeating determining whether the priority is queue is empty. The method can further include when the priority queue is empty, setting the logical clock to an idle state.

The method can also include performing a backtesting operation by providing predetermined input data to the update propagation graph as one or more events and updates to one or more data sources, and receiving output results from the update propagation graph for each logical clock cycle. The method can further include comparing the output results received from the update propagation graph with one or more reference values, and generating an output signal based on the comparing.

Some implementations can include a nontransitory computer readable medium having stored thereon software instructions that, when executed by one or more processors, cause the one or more processors to perform operations. The operations can include constructing a cyclicality feedback provider including a cyclicality feedback provider object including one or more feedback data fields, and obtaining a reference to the cyclicality feedback provider object. The operations can also include constructing a computer data system update propagation graph having one or more update propagation graph data fields that correspond to the one or more feedback data fields, and adding a feedback provider listener to the computer data system update propagation graph, wherein the feedback provider listener provides feedback updates to the one or more feedback data fields of the cyclicality feedback provider object when changes to the one or more update propagation graph data fields corresponding to the one or more feedback data fields are detected, and wherein the feedback updates are provided to the one or more feedback data fields of the cyclicality feedback provider object based on a state of a logical clock and on completion of update processing for a given logical clock cycle.

The cyclicality feedback provider object can include a computer data system table object. The update propagation graph can include a hybrid directed acyclic graph having a clock-state controlled cyclicality feedback provided by the cyclicality feedback provider and a state of a logical clock. The operations can also include determining that a logical clock has transitioned to an update state, and processing events and updates to data sources of the update propagation graph for a current logical clock cycle, wherein processing the events and updates are performed on the hybrid directed acyclic graph as if the cyclicality feedback is not present. The operations can further include, after the processing events and updates has completed, providing events and updates from the cyclicality feedback provider object to one or more data objects within the update propagation graph, wherein the events and updates from the cyclicality feedback provider object will be processed through the update propagation graph in a next logical clock cycle.

Processing events and updates to the data sources can include invoking a data source refresh method for a data source for which changes are being processed, and determining whether a priority queue for the data source is empty. The operations can also include, when the priority queue is not empty, retrieving a next change notification message from the priority queue and delivering the change notification to a corresponding data source and repeating determining whether the priority is queue is empty. The operations can further include, when the priority queue is empty, setting the logical clock to an idle state.

The operations can also include performing a backtesting operation by providing predetermined input data to the update propagation graph as one or more events and updates to one or more data sources, and receiving output results from the update propagation graph for each logical clock cycle. The operations can further include comparing the output results received from the update propagation graph with one or more reference values, and generating an output signal based on the comparing.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a diagram of an example computer data system showing an example data distribution configuration in accordance with some implementations.

FIG. 2 is a diagram of an example computer data system showing an example administration/process control arrangement in accordance with some implementations.

FIG. 3 is a diagram of an example computing device configured for GUI control element processing in accordance with some implementations.

FIGS. 4A and 4B show example data source definitions and a corresponding directed acyclic graph (DAG) in accordance with some implementations.

FIGS. 5A and 5B show example data source definitions and a corresponding hybrid DAG having a cyclicality feedback provider in accordance with some implementations.

FIG. 6 is a flowchart of an example cyclicality feedback provider process in accordance with some implementations.

FIG. 7 is a flowchart of an example data source refresh process using a hybrid DAG having a cyclicality feedback provider in accordance with some implementations.

FIG. 8 is a diagram of an example computer data system and backtesting application using a hybrid DAG having a cyclicality feedback provider in accordance with some implementations.

FIG. 9 is a diagram of an example machine learning system using a hybrid DAG having a cyclicality feedback provider in accordance with some implementations.

DETAILED DESCRIPTION

Reference may be made herein to the Java programming language, Java classes, Java bytecode and the Java Virtual Machine (JVM) for purposes of illustrating example implementations. It will be appreciated that implementations can include other programming languages (e.g., groovy, Scala, R, Go, etc.), other programming language structures as an alternative to or in addition to Java classes (e.g., other language classes, objects, data structures, program units, code portions, script portions, etc.), other types of bytecode, object code and/or executable code, and/or other virtual machines or hardware implemented machines configured to execute a data system query.

FIG. 1 is a diagram of an example computer data system and network 100 showing an example data distribution configuration in accordance with some implementations. In particular, the system 100 includes an application host 102, a periodic data import host 104, a query server host 106, a long-term file server 108, and a user data import host 110. While tables are used as an example data object in the description below, it will be appreciated that the data system described herein can also process other data objects such as mathematical objects (e.g., a singular value decomposition of values in a given range of one or more rows and columns of a table), TableMap objects, etc. A TableMap object provides the ability to lookup a Table by some key. This key represents a unique value (or unique tuple of values) from the columns aggregated on in a by External( ) statement execution, for example. A TableMap object can be the result of a by External( ) statement executed as part of a query. It will also be appreciated that the configurations shown in FIGS. 1 and 2 are for illustration purposes and in a given implementation each data pool (or data store) may be directly attached or may be managed by a file server.

The application host 102 can include one or more application processes 112, one or more log files 114 (e.g., sequential, row-oriented log files), one or more data log tailers 116 and a multicast key-value publisher 118. The periodic data import host 104 can include a local table data server, direct or remote connection to a periodic table data store 122 (e.g., a column-oriented table data store) and a data import server 120. The query server host 106 can include a multicast key-value subscriber 126, a performance table logger 128, local table data store 130 and one or more remote query processors (132, 134) each accessing one or more respective tables (136, 138). The long-term file server 108 can include a long-term data store 140. The user data import host 110 can include a remote user table server 142 and a user table data store 144. Row-oriented log files and column-oriented table data stores are discussed herein for illustration purposes and are not intended to be limiting. It will be appreciated that log files and/or data stores may be configured in other ways. In general, any data stores discussed herein could be configured in a manner suitable for a contemplated implementation.

In operation, the input data application process 112 can be configured to receive input data from a source (e.g., a securities trading data source), apply schema-specified, generated code to format the logged data as it's being prepared for output to the log file 114 and store the received data in the sequential, row-oriented log file 114 via an optional data logging process. In some implementations, the data logging process can include a daemon, or background process task, that is configured to log raw input data received from the application process 112 to the sequential, row-oriented log files on disk and/or a shared memory queue (e.g., for sending data to the multicast publisher 118). Logging raw input data to log files can additionally serve to provide a backup copy of data that can be used in the event that downstream processing of the input data is halted or interrupted or otherwise becomes unreliable.

A data log tailer 116 can be configured to access the sequential, row-oriented log file(s) 114 to retrieve input data logged by the data logging process. In some implementations, the data log tailer 116 can be configured to perform strict byte reading and transmission (e.g., to the data import server 120). The data import server 120 can be configured to store the input data into one or more corresponding data stores such as the periodic table data store 122 in a column-oriented configuration. The periodic table data store 122 can be used to store data that is being received within a time period (e.g., a minute, an hour, a day, etc.) and which may be later processed and stored in a data store of the long-term file server 108. For example, the periodic table data store 122 can include a plurality of data servers configured to store periodic securities trading data according to one or more characteristics of the data (e.g., a data value such as security symbol, the data source such as a given trading exchange, etc.).

The data import server 120 can be configured to receive and store data into the periodic table data store 122 in such a way as to provide a consistent data presentation to other parts of the system. Providing/ensuring consistent data in this context can include, for example, recording logged data to a disk or memory, ensuring rows presented externally are available for consistent reading (e.g., to help ensure that if the system has part of a record, the system has all of the record without any errors), and preserving the order of records from a given data source. If data is presented to clients, such as a remote query processor (132, 134), then the data may be persisted in some fashion (e.g., written to disk).

The local table data server 124 can be configured to retrieve data stored in the periodic table data store 122 and provide the retrieved data to one or more remote query processors (132, 134) via an optional proxy.

The remote user table server (RUTS) 142 can include a centralized consistent data writer, as well as a data server that provides processors with consistent access to the data that it is responsible for managing. For example, users can provide input to the system by writing table data that is then consumed by query processors.

The remote query processors (132, 134) can use data from the data import server 120, local table data server 124 and/or from the long-term file server 108 to perform queries. The remote query processors (132, 134) can also receive data from the multicast key-value subscriber 126, which receives data from the multicast key-value publisher 118 in the application host 102. The performance table logger 128 can log performance information about each remote query processor and its respective queries into a local table data store 130. Further, the remote query processors can also read data from the RUTS, from local table data written by the performance logger, or from user table data read over NFS, for example.

It will be appreciated that the configuration shown in FIG. 1 is a typical example configuration that may be somewhat idealized for illustration purposes. An actual configuration may include one or more of each server and/or host type. The hosts/servers shown in FIG. 1 (e.g., 102-110, 120, 124 and 142) may each be separate or two or more servers may be combined into one or more combined server systems. Data stores can include local/remote, shared/isolated and/or redundant. Any table data may flow through optional proxies indicated by an asterisk on certain connections to the remote query processors. Also, it will be appreciated that the term “periodic” is being used for illustration purposes and can include, but is not limited to, data that has been received within a given time period (e.g., millisecond, second, minute, hour, day, week, month, year, etc.) and which has not yet been stored to a long-term data store (e.g., 140).

FIG. 2 is a diagram of an example computer data system 200 showing an example administration/process control arrangement in accordance with some implementations. The system 200 includes a production client host 202, a controller host 204, a GUI host or workstation 206, and query server hosts 208 and 210. It will be appreciated that there may be one or more of each of 202-210 in a given implementation.

The production client host 202 can include a batch query application 212 (e.g., a query that is executed from a command line interface or the like) and a real time query data consumer process 214 (e.g., an application that connects to and listens to tables created from the execution of a separate query). The batch query application 212 and the real time query data consumer 214 can connect to a remote query dispatcher 222 and one or more remote query processors (224, 226) within the query server host 1208.

The controller host 204 can include a persistent query controller 216 configured to connect to a remote query dispatcher 232 and one or more remote query processors 228-230. In some implementations, the persistent query controller 216 can serve as the “primary client” for persistent queries and can request remote query processors from dispatchers, and send instructions to start persistent queries. For example, a user can submit a query to 216, and 216 starts and runs the query every day. In another example, a securities trading strategy could be a persistent query. The persistent query controller can start the trading strategy query every morning before the market opened, for instance. It will be appreciated that 216 can work on times other than days. In some implementations, the controller may require its own clients to request that queries be started, stopped, etc. This can be done manually, or by scheduled (e.g., cron) jobs. Some implementations can include “advanced scheduling” (e.g., auto-start/stop/restart, time-based repeat, etc.) within the controller.

The GUI/host workstation can include a user console 218 and a user query application 220. The user console 218 can be configured to connect to the persistent query controller 216. The user query application 220 can be configured to connect to one or more remote query dispatchers (e.g., 232) and one or more remote query processors (228, 230).

FIG. 3 is a diagram of an example computing device 300 in accordance with at least one implementation. The computing device 300 includes one or more processors 302, operating system 304, computer readable medium 306 and network interface 308. The memory 306 can include a feedback cyclicality application 310 and a data section 312 (e.g., for storing in-memory tables, etc.).

In operation, the processor 302 may execute the application 310 stored in the memory 306. The application 310 can include software instructions that, when executed by the processor, cause the processor to perform operations for data source refreshing using an update propagation graph with feedback cyclicality in accordance with the present disclosure (e.g., performing one or more of 602-618 and/or 702-708 described below).

The application program 310 can operate in conjunction with the data section 312 and the operating system 304.

As used herein, a data source can include, but is not limited to, a real time or near real time data source such as securities market data (e.g., over a multicast distribution mechanism (e.g., 118/126) or through a tailer (e.g., 116), system generated data, historical data, user input data from a remote user table server, tables programmatically generated in-memory, or an element upstream in an update propagation graph (UPG) such as a directed acyclic graph (DAG), and/or any data (e.g., a table, mathematical object, etc.) having a capability to refresh itself/provide updated data.

When a data source is updated, it will send add, delete, modify, reindex (AMDR) notifications through the DAG. It will be appreciated that a DAG is used herein for illustration purposes of a possible implementation of the UPG, and that the UPG can include other implementations. A reindex message is a message to change the indexing of a data item, but not change the value. When a table is exported from the server to a client, there is an exported table handle created and that handle attaches itself to the DAG; as a child of the table to be displayed. When the DAG updates, that handle's node in the DAG is reached and a notification is sent across the network to the client that includes the rows which have been added/modified/deleted/reindexed. On the client side, those rows are reconstructed and an in-memory copy of the table (or portion thereof) is maintained for display (or other access).

There can be two cases in which a view is updated. In the first case, a system clock ticks, and there is new data for one or more source (parent) nodes in the DAG, which percolates down to the exported table handle. In the second case, a user changes the “viewport”, which is the active set of rows and columns.

There can be various ways the viewport is caused to be updated, such as: (i) scrolling the view of the table, (ii) showing or hiding a table, (iii) when the user or client program programmatically accesses the table, and/or (iv) adding/removing columns from a view. When the viewport is updated, the viewport is automatically adjusted to include the rows/columns that the user is trying to access with exponential expansion up to a limit for efficiency. After a timeout, any automatically created viewports are closed.

A query result may not change without a clock tick that has one or more AMDR messages which traverse the DAG. However, the portion of a query result that is displayed by the user (e.g., the viewport) might change. When a user displays a table, a set of visible columns and rows is computed. In addition to the visible set of rows/columns, the system may compute (and make available for possible display) more data than is visible. For example, the system may compute and make available for possible display three screens of data: the currently visible screen and one screen before and one screen after. If there are multiple views of the same table, either multiple exported table handles are created in which case the views are independent or if a single exported table handle is created, the viewport is the union of the visible sets. As the user scrolls the table, the viewport may change. When the viewport changes, the visible area (with a buffer of rows up and down, and columns left and right, so that scrolling is smooth) is computed and the updated visible area is sent to the server. In response, the server sends a snapshot with relevant portions of those newly visible rows/columns. For non-displayed tables, the visible area can be considered the whole table by the system for further processing so that a consistent table view is available for further processing (e.g., all rows and one or more columns of the data object may be sent to the client).

The snapshot can be generated asynchronously from the DAG update/table refresh loop under the condition that a consistent snapshot (i.e., the clock value remains the same throughout the snapshot) is able to be obtained. If a consistent snapshot is not obtained after a given number of attempts (e.g., three attempts), a lock can be obtained (e.g., the LiveTableMonitor lock) at the end of the current DAG update cycle to lock out updates while the snapshot is created.

Further, the remote query processor (or server) has knowledge of the visible regions and will send data updates for the visible rows/columns (e.g., it can send the entire AMDR message information so the client has information about what has been updated, just not what the actual data is outside of its viewport). This enables the client optionally to cache data even if the data is outside the viewport and only invalidate the data once the data actually changes.

The DAG structure can be maintained in the memory of a remote query processor. Child nodes have hard references back to their parents, and parents have weak references to their children. This ensures that if a child exists, its parent will also exist, but if there are no external references to a child, then a garbage collection event can properly clean the child up (and the parent won't hold onto the child). For the exported table handles, a component (e.g., an ExportedTableHandleManager component) can be configured to hold hard references to the exported tables. If a client disconnects, then the references for its tables can be cleaned up. Clients can also proactively release references.

FIGS. 4A and 4B show data source definitions and a corresponding directed acyclic graph (DAG) in accordance with some implementations. In FIG. 4A, example code defines the data sources as tables (t1-t5). From the code for the data sources, a DAG can be generated as shown by the graph in FIG. 4B. The DAG in FIG. 4B shows dependencies between the nodes, which correspond to table data sources.

Data sources can include market data (e.g., data received via multicast distribution mechanism or through a tailer), system generated data, historical data, user input data from the remote user table server, tables programmatically generated in-memory, or something further upstream in the DAG. In general, anything represented in the data system as a table and which can refresh itself/provide data can be a data source. Also, data sources can include non-table data structures which update, for example, mathematical data structures such as a singular value decomposition (SVD) of a table. Similarly, correlation matrices, linear algebra, PDE solvers, a non-matrix, non-tabular data object, etc. can be supported.

In some implementations, code can be converted into the in-memory data structures holding the DAG. For example, the source code of FIG. 4A gets converted into the DAG data structure in memory. The DAG connectivity can change by executing code. For example, assume a set of code CODE1 is executed. CODE1 leads to a DAG1 being created. Data can be processed through DAG1, leading to table updates. Now assume that the user wants to compute a few more tables. The user can run a few more lines of code CODE2, which use variables computed in CODE1. The execution of CODE2 leads to a change in the DAG. As a simple example, assume that the first 3 lines in FIG. 4A are executed. The user could come along later and execute line 4, which would modify the DAG data structure. Also, some implementations can permit other programs to listen to changes from a node representing a data object (e.g., table or non-table object) or an internal node.

In some implementations, when a table changes, an application programming interface (API) can specify rows where add, modify, delete, or reindex (AMDR) changes were made. A reindex is a change in which a row is moved but the value contained in the row is not modified. The API can also provide a mechanism to obtain a value prior to the most recent change. When the DAG is processed during the refresh, the AMD info on “upstream” data objects (e.g., tables, etc.) or nodes is used to compute changes in “downstream” data objects or nodes. In some implementations, the entire DAG can be processed during the refresh cycle.

In general, a DAG can be comprised of a) dynamic nodes (DN); b) static nodes (SN); and c) internal nodes (IN) that can include nodes with DN and/or SN and/or IN as inputs.

DNs are nodes of the graph that can change. For example, DN can be data sources that update as new data comes in. DN could also be timers that trigger an event based on time intervals. In other examples, DN could also be MySQL monitors, specialized filtering criteria (e.g., update a “where” filter only when a certain event happens). Because these nodes are “sources”, they may occur as root nodes in the DAG. At the most fundamental level, DN are root DAG nodes which change (e.g., are “alive”).

SNs are nodes of the DAG that do not change. For example, historical data does not change. IN are interior nodes of the DAG. The state of an IN can be defined by its inputs, which can be DN, SN, and or IN. If all of the IN inputs are “static”, the IN will be static. If one or more of the IN inputs is “dynamic”, the IN will be dynamic. IN can be tables or other data structures. For example, a “listener IN” can permit code to listen to a node of the DAG. A listener node or associated listener monitoring code can place (or “fire”) additional events (or notifications) into a priority queue of a DAG.

In general, a DAG can be composed of static and/or dynamic subgraphs. Update processing occurs on dynamic subgraphs (because static subgraphs are not changing). Only dynamic nodes are in the DataMonitor loop. For Tables, AMDR messages are used for communication within the DAG.

When query code is executed, the DAG is created or modified. As part of this process, the system records the order in which the DAG nodes were constructed in. This “construction ordering” can be used to determine the order that nodes are processed in the DAG.

For example, consider:

a=db.i( . . . ), where a is a dynamic node (or DN)

b=a.where(“A=1”)

c=b.where(“B=2”)

d=c.join(b)

Assume (a) has changes to be processed during a refresh cycle. The order of processing will be (a), (b), (c), and then (d).

When (d) is processed, it will process input changes from both (b) and (c) before creating AMDRs notification messages for (d). This ordering prevents (d) from creating more than one set of AMDRs per input change, and it can help ensure that all AMDRs are consistent with all data being processed for the clock cycle. If this ordering were not in place, it may be possible to get multiple ticks per cycle and some of the data can be inconsistent. Also, the ordering can help ensure that joins produce consistent results.

FIGS. 5A and 5B show example data source definitions and a corresponding hybrid DAG having a cyclicality feedback provider in accordance with some implementations. In FIG. 5A, example code defines the data sources as tables (t1-t5), where table t1 includes data from the cyclicality feedback provider that provides a feedback path within the hybrid DAG for data from table t3. From the code for the data sources in FIG. 5A, a hybrid DAG can be generated as shown by the graph in FIG. 5B. The hybrid DAG in FIG. 5B shows dependencies between the nodes, which correspond to table data sources, and also shows the cyclicality feedback provider that provides feedback data from table t3 to table t1. It will be appreciated that a cyclicality feedback provider can provide feedback from one or more tables (or other data sources) within an update propagation graph to one or more other tables (or other data sources) at a higher level within the update propagation graph. The cyclicality feedback provider can include a table that is maintained in memory or stored to a storage device such as a disk and/or a non-table data source as mentioned above.

While the cyclicality feedback provider is shown as a separate table for illustration purposes in FIG. 5B, it will be appreciated that in some implementations, the cyclicality feedback provider could be part of another data source (e.g., table). The cyclicality feedback provider can be configured to listen to one or more tables (or other data sources). In addition to listening (or monitoring) for changes in data sources, a cyclicality feedback provider listener can listen for events such as state changes, state updates or other changes in addition to or as an alternative to listening for AMDR-type changes or events.

FIG. 6 is a flowchart of an example cyclicality feedback provider method 600 in accordance with some implementations. Processing begins at 602, where a new cyclicality feedback provider is created. For example, the example pseudo code of FIG. 5A shows example pseudo code lines (e.g., first three lines of FIG. 5A) that create a new cyclicality feedback provider. Creation of a cyclicality feedback provider can include instantiating an object of a class that has been configured to operate as a cyclicality feedback provider object. Processing continues to 604.

At 604, a reference to the cyclicality feedback provider object is obtained. For example, a reference to a table of the cyclicality feedback provider can be requested as illustrated in line four of the pseudo code in FIG. 5A. Processing continues to 606.

At 606, an update propagation graph is constructed. For example, using techniques described above, an update propagation graph (e.g., as represented by FIG. 5B) is constructed. Processing continues to 608.

At 608, the cyclicality feedback provider is configured to listen to changes (e.g., events and/or updates) of the data sources to which the cyclicality feedback provider is associated (e.g., the data sources or tables having data fields that the cyclicality feedback provider is configured to listen for events or changes to occur. Processing continues to 610.

At 610, events and/or updates for the data sources (e.g., tables) within the update propagation graph are processed according to the techniques mentioned above. Processing continues to 612.

At 612, the cyclicality feedback provider listener listens for changes. For example, the listener could programmatically detect a change in one or more data sources or data fields that the cyclicality feedback provider is listening to. Processing continues to 614.

At 614, it is determined whether any changes were detected. If so, processing continues to 616. If no changes were detected, processing continues to 610.

At 616, the cyclicality feedback provider data object (e.g., table) is updated once the events and/or changes for the update propagation graph have been processed for the current logical clock cycle. For example, data changes in data sources monitored by the cyclicality feedback provider can be reflected in the cyclicality feedback provider data object. Processing continues to 618.

At 618, data sources that depend on the cyclicality feedback provider are updated at the end of the current logical cycle so that the feedback data is available for processing during the next logical clock cycle.

It will be appreciated that 602-618 can be repeated in whole or in part to perform a cyclicality feedback operation.

FIG. 7 is a flowchart of an example data source refresh process 700 using a hybrid DAG having a cyclicality feedback provider in accordance with some implementations. Processing begins at 702, where it is determined that a logical clock has transitioned to an update state. The transition to an update state can mark the beginning of a logical clock cycle for updating an update propagation graph as discussed above. Processing continues to 704.

At 704, events and updates that have been queued are processed through the update propagation graph as described above. Processing continues to 706.

At 706, after the events and updates for the update propagation graph have been processed, events and updates from the cyclicality feedback provider are provided to the data sources that are dependent on the cyclicality feedback provider. Processing continues to 708.

At 708, the logical clock transitions to an idle state, which can indicate the end of a current logical clock cycle in advance of a next logical clock cycle.

FIG. 8 is a diagram of an example computer data system 802 and backtesting application 804 using a hybrid DAG having a cyclicality feedback provider 806 in accordance with some implementations.

In operation, the backtesting application 804 can provide predetermined input data 808 to the update propagation graph of the computer data system 802. As the input data 808 is processed across one or more logical clock cycles, the cyclicality feedback provider 806 can provide feedback within the update propagation graph. Output 810 from the computer data system 802 can be received by the backtesting application/system 804 and programmatically evaluated to generate an output signal (e.g., new/modified input data 808, and/or an indication of how the computer data system 802 is performing based on the input data 808).

It will be appreciated that the backtesting application/system 804 can be part of the computer data system 802, or can be a separate system. The cyclicality feedback provider 806 can be part of the backtesting application/system 804 or part of the computer data system 802, or distributed between the two.

FIG. 9 is a diagram of an example machine learning system 906 using a hybrid DAG having a cyclicality feedback provider in accordance with some implementations. In operation, data from data sources 1 and 2 (902 and 904) is provided to the machine learning model or system 906, which processes the input data according to a machine learning technique (e.g., neural network, etc.). The machine learning model or system 906 generates a prediction 908 (or other inference or estimation). A cyclicality feedback provider 910 listens to changes in the machine learning model or system 906 can provides updates to the machine learning model or system 906 as described above.

It will be appreciated that the modules, processes, systems, and sections described above can be implemented in hardware, hardware programmed by software, software instructions stored on a nontransitory computer readable medium or a combination of the above. A system as described above, for example, can include a processor configured to execute a sequence of programmed instructions stored on a nontransitory computer readable medium. For example, the processor can include, but not be limited to, a personal computer or workstation or other such computing system that includes a processor, microprocessor, microcontroller device, or is comprised of control logic including integrated circuits such as, for example, an Application Specific Integrated Circuit (ASIC), a field programmable gate array (FPGA), GPGPU, GPU, or the like. The instructions can be compiled from source code instructions provided in accordance with a programming language such as Java, C, C++, C#.net, assembly or the like. The instructions can also comprise code and data objects provided in accordance with, for example, the Visual Basic™ language, a specialized database query language, or another structured or object-oriented programming language. The sequence of programmed instructions, or programmable logic device configuration software, and data associated therewith can be stored in a nontransitory computer-readable medium such as a computer memory or storage device which may be any suitable memory apparatus, such as, but not limited to ROM, PROM, EEPROM, RAM, flash memory, disk drive and the like.

Furthermore, the modules, processes systems, and sections can be implemented as a single processor or as a distributed processor. Further, it should be appreciated that the steps mentioned above may be performed on a single or distributed processor (single and/or multi-core, or cloud computing system). Also, the processes, system components, modules, and sub-modules described in the various figures of and for embodiments above may be distributed across multiple computers or systems or may be co-located in a single processor or system. Example structural embodiment alternatives suitable for implementing the modules, sections, systems, means, or processes described herein are provided below.

The modules, processors or systems described above can be implemented as a programmed general purpose computer, an electronic device programmed with microcode, a hard-wired analog logic circuit, software stored on a computer-readable medium or signal, an optical computing device, a networked system of electronic and/or optical devices, a special purpose computing device, an integrated circuit device, a semiconductor chip, and/or a software module or object stored on a computer-readable medium or signal, for example.

Embodiments of the method and system (or their sub-components or modules), may be implemented on a general-purpose computer, a special-purpose computer, a programmed microprocessor or microcontroller and peripheral integrated circuit element, an ASIC or other integrated circuit, a digital signal processor, a hardwired electronic or logic circuit such as a discrete element circuit, a programmed logic circuit such as a PLD, PLA, FPGA, PAL, or the like. In general, any processor capable of implementing the functions or steps described herein can be used to implement embodiments of the method, system, or a computer program product (software program stored on a nontransitory computer readable medium).

Furthermore, embodiments of the disclosed method, system, and computer program product (or software instructions stored on a nontransitory computer readable medium) may be readily implemented, fully or partially, in software using, for example, object or object-oriented software development environments that provide portable source code that can be used on a variety of computer platforms. Alternatively, embodiments of the disclosed method, system, and computer program product can be implemented partially or fully in hardware using, for example, standard logic circuits or a VLSI design. Other hardware or software can be used to implement embodiments depending on the speed and/or efficiency requirements of the systems, the particular function, and/or particular software or hardware system, microprocessor, or microcomputer being utilized. Embodiments of the method, system, and computer program product can be implemented in hardware and/or software using any known or later developed systems or structures, devices and/or software by those of ordinary skill in the applicable art from the function description provided herein and with a general basic knowledge of the software engineering and computer networking arts.

Moreover, embodiments of the disclosed method, system, and computer readable media (or computer program product) can be implemented in software executed on a programmed general purpose computer, a special purpose computer, a microprocessor, or the like.

It is, therefore, apparent that there is provided, in accordance with the various embodiments disclosed herein, methods, systems and computer readable media for data source refreshing using an update propagation graph with feedback cyclicality.

Application Ser. No. 15/154,974, entitled “DATA PARTITIONING AND ORDERING” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/154,975, entitled “COMPUTER DATA SYSTEM DATA SOURCE REFRESHING USING AN UPDATE PROPAGATION GRAPH” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/154,979, entitled “COMPUTER DATA SYSTEM POSITION-INDEX MAPPING” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/154,980, entitled “SYSTEM PERFORMANCE LOGGING OF COMPLEX REMOTE QUERY PROCESSOR QUERY OPERATIONS” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/154,983, entitled “DISTRIBUTED AND OPTIMIZED GARBAGE COLLECTION OF REMOTE AND EXPORTED TABLE HANDLE LINKS TO UPDATE PROPAGATION GRAPH NODES” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/154,984, entitled “COMPUTER DATA SYSTEM CURRENT ROW POSITION QUERY LANGUAGE CONSTRUCT AND ARRAY PROCESSING QUERY LANGUAGE CONSTRUCTS” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/154,985, entitled “PARSING AND COMPILING DATA SYSTEM QUERIES” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/154,987, entitled “DYNAMIC FILTER PROCESSING” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/154,988, entitled “DYNAMIC JOIN PROCESSING USING REAL-TIME MERGED NOTIFICATION LISTENER” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/154,990, entitled “DYNAMIC TABLE INDEX MAPPING” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/154,991, entitled “QUERY TASK PROCESSING BASED ON MEMORY ALLOCATION AND PERFORMANCE CRITERIA” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/154,993, entitled “A MEMORY-EFFICIENT COMPUTER SYSTEM FOR DYNAMIC UPDATING OF JOIN PROCESSING” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/154,995, entitled “QUERY DISPATCH AND EXECUTION ARCHITECTURE” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/154,996, entitled “COMPUTER DATA DISTRIBUTION ARCHITECTURE” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/154,997, entitled “DYNAMIC UPDATING OF QUERY RESULT DISPLAYS” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/154,998, entitled “DYNAMIC CODE LOADING” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/154,999, entitled “IMPORTATION, PRESENTATION, AND PERSISTENT STORAGE OF DATA” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/155,001, entitled “COMPUTER DATA DISTRIBUTION ARCHITECTURE” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/155,005, entitled “PERSISTENT QUERY DISPATCH AND EXECUTION ARCHITECTURE” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/155,006, entitled “SINGLE INPUT GRAPHICAL USER INTERFACE CONTROL ELEMENT AND METHOD” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/155,007, entitled “GRAPHICAL USER INTERFACE DISPLAY EFFECTS FOR A COMPUTER DISPLAY SCREEN” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/155,009, entitled “COMPUTER ASSISTED COMPLETION OF HYPERLINK COMMAND SEGMENTS” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/155,010, entitled “HISTORICAL DATA REPLAY UTILIZING A COMPUTER SYSTEM” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/155,011, entitled “DATA STORE ACCESS PERMISSION SYSTEM WITH INTERLEAVED APPLICATION OF DEFERRED ACCESS CONTROL FILTERS” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/155,012, entitled “REMOTE DATA OBJECT PUBLISHING/SUBSCRIBING SYSTEM HAVING A MULTICAST KEY-VALUE PROTOCOL” and filed in the United States Patent and Trademark Office on May 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/351,429, entitled “QUERY TASK PROCESSING BASED ON MEMORY ALLOCATION AND PERFORMANCE CRITERIA” and filed in the United States Patent and Trademark Office on Nov. 14, 2016, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/813,112, entitled “COMPUTER DATA SYSTEM DATA SOURCE REFRESHING USING AN UPDATE PROPAGATION GRAPH HAVING A MERGED JOIN LISTENER” and filed in the United States Patent and Trademark Office on Nov. 14, 2017, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/813,127, entitled “COMPUTER DATA DISTRIBUTION ARCHITECTURE CONNECTING AN UPDATE PROPAGATION GRAPH THROUGH MULTIPLE REMOTE QUERY PROCESSORS” and filed in the United States Patent and Trademark Office on Nov. 14, 2017, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

Application Ser. No. 15/813,119, entitled “KEYED ROW SELECTION” and filed in the United States Patent and Trademark Office on Nov. 14, 2017, is hereby incorporated by reference herein in its entirety as if fully set forth herein.

While the disclosed subject matter has been described in conjunction with a number of embodiments, it is evident that many alternatives, modifications and variations would be, or are, apparent to those of ordinary skill in the applicable arts. Accordingly, Applicants intend to embrace all such alternatives, modifications, equivalents and variations that are within the spirit and scope of the disclosed subject matter.

Claims

1. A system for updating a data object using an update propagation graph having a cyclicality feedback provider, the system comprising: one or more hardware processors coupled to a nontransitory computer readable medium having stored thereon software instructions that, when executed by the one or more processors, cause the one or more processors to perform operations including:constructing a cyclicality feedback provider including a cyclicality feedback provider object including one or more feedback data fields;obtaining a reference to the cyclicality feedback provider object;constructing a computer data system update propagation graph having one or more update propagation graph data fields that correspond to the one or more feedback data fields; andadding a feedback provider listener to the computer data system update propagation graph, wherein the feedback provider listener provides feedback updates to the one or more feedback data fields of the cyclicality feedback provider object when changes to the one or more update propagation graph data fields corresponding to the one or more feedback data fields are detected, and wherein the feedback updates are provided to the one or more feedback data fields of the cyclicality feedback provider object based on a state of a logical clock and on completion of update processing for a given logical clock cycle.

2. The system of claim 1, wherein the cyclicality feedback provider object includes a computer data system table object.

3. The system of claim 1, wherein the update propagation graph includes a hybrid directed acyclic graph having a clock-state controlled cyclicality feedback provided by the cyclicality feedback provider and a state of a logical clock.

4. The system of claim 3, wherein the operations further include: determining that a logical clock has transitioned to an update state;processing events and updates to data sources of the update propagation graph for a current logical clock cycle, wherein processing the events and updates are performed on the hybrid directed acyclic graph as if the cyclicality feedback is not present; andafter the processing events and updates has completed, providing events and updates from the cyclicality feedback provider object to one or more data objects within the update propagation graph, wherein the events and updates from the cyclicality feedback provider object will be processed through the update propagation graph in a next logical clock cycle.

5. The system of claim 4, wherein processing events and updates to the data sources includes: invoking a data source refresh method for a data source for which changes are being processed;determining whether a priority queue for the data source is empty;when the priority queue is not empty, retrieving a next change notification message from the priority queue and delivering the change notification to a corresponding data source and repeating determining whether the priority is queue is empty; andwhen the priority queue is empty, setting the logical clock to an idle state.

6. The system of claim 5, wherein the operations further include: performing a backtesting operation by providing predetermined input data to the update propagation graph as one or more events and updates to one or more data sources; andreceiving output results from the update propagation graph for each logical clock cycle.

7. The system of claim 6, wherein the operations further include: comparing the output results received from the update propagation graph with one or more reference values; andgenerating an output signal based on the comparing.

8. A computer-implemented method for updating a data object using an update propagation graph having a cyclicality feedback provider, the method comprising: constructing a cyclicality feedback provider including a cyclicality feedback provider object including one or more feedback data fields;obtaining a reference to the cyclicality feedback provider object;constructing a computer data system update propagation graph having one or more update propagation graph data fields that correspond to the one or more feedback data fields; andadding a feedback provider listener to the computer data system update propagation graph, wherein the feedback provider listener provides feedback updates to the one or more feedback data fields of the cyclicality feedback provider object when changes to the one or more update propagation graph data fields corresponding to the one or more feedback data fields are detected, and wherein the feedback updates are provided to the one or more feedback data fields of the cyclicality feedback provider object based on a state of a logical clock and on completion of update processing for a given logical clock cycle.

9. The computer-implemented method of claim 8, wherein the cyclicality feedback provider object includes a computer data system table object.

10. The computer-implemented method of claim 8, wherein the update propagation graph includes a hybrid directed acyclic graph having a clock-state controlled cyclicality feedback provided by the cyclicality feedback provider and a state of a logical clock.

11. The computer-implemented method of claim 10, further comprising: determining that a logical clock has transitioned to an update state;processing events and updates to data sources of the update propagation graph for a current logical clock cycle, wherein processing the events and updates are performed on the hybrid directed acyclic graph as if the cyclicality feedback is not present; andafter the processing events and updates has completed, providing events and updates from the cyclicality feedback provider object to one or more data objects within the update propagation graph, wherein the events and updates from the cyclicality feedback provider object will be processed through the update propagation graph in a next logical clock cycle.

12. The computer-implemented method of claim 11, wherein processing events and updates to the data sources includes: invoking a data source refresh method for a data source for which changes are being processed;determining whether a priority queue for the data source is empty;when the priority queue is not empty, retrieving a next change notification message from the priority queue and delivering the change notification to a corresponding data source and repeating determining whether the priority is queue is empty; andwhen the priority queue is empty, setting the logical clock to an idle state.

13. The computer-implemented method of claim 12, further comprising: performing a backtesting operation by providing predetermined input data to the update propagation graph as one or more events and updates to one or more data sources; andreceiving output results from the update propagation graph for each logical clock cycle.

14. The computer-implemented method of claim 13, further comprising: comparing the output results received from the update propagation graph with one or more reference values; andgenerating an output signal based on the comparing.

15. A nontransitory computer readable medium having stored thereon software instructions that, when executed by one or more processors, cause the one or more processors to perform operations including: constructing a cyclicality feedback provider including a cyclicality feedback provider object including one or more feedback data fields;obtaining a reference to the cyclicality feedback provider object;constructing a computer data system update propagation graph having one or more update propagation graph data fields that correspond to the one or more feedback data fields; andadding a feedback provider listener to the computer data system update propagation graph, wherein the feedback provider listener provides feedback updates to the one or more feedback data fields of the cyclicality feedback provider object when changes to the one or more update propagation graph data fields corresponding to the one or more feedback data fields are detected, and wherein the feedback updates are provided to the one or more feedback data fields of the cyclicality feedback provider object based on a state of a logical clock and on completion of update processing for a given logical clock cycle.

16. The nontransitory computer readable medium of claim 15, wherein the cyclicality feedback provider object includes a computer data system table object.

17. The nontransitory computer readable medium of claim 15, wherein the update propagation graph includes a hybrid directed acyclic graph having a clock-state controlled cyclicality feedback provided by the cyclicality feedback provider and a state of a logical clock.

18. The nontransitory computer readable medium of claim 17, wherein the operations further include: determining that a logical clock has transitioned to an update state;processing events and updates to data sources of the update propagation graph for a current logical clock cycle, wherein processing the events and updates are performed on the hybrid directed acyclic graph as if the cyclicality feedback is not present; andafter the processing events and updates has completed, providing events and updates from the cyclicality feedback provider object to one or more data objects within the update propagation graph, wherein the events and updates from the cyclicality feedback provider object will be processed through the update propagation graph in a next logical clock cycle.

19. The nontransitory computer readable medium of claim 18, wherein processing events and updates to the data sources includes: invoking a data source refresh method for a data source for which changes are being processed;determining whether a priority queue for the data source is empty;when the priority queue is not empty, retrieving a next change notification message from the priority queue and delivering the change notification to a corresponding data source and repeating determining whether the priority is queue is empty; andwhen the priority queue is empty, setting the logical clock to an idle state.

20. The nontransitory computer readable medium of claim 19, wherein the operations further include: performing a backtesting operation by providing predetermined input data to the update propagation graph as one or more events and updates to one or more data sources;receiving output results from the update propagation graph for each logical clock cycle;comparing the output results received from the update propagation graph with one or more reference values; andgenerating an output signal based on the comparing.