Low latency architecture with directory service for integration of transactional data system with analytical data structures

Patent metadata

Field	Value
Publication number	US-2016103702-A1
Application number	US-201414512240-A
Country	US
Kind code	A1
Filing date	Oct 10, 2014
Priority date	Oct 10, 2014
Publication date	Apr 14, 2016
Grant date	—

Abstract

Low latency communication between a transactional system and analytic data store resources can be accomplished through a low latency key-value store with purpose-designed queues and status reporting channels. Posting by the transactional system to input queues and complementary posting by analytic system workers to output queues is described. On-demand production and splitting of analytic data stores requires significant elapsed processing time, so a separate process status reporting channel is described to which workers can periodically post their progress, thereby avoiding progress inquiries and interruptions of processing to generate report status. This arrangement produces low latency and reduced overhead for interactions between the transactional system and the analytic data store system.

First claim

What is claimed is:

1 . A method of low latency tasking and task monitoring between a transaction processing system and an analytics processing system, the method including:
a transaction processing system generating an analytic data store creation task request that specifies creation of an analytic data store based on at least one data set stored by at least one transactional data management system; queuing the task request to a named key-value task start queue; a worker thread picking up the task request from the named key-value task start queue; the worker thread reporting progress on the task request to a monitoring data structure independent of the task start queue; the worker thread registering a completed analytic data store with the transaction processing system; and upon completion of creating the analytic data store specified by the task request, the worker thread queuing a task complete report to a named key-value task complete queue complementary to the named key-value start queue.

2 . The method of claim 1 , further including the transaction processing system selecting one of more than a hundred named key-value task start queues to which to queue the task request.

3 . The method of claim 1 , further including the transaction processing system selecting one of a multiplicity of named key-value task start queues to which to queue the task request, wherein the task start queues are between 5 and 50 times as numerous as worker threads.

4 . The method of claim 1 , further including the transaction processing system selecting a named key-value task start queue based at least in part on affinity to an entity that owns the data set stored on the transactional data management system.

5 . The method of claim 1 , wherein failure of the worker thread to complete the task request within a predetermined time causes generating and queuing a second analytic data store creation task request that repeats the task request that the worker thread did not complete.

6 . The method of claim 1 , wherein the worker thread, prior to picking up the task request, resumes operation from an error state, further including:
the worker thread querying a directory service to determine one or more additional worker threads in an affinity group with the worker thread that possess current data for the affinity group; and the worker thread obtaining from the additional worker threads the current data for the affinity group.

7 . The method of claim 1 , further including the worker thread setting the named key-value task start queue to blocked status for pulling additional task requests from the named key-value task start queue while processing the task request.

8 . The method of claim 7 , wherein the worker thread, after picking up the task request, enters an error state, further including:
a queue processing system detecting a time out condition following passage of a predetermined period following the worker thread picking up the task request; the queue processing system clearing the blocked status from the named key-value task start queue; and responsive to the detection of the time out condition, generating and queuing a second analytic data store creation task request that repeats the task request that the worker thread did not complete.

9 . The method of claim 1 , further including running a queuing process that manages the task start queue and the task complete queue and storing data for both the task start queue and the task complete queue in a volatile memory instead of rotating or non-volatile memory.

10 . The method of claim 1 , further including:
running in a volatile memory a queuing process that manages the task start queue and the task complete queue; and storing data for both the task start queue and the task complete queue in volatile memory without redundant storage in persistent memory.

11 . A method of low latency tasking and task monitoring system, the method including:
an analytic data store creation task request maker running on hardware that generates creation task requests and specifies creation of an analytic data store based on at least one data set stored by at least one transactional data management system; queuing the task request to a named key-value task start queue; a worker thread picking up the task request from the named key-value task start queue; the worker thread reporting progress on the task request to a monitoring data structure independent of the task start queue; the worker thread registering a completed analytic data store with the transaction processing system; and upon completion of creating the analytic data store specified by the task request, the worker thread queuing a task complete report to a named key-value task complete queue complementary to the named key-value start queue.

12 . A method of low latency query tasking and query processing monitoring between a transaction processing system and an analytics processing system, the method including:
a transaction processing system generating a query task request that specifies querying a read-only analytic data store that stores a data set retrieved from at least one transactional data management system; queuing the task request to a named key-value task start queue; a worker thread picking up the task request from the named key-value task start queue; and upon completion of assembling query results specified by the task request, the worker thread queuing a task complete report to a named key-value task complete queue complementary to the named key-value start queue and reporting the assembled query results.

13 . The method of claim 12 , further including the transaction processing system selecting a named key-value task start queues based at least in part on affinity to an entity that owns the data set stored on the transactional data management system.

14 . The method of claim 12 , further including the transaction processing system selecting one of more than a hundred named key-value task start queues to which to queue the task request.

15 . The method of claim 12 , further including the transaction processing system selecting one of a multiplicity of named key-value task start queues to which to queue the task request, wherein the task start queues are between 5 and 50 times as numerous as worker threads.

16 . The method of claim 12 , wherein failure of the worker thread to complete the task request within a predetermined time causes generating and queuing a second analytic data store creation task request that repeats the task request that the worker thread did not complete.

17 . The method of claim 12 , wherein the worker thread, prior to picking up the task request, resumes operation from an error state, further including:
the worker thread querying a directory service to determine one or more additional worker threads in an affinity group with the worker thread that possess current data for the affinity group; and the worker thread obtaining from the additional worker threads the current data for the affinity group.

18 . The method of claim 12 , further including the worker thread setting the named key-value task start queue to blocked status while processing the task request.

19 . The method of claim 18 , wherein the worker thread, after picking up the task request, enters an error state, further including:
a queue processing system detecting a time out condition following passage of a predetermined period following the worker thread picking up the task request; the queue processing syste

Assignees

• Salesforce Com Inc

Inventors

• Schneider Donovan
• Im Fred
• Silver Daniel C
• Chakravarthy Vijayasarathy

Classifications

• G06F9/466 Primary

  Transaction processing · CPC title

• G06F9/4881

  Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues · CPC title

• G06F16/25 Primary

  Integrating or interfacing systems involving database management systems · CPC title