Optimal persistence of a business process

• US 8,892,557 B2
• Filed: 08/23/2013
• Issued: 11/18/2014
• Est. Priority Date: 11/16/2010
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Nagpal, Abhinay R., Ramanathan, Sri, Ramesh, Sandeep, Sivakumar, Gandhi, Trevathan, Matthew B.
• Primary Examiner(s): Cao, Phuong Thao
• Application Number:
  US13/974,133
• Publication Number:
  US 20130339279A1
• Time in Patent Office: 452 Days
• Field of Search: 707/736, 707/769, 707/999.002, 706/21
• US Class Current: 707/736
• CPC Class Codes: G06F 16/2438 Embedded query languages
• CPC Class Codes: G06F 16/289 Object oriented databases
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G06N 3/084 Backpropagation, e.g. using gradient descent
• CPC Class Codes: G06Q 10/06 Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling

Abstract

Aspects of the invention provide for automatically selecting optimal fetch settings for business processes as a function of database query load and relational context by determining whether data loaded for data retrieval points is dependent upon a query result from another query process and automatically selecting an eager fetch setting if dependent upon a query result from another query process, or a lazy fetch setting if not. Usage of the data retrieval points is monitored with respect to defined units of work to define retrieval patterns and automatically update the fetch settings, including by revising selected eager fetch settings to lazy fetch settings if a data size of a defined retrieval pattern is larger than a permissible memory resource threshold.

Claims

1. A method for automatically selecting optimal fetch settings for each of a plurality of business processes as a function of database query load and relational context, the method comprising:

identifying each of a plurality of data retrieval points enabled during a creation of a business process execution language process;

determining for each of the plurality of data retrieval points via a processor whether data loaded is dependent upon a query result from another query process, wherein determining that the data loaded is dependent upon a query result from another query process is in response to determining that a task corresponding to the respective data retrieval point always requires at least one additional query to acquire all information needed from a database;

the processor automatically selecting an eager fetch setting for each of the plurality of data retrieval points that are determined to be dependent upon a query result from another query process; and

the processor automatically selecting a lazy fetch setting for each of the plurality of data retrieval points that are determined to be not dependent upon a query result from another query process.

2. The method of claim 1, further comprising:

monitoring usage of the business process execution language process data retrieval points with respect to a defined unit of work and defining retrieval patterns for each of the data retrieval points as a function of the monitoring; and

automatically updating the fetch settings as a function of the defined retrieval patterns by revising each of the selected eager fetch settings for the data retrieval points that have defined retrieval patterns that are larger than a permissible memory resource threshold to lazy fetch settings.

3. The method of claim 2, further comprising:

iteratively repeating for each of a plurality of periodic retraining periods the monitoring the usage of the business process execution language process data retrieval points with respect to the defined unit of work, the defining the retrieval patterns for each of the data retrieval points as the function of the monitoring and the automatically updating the fetch settings as the function of the defined retrieval patterns by revising each of the selected eager fetch settings for the data retrieval points that have defined retrieval patterns that are larger than the permissible memory resource threshold to lazy fetch settings.

4. The method of claim 2, further comprising:

predicting that a data size of results from a selected eager fetch setting for one of the data retrieval points would be larger than the specified permissible memory resource threshold; and

overriding an automatically selected eager fetch setting for the one data retrieval point and selecting instead the lazy fetch setting in response to the predicting that the data size of results from the selected eager fetch setting for the one data retrieval point would be larger than the specified permissible memory resource threshold.

5. The method of claim 4, wherein the predicting that the data size of results from a selected eager fetch setting for one of the data retrieval points would be larger than the specified permissible memory resource threshold comprises using a multilayer feed-forward neural network to predict future values based on training sets composed of historical data.

6. The method of claim 5, wherein the using the multilayer feed-forward neural network to predict future values based on training sets composed of historical data comprises:

collecting periodic resource demand statistics and compiling the collected periodic resource demand statistics over a specified historic timeframe;

translating the compiled periodic resource demand statistics into logged traffic load measurements and deriving associated demand requirements into an initial training data set comprising day and time statistics;

training a multilayer feed-forward neural network using the initial training data set;

using the trained multilayer feed-forward neural network to predict frequency patterns for each of the data retrieval points using the compiled periodic resource demand statistics over a previous specified timeframe;

assigning the predicted frequency patterns to each of the data retrieval points;

updating the initial training data set to contain statistics for a current time period; and

periodically retraining the multilayer feed-forward neural network through a back-propagation learning algorithm and a nonlinear sigmoid activation function.

7. The method of claim 5, wherein the defined unit of work is an application session defined from an open method to a close method, the method further comprising:

capturing a thread stack trace at a point when the open method is invoked; and

using the thread stack trace to uniquely identify and match a business process to tuned optimal settings calculated for it in the defined retrieval patterns.

8. The method of claim 5, further comprising:

automatically selecting a fetch setting for each of the plurality of data retrieval points or automatically updating the fetch settings as the function of the defined retrieval patterns by;

selecting a batch fetch setting or a join fetch setting.

9. A computer system for automatically selecting optimal fetch settings for each of a plurality of business processes as a function of database query load and relational context, the computer system comprising:

a processor;

a computer readable memory in circuit communication with the processor; and

a computer readable storage system in circuit communication with the processor and with the computer readable memory;

wherein the processor, when executing program instructions stored on the computer-readable storage system via the computer readable memory;

identifies each of a plurality of data retrieval points enabled during a creation of a business process execution language process;

determines for each of the plurality of data retrieval points whether data loaded is dependent upon a query result from another query process, wherein determining that the data loaded is dependent upon a query result from another query process is in response to determining that a task corresponding to the respective data retrieval point always requires at least one additional query to acquire all information needed from a database;

automatically selects an eager fetch setting for each of the plurality of data retrieval points that are determined to be dependent upon a query result from another query process; and

automatically selects a lazy fetch setting for each of the plurality of data retrieval points that are determined to be not dependent upon a query result from another query process.

10. The computer system of claim 9, wherein the processor, when executing the program instructions stored on the computer-readable storage system via the computer readable memory, further:

monitors usage of the business process execution language process data retrieval points with respect to a defined unit of work and defines retrieval patterns for each of the data retrieval points as a function of the monitoring; and

automatically updates the fetch settings as a function of the defined retrieval patterns by revising each of the selected eager fetch settings for the data retrieval points that have defined retrieval patterns that are larger than a permissible memory resource threshold to lazy fetch settings.

11. The computer system of claim 10, wherein the processor, when executing the program instructions stored on the computer-readable storage system via the computer readable memory, further:

predicts that a data size of results from a selected eager fetch setting for one of the data retrieval points would be larger than the specified permissible memory resource threshold; and

overrides an automatically selected eager fetch setting for the one data retrieval point and selects the lazy fetch setting in response to the prediction that the data size of results from the selected eager fetch setting for the one data retrieval point would be larger than the specified permissible memory resource threshold.

12. The computer system of claim 11, wherein the processor, when executing the program instructions stored on the computer-readable storage system via the computer readable memory, further:

predicts that the data size of results from a selected eager fetch setting for one of the data retrieval points would be larger than the specified permissible memory resource threshold through using a multilayer feed-forward neural network to predict future values based on training sets composed of historical data by;

collecting periodic resource demand statistics and compiling the collected periodic resource demand statistics over a specified historic timeframe;

translating the compiled periodic resource demand statistics into logged traffic load measurements and deriving associated demand requirements into an initial training data set comprising day and time statistics;

training a multilayer feed-forward neural network using the initial training data set;

using the trained multilayer feed-forward neural network to predict frequency patterns for each of the data retrieval points using the compiled periodic resource demand statistics over a previous specified timeframe;

assigning the predicted frequency patterns to each of the data retrieval points;

updating the initial training data set to contain statistics for a current time period; and

periodically retraining the multilayer feed-forward neural network through a back-propagation learning algorithm and a nonlinear sigmoid activation function.

13. The computer system of claim 12, wherein the defined unit of work is an application session defined from an open method to a close method;

and

wherein the processor, when executing the program instructions stored on the computer-readable storage system via the computer readable memory, further captures a thread stack trace at a point when the open method is invoked and uses the thread stack trace to uniquely identify and match a business process to tuned optimal settings calculated for it in the defined retrieval patterns.

14. The computer system of claim 12, wherein the processor, when executing the program instructions stored on the computer-readable storage system via the computer readable memory, further automatically selects a fetch setting for each of the plurality of data retrieval points, or automatically updates the fetch settings as the function of the defined retrieval patterns, by selecting a batch fetch setting or a join fetch setting.

15. A computer program product for automatically selecting optimal fetch settings for each of a plurality of business processes as a function of database query load and relational context, the computer program product comprising:

a computer readable storage device having computer readable program code embodied therewith, the computer readable program code comprising instructions that, when executed by a processor, cause the processor to;

identify each of a plurality of data retrieval points enabled during a creation of a business process execution language process;

determine for each of the plurality of data retrieval points whether data loaded is dependent upon a query result from another query process, wherein determining that the data loaded is dependent upon a query result from another query process is in response to determining that a task corresponding to the respective data retrieval point always requires at least one additional query to acquire all information needed from a database;

automatically select an eager fetch setting for each of the plurality of data retrieval points that are determined to be dependent upon a query result from another query process; and

automatically select a lazy fetch setting for each of the plurality of data retrieval points that are determined to be not dependent upon a query result from another query process.

16. The computer program product of claim 15, wherein the computer readable program code instructions, when executed by the processor, further cause the processor to:

monitor usage of the business process execution language process data retrieval points with respect to a defined unit of work and define retrieval patterns for each of the data retrieval points as a function of the monitoring; and

automatically update the fetch settings as a function of the defined retrieval patterns by revising each of the selected eager fetch settings for the data retrieval points that have defined retrieval patterns that are larger than a permissible memory resource threshold to lazy fetch settings.

17. The computer program product of claim 16, wherein the computer readable program code instructions, when executed by the processor, further cause the processor to:

predict that a data size of results from a selected eager fetch setting for one of the data retrieval points would be larger than the specified permissible memory resource threshold; and

override an automatically selected eager fetch setting for the one data retrieval point and select the lazy fetch setting in response to the prediction that the data size of results from the selected eager fetch setting for the one data retrieval point would be larger than the specified permissible memory resource threshold; and

iteratively repeat for each of a plurality of periodic retraining periods the monitoring the usage of the business process execution language process data retrieval points with respect to the defined unit of work, the defining the retrieval patterns for each of the data retrieval points as the function of the monitoring and the automatically updating of the fetch settings as the function of the defined retrieval patterns by revising each of the selected eager fetch settings for the data retrieval points that have defined retrieval patterns that are larger than the permissible memory resource threshold to lazy fetch settings.

18. The computer program product of claim 17, wherein the computer readable program code instructions, when executed by the processor, further cause the processor to:

predict that the data size of results from a selected eager fetch setting for one of the data retrieval points would be larger than the specified permissible memory resource threshold through using a multilayer feed-forward neural network to predict future values based on training sets composed of historical data by;

collecting periodic resource demand statistics and compiling the collected periodic resource demand statistics over a specified historic timeframe;

translating the compiled periodic resource demand statistics into logged traffic load measurements and deriving associated demand requirements into an initial training data set comprising day and time statistics;

training a multilayer feed-forward neural network using the initial training data set;

using the trained multilayer feed-forward neural network to predict frequency patterns for each of the data retrieval points using the compiled periodic resource demand statistics over a previous specified timeframe;

assigning the predicted frequency patterns to each of the data retrieval points;

updating the initial training data set to contain statistics for a current time period; and

periodically retraining the multilayer feed-forward neural network through a back-propagation learning algorithm and a nonlinear sigmoid activation function.

19. The computer program product of claim 18, wherein the defined unit of work is an application session defined from an open method to a close method;

and wherein the computer readable program code instructions, when executed by the processor, further cause the processor to capture a thread stack trace at a point when the open method is invoked and use the thread stack trace to uniquely identify and match a business process to tuned optimal settings calculated for it in the defined retrieval patterns.