Autonomic workload classification using predictive assertion for wait queue and thread pool selection

• US 7,703,101 B2
• Filed: 02/13/2004
• Issued: 04/20/2010
• Est. Priority Date: 02/13/2004
• Status: Active Grant

• Current Assignee: X Corp. (f/k/a Twitter, Inc.) (X Holdings Corp.)
• Original Assignee: International Business Machines Corporation
• Inventors: Daughtrey, Erik A., Cuomo, Gennaro A.
• Primary Examiner(s): An; Meng-Ai
• Assistant Examiner(s): To; Jennifer N
• Application Number:
  US10/778,584
• Publication Number:
  US 20050183084A1
• Time in Patent Office: 2,258 Days
• Field of Search: 718100-106, 709201-203, 709223-226, 717127-131
• US Class Current: 718/104
• CPC Class Codes: G06F 9/505 considering the load

Abstract

Incoming work units (e.g., requests) in a computing workload are analyzed and classified according to predicted execution. Preferred embodiments track which instrumented wait points are encountered by the executing work units, and this information is analyzed to dynamically and autonomically create one or more recognizers to programmatically recognize similar, subsequently-received work units. When a work unit is recognized, its execution behavior is then predicted. Execution resources are then allocated to the work units in view of these predictions. The recognizers may be autonomically evaluated or tuned, thereby adjusting to changing workload characteristics. The disclosed techniques may be used advantageously in application servers, message-processing software, and so forth.

Claims

1. A computer program product for programmatically analyzing execution behavior of work units, the computer program product comprising at least one computer-usable storage media having computer-usable program code embodied therein for:

recording, for each of a plurality of work units, information regarding each of at least one instrumented wait point encountered in code executed by that one of the work units;

determining, for each of the plurality of work units from the recorded information, a set comprising the wait points encountered by that one of the work units;

determining all unique ones of the determined sets of wait points;

determining characteristics of the work units that encountered the wait points in each of the unique ones of the wait point sets;

using the determined characteristics to programmatically build one or more recognizers which are adapted to recognizing those characteristics in arriving work units that are to be subsequently executed;

executing the recognizers to programmatically recognize the characteristics in the arriving work units; and

using the programmatically-recognized characteristics when allocating execution resources to the arriving work units.

2. The computer program product according to claim 1, wherein the allocated execution resources are threads in thread pools.

3. The computer program product according to claim 1, wherein the work units are execution requests.

4. The computer program product according to claim 1, wherein the recorded information further comprises time spent waiting at each encountered wait point.

5. A method of programmatically analyzing execution behavior of work units using computer-readable code executed by a computer, comprising:

recording, for each of a plurality of work units, information regarding each of at least one instrumented wait point encountered in code executed by that one of the work units;

determining, for each of the plurality of work units from the recorded information, a set comprising the wait points encountered by that one of the work units;

determining all unique ones of the determined sets of wait points;

determining characteristics of the work units that encountered the wait points in each of the unique ones of the wait point sets;

using the determined characteristics to programmatically build one or more recognizers which are adapted to recognizing those characteristics in arriving work units that are to be subsequently executed; and

using the programmatically-built recognizers when allocating execution resources to the arriving work units.

6. The method according to claim 5, wherein the work units are execution requests.

7. The method according to claim 6, wherein the execution requests are content requests.

8. The method according to claim 6, wherein the execution requests are Web services requests.

9. The method according to claim 6, wherein the execution requests specify method/class names to be executed.

10. The method according to claim 5, wherein the recorded information comprises an identifier of each encountered wait point.

11. The method according to claim 10, wherein the recorded information further comprises time spent waiting at each encountered wait point.

12. The method according to claim 5, wherein the characteristics comprise parameter names and values thereof.

13. The method according to claim 5, wherein determining the characteristics further comprises evaluating occurrence of the characteristics among the unique ones of the wait point sets.

14. The method according to claim 5, wherein determining the characteristics further comprises determining which of the characteristics are discriminators as to the wait points that are encountered.

15. The method according to claim 5, wherein using the programmatically-built recognizers further comprises executing the recognizers to recognize the characteristics in the arriving work units in order to allocate the execution resources to the arriving work units.

16. The method according to claim 5, wherein using the pro grammatically-built recognizers further comprises classifying the arriving work units using the recognizers.

17. The method according to claim 5, wherein the allocated execution resources are threads in thread pools.

18. The method according to claim 5, further comprising programmatically evaluating effectiveness of one or more of the recognizers, in view of its ability to predict which of the instrumented wait points will be encountered in code executed by the arriving work units.

19. The method according to claim 18, further comprising programmatically destroying ones of the programmatically-evaluated recognizers for which the effectiveness falls below a threshold value.

20. A system for programmatically analyzing execution behavior of work units, comprising:

a computer comprising at least one processor; and

instructions which execute using at least one of the at least one processor to implement functions comprising;

recording, for each of a plurality of work units, information regarding at each of least one instrumented wait point encountered in code executed by that one of the work units;

determining, for each of the plurality of work units from the recorded information, a set comprising the wait points encountered by that one of the work units;

determining all unique ones of the determined sets of wait points;

determining characteristics of the work units that encountered the wait points in each of the unique ones of the wait point sets;

using the determined characteristics to programmatically build one or more recognizers which are adapted to recognizing those characteristics in arriving work units that are to be subsequently executed;

executing the recognizers to programmatically recognize the characteristics in the arriving work units; and

using the programmatically-recognized characteristics when allocating execution resources to the arriving work units.

21. The system according to claim 20, wherein the allocated execution resources are threads in thread pools.

22. The system according to claim 20, wherein the work units are execution requests.

23. The system according to claim 20, wherein the recorded information further comprises time spent waiting at each encountered wait point.

24. A method of analyzing execution information for work units using computer-readable code executed by a computer, comprising:

determining, from information that has been recorded for each of a plurality of work units, a set comprising each of at least one instrumented wait point encountered in code executed by that one of the work units;

determining all unique ones of the determined sets of wait points;

determining characteristics of the work units that encountered the wait points in each of the unique ones of the wait point sets;

determining which of the characteristics are discriminators as to the wait points that are encountered in each of the unique ones of the wait point sets;

programmatically building one or more recognizers which are adapted to recognizing, in arriving work units that are to be executed, the characteristics which are determined to be discriminators;

executing the recognizers to recognize, in the arriving work units, the characteristics which are determined to be discriminators; and

allocating execution resources to each of the arriving work units according to which of the recognizers recognizes, for that one of the arriving work units, the characteristics which are determined to be discriminators.

25. The method according to claim 24, wherein the recorded information further comprises time spent waiting at each encountered wait point.

26. The method according to claim 24, wherein the work units are execution requests.

27. The method according to claim 24, wherein the allocated execution resources are threads in thread pools.