Method and system for apportioning changes in metric variables in an symmetric multiprocessor (SMP) environment

• US 6,904,594 B1
• Filed: 07/06/2000
• Issued: 06/07/2005
• Est. Priority Date: 07/06/2000
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Berry, Robert Francis, Howard, John Day, Levine, Frank Eliot, Urquhart, Robert J.
• Primary Examiner(s): An, Meng-Al T.
• Assistant Examiner(s): Vo, Lilian
• Application Number:
  US09/612,340
• Time in Patent Office: 1,797 Days
• Field of Search: 718/102, 718/100, 718/107, 718/108, 712/228, 717/145, 717/100, 717/130, 717/158
• US Class Current: 718/100
• CPC Class Codes: G06F 11/3404 for parallel or distributed programming
• CPC Class Codes: G06F 11/3466 Performance evaluation by tracing or monitoring
• CPC Class Codes: G06F 2201/86 Event-based monitoring
• CPC Class Codes: G06F 2201/865 Monitoring of software
• CPC Class Codes: G06F 2201/88 Monitoring involving counting

Abstract

A method and system for monitoring performance of a program using global metric variables to provide the support in an symmetric multiprocessor (SMP) system. A Java virtual machine (Jvm) either calls the profiler whenever bytes are allocated or provides an interface to allow the profiler to determine the value of the change in the metric for the current thread. The profiler then applies the changes to a metric for the current thread. Alternatively, per processor data areas are maintained for storing per processor metric values. Whenever a thread switch occurs or there is a request for the metric on a specified thread, an operating system kernel updates the thread level metric values with changes in the values per processor metrics.

Claims

1. A method for monitoring performance of a program being executed using symmetric multiprocessing (SMP) functionality comprising:

executing a native code routine;

executing a first thread of the native code routine on a first symmetric multiprocessing (SMP) processor;

ascertaining first profile information, wherein the first profile information is processor profile information for the first SMP processor that identifies metrics associated with the execution of the first thread on the first SMP processor and is maintained by an operating system kernel;

updating first thread profile information, maintained by a profiler application, with the first profile information;

executing the first thread of the native code routine on a second SMP processor;

ascertaining second profile information, wherein the second profile information is processor profile information for the second SMP processor that identifies metrics associated with the execution of the first thread on the second SMP processor and is maintained by the operating system kernels; and

updating the first thread profile information, maintained by the profiler application, with the second profile information.

2. The method recited in claim 1 above, further comprises:

executing a second thread of the native code routine on the first SMP processor;

ascertaining third profile information; and

updating second thread profile information with the third profile information.

3. The method recited in claim 1 above, wherein the first profile information releases to the execution of the first thread on a virtual machine.

4. The method recited in claim 1 above, wherein ascertaining first profile information further comprises:

retrieving processor accumulated profile information;

retrieving processor last profile information; and

calculating the first profile information by ascertaining the processing accumulated profile information and the processor last profile information.

5. The method recited in claim 4 above further comprises:

resetting the processor last profile information by replacing the processor last profile information with the processor accumulated profile information.

6. The method of claim 4, wherein the processor accumulated profile information is a total accumulated amount of a particular metric for a processor since initialization of monitoring the performance of the program, and wherein the processor last accumulated profile information is an accumulated amount of a particular metric for a processor at a last time that the first thread profile information was updated.

7. The method recited in claim 1 above, wherein ascertaining first profile information further comprises:

accessing a first processor data area containing first processor accumulated profile information; and

updating the first processor accumulated profile information with virtual machine profile information, wherein the first profile information is calculated from the first processor accumulated profile information.

8. The method recited in claim 7 above, wherein updating the first processor accumulated profile information with virtual machine profile information is performed by a virtual machine.

9. The method recited in claim 7 above, wherein updating the first processor accumulated profile information with virtual machine profile information is performed by a virtual machine further comprises:

receiving a request to update the first processor accumulated profile information with virtual machine profile information;

accessing a first processor data area containing first processor accumulated profile information; and

updating the first processor accumulated profile information with virtual machine profile information, wherein updating the first processor accumulated profile information with virtual machine profile information is performed by an operating system kernel.

10. The method recited in claim 1 above, wherein ascertaining first profile information further comprises:

receiving virtual machine profile information;

updating processor accumulated profile information with virtual machine profile information;

retrieving processor last profile information; and

calculating the first profile information by comparing the processor accumulated profile information and the processor last profile information.

11. The method recited in claim 7 above, wherein the first processor data area is a predefined data area.

12. The method recited in claim 7 above, wherein the first processor data area is allocated for first processor accumulated profile information at initialization.

13. The method recited in claim 1 above, wherein the first processor accumulated profile comprises one of allocation bytes, allocation objects, time, live object and live bytes.

14. The method of claim 1, wherein the processor profile information for the first SMP processor and the processor profile information for the second SMP processor are stored in per-processor data areas for each metric being tracked for each processor, and wherein the per-processor data areas are allocated by the operating system kernel at initialization of monitoring the performance of the program being executed.

15. A method for monitoring performance of a program being executed by a symmetric multiprocessing (SMP) functionality comprising:

launching a current thread of a native code routine on a first symmetric multiprocessing (SMP) processor;

stopping a last thread on the first SMP processor;

accessing a first SMP processor data area;

retrieving first SMP processor accumulated profile information;

retrieving first SMP processor last accumulated profile information;

ascertaining first profile information from the first SMP processor accumulated profile information and the first SMP processor last accumulated profile information;

updating first thread profile information with the first profile information;

setting first SMP processor last accumulated profile information in the first SMP processor data area equal to the first SMP processor accumulated profile information; and

starting the first thread on the first SMP processor.

16. The method of claim 15, wherein the first SMP processor accumulated profile information is a total accumulated amount of a particular metric for the first SMP processor since initialization of monitoring the performance of the program, and wherein the first SMP processor last accumulated profile information is an accumulated amount of a particular metric for the first SMP processor at a last time that the first thread profile information was updated.

17. A method for monitoring performance of a program being executed by a symmetric multiprocessing (SMP) functionality comprising:

receiving a request for current thread profile information;

retrieving SMP processor accumulated profile information from each SMP processor'"'"'s data area;

retrieving SMP processor last accumulated profile information from each processor'"'"'s data area;

ascertaining profile information from the SMP processor accumulated profile information and the SMP processor last accumulated profile information for each thread running on a processor;

updating thread profile information with the profile information for each thread running on a processor;

setting each SMP processor last accumulated profile information in a respective SMP processor'"'"'s data area equal to the SMP processor accumulated profile information for that SMP processor; and

transferring the thread profile information to the requester.

18. The method of claim 17, wherein the SMP processor accumulated profile information is a total accumulated amount of a particular metric for a corresponding SMP processor since initialization of monitoring the performance of the program, and wherein the SMP processor last accumulated profile information is an accumulated amount of a particular metric for a corresponding SMP processor at a last time that the thread profile information was updated.

19. A data processing system for monitoring performance of a program being executed by symmetric multiprocessing (SMP) functionality comprising:

executing means for executing a native code routine;

executing means for executing a first thread of the native code routine on a first symmetric multiprocessing (SMP) processor;

ascertaining means for ascertaining the profile information, wherein the first profile information is processor profile information for the first SMP processor that, identifies metrics associated with the execution of the first thread on the first SMP processor and is maintained by an operating system kernel;

updating means for updating first thread profile information, maintained by a profiler application, with the first profile information;

executing means for executing the first thread of the native code routine on a second SMP processor;

ascertaining means for ascertaining second profile information, wherein the second profile information is processor profile information for the second SMP processor that identifies metrics associated with the execution of the first thread on the second SMP processor and is maintained by the operating system kernel; and

updating means for updating first thread profile information, maintained by the profiler application, with the second profile information.

20. The system recited in claim 19 above, further comprises:

executing means for executing a second thread of the native code routine on the first SMP processor;

ascertaining means for ascertaining third profile information; and

updating means for updating second thread profile information with the third profile information.

21. The system recited in claim 19 above, wherein the first profile information relates to the execution of the first thread on a virtual machine.

22. The system recited in claim 19 above, wherein the ascertaining means for ascertaining first profile information further comprises:

retrieving means for retrieving processor accumulated profile information;

retrieving means for retrieving processor last profile information; and

calculating means for calculating the first profile information by comparing the processor accumulated profile information and the processor last profile information.

23. The system recited in claim 22 above further comprises:

resetting means for resetting the processor last profile information by replacing the last processor profile information with the processor accumulated profile information.

24. The system recited in claim 19 above, wherein the ascertaining means for ascertaining first profile information further comprises:

accessing means for accessing a first processor data area containing first processor accumulated profile information; and

updating means for updating the first processor accumulated profile information with virtual machine profile information, wherein the first profile information is calculated from the first processor accumulated profile information.

25. The system recited in claim 24 above, wherein the updating means for updating the first processor accumulated profile information with virtual machine profile information is performed by a virtual machine.

26. The system recited in claim 24 above, wherein the updating means for updating the first processor accumulated profile information with virtual machine profile information is performed by a virtual machine further comprises:

receiving means for receiving a request to update the first processor accumulated profile information with virtual machine profile information;

accessing means for accessing a first processor data area containing first processor accumulated profile information; and

updating means for updating the first processor accumulated profile information with virtual machine profile information, wherein updating the first processor accumulated profile information with virtual machine profile information is performed by an operating system kernel.

27. The system recited in claim 24 above, wherein the first processor data area is a predefined data area.

28. The system recited in claim 19 above, wherein the ascertaining means for ascertaining first profile information further comprises:

receiving means for receiving virtual machine profile information;

updating means for updating processor accumulated profile information with virtual machine profile information;

retrieving means for retrieving processor last profile information; and

calculating means for calculating the first profile information by comparing the processor accumulated profile information and the processor last profile information.

29. The system recited in claim 24 above, wherein the first processor data area is allocated for first processor accumulated profile information at initialization.

30. The system recited in claim 19 above, wherein the first processor accumulated profile comprises one of allocation bytes, allocation objects, time, live object and live bytes.

31. A computer program product for monitoring performance of a program being executed by a symmetric multiprocessing (SMP) functionality comprising:

executing instructions for executing a native code routine;

executing instruction for executing a first thread of the native code routine on a first symmetric multiprocessing (SMP) processor;

ascertaining instructions for ascertaining first profile information, wherein the first profile information is processor profile information for the first SMP processor that identifies metrics associated with the execution of the first thread on the first SMP processor and is maintained by an operating system kernel;

updating instructions for updating first thread profile information, maintained by a profiler application, with the first profile information;

executing instructions for executing the first thread of the native code routine on a second SMP processor;

ascertaining instructions for ascertaining second profile information, wherein the second profile information is processor profile information for the second SMP processor that identifies metrics associated with the execution of the first thread on the second SMP processor and is maintained by the operating system kernel; and

updating instructions for updating first thread profile information, maintained by the profiler application, with the second profile information.

32. The computer program product recited in claim 31 above, further comprises:

executing means for executing a second thread of the native code routine on the first SMP processor;

ascertaining means for ascertaining third profile information; and

updating means for updating second thread profile information with the third profile information.

33. The computer program product recited in claim 31 above, wherein the first profile information relates to the execution of the first threaded on a virtual machine.

34. The computer program product recited in claim 31 above, wherein the ascertaining instructions for ascertaining first profile information further comprises:

retrieving instructions for retrieving processor accumulated profile information;

retrieving instructions for retrieving processor last profile information; and

calculating instructions for calculating the first profile information by comparing the processor accumulated profile information and the processor last profile information.

35. The computer program product recited in claim 34 above further comprises:

resetting instructions for resetting the processor last profile information by replacing the last processor profile information with the processor accumulated profile information.

36. The computer program product recited in claim 31 above, wherein the ascertaining instructions for ascertaining first profile information further comprises:

accessing instructions for accessing a first processor data area containing first processor accumulated profile information; and

updating instructions for updating the first processor accumulated profile information with virtual machine profile information, wherein the first profile information is calculated from the first processor accumulated profile information.

37. The computer program product recited in claim 36 above, wherein the updating instructions for updating the first processor accumulated profile information with virtual machine profile information is performed by a virtual machine.

38. The computer program product recited in claim 36 above, wherein the updating instructions for updating the first processor accumulated profile information with virtual machine profile information is performed by a virtual machine further comprises:

receiving instructions for retrieving for receiving a request to update the first processor accumulated profile information with virtual machine profile information;

accessing instructions for accessing a first processor data area containing first processor accumulated profile instructions; and

updating instructions for updating the first processor accumulated profile information with virtual machine profile information, wherein updating the first processing accumulated profile information with virtual machine profile information is performed by an operating system kernel.

39. The computer program product recited in claim 36 above, wherein the first processor data area is a predefined data area.

40. The computer program product recited in claim 36 above, wherein the first processor data area is allocated for first processor profile information at initialization.

41. The computer program product recited in claim 31 above, wherein the ascertaining instructions for ascertaining first profile information further comprises:

retrieving instructions for receiving virtual machine profile information;

updating instructions for updating processor accumulated profile information with virtual machine profile information;

retrieving instructions for retrieving processor last profile information; and

calculating instructions for calculating the first profile information by comparing the processor accumulated profile information and the processor last profile information.

42. The computer program product recited in claim 31 above, wherein the first processor accumulated profile comprises one of allocation bytes, allocation objects, time, live object and live bytes.

43. A method of monitoring performance of a program being executed by a symmetric multiprocessing (SMP) system, comprising:

using an operating system kernel to maintain a first set of metrics associated with execution of a thread on a plurality of processors in the SMP system, wherein per-processor data areas are allocated by the operating system kernel, and associated with corresponding processors in the plurality of processors, to store metrics for the corresponding processors, and wherein the first set of metrics are at a processor-level;

using a profiler application to maintain a second set of metrics associated with threads of execution of the program, wherein the second set of metrics are at a thread level;

updating the second set of metrics based on the first set of metrics in response to a request from a requestor to update the second set of metrics to thereby generate an update second set of metrics; and

providing the updated second set of metrics to the requestor.