MANAGING A SET OF WEAR-LEVELING DATA USING A SET OF THREAD EVENTS

• US 20170199798A1
• Filed: 01/13/2016
• Published: 07/13/2017
• Est. Priority Date: 01/13/2016
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Jain, Chethan, Ward, Maria R.
• Granted Patent: US 10,095,597 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 11/008 Reliability or availability analysis
• CPC Class Codes: G06F 11/30 Monitoring
• CPC Class Codes: G06F 11/3024 where the computing system component is a central processing unit [CPU]
• CPC Class Codes: G06F 2009/4557 Distribution of virtual machine instances; Migration and load balancing
• CPC Class Codes: G06F 3/0616 in relation to life time, e.g. increasing Mean Time Between Failures [MTBF]
• CPC Class Codes: G06F 9/505 considering the load
• CPC Class Codes: G06F 9/5094 where the allocation takes into account power or heat criteria power management in computers in general G06F1/3203; thermal management in computers in general G06F1/206

Abstract

Disclosed aspects include managing a set of wear-leveling data with respect to a set of physical cores of a set of compute nodes. A set of physical cores of the set of compute nodes may be monitored using a set of processor utilization resource registers (PURRs) to identify the set of wear-leveling data. By monitoring the set of physical cores of the set of compute nodes, a set of thread events with respect to the set of physical cores of the set of compute nodes may be detected. Based on the set of thread events, the set of wear-leveling data may be determined. The set of wear-leveling data may then be established in a data store. The wear leveling data may be used to manage asset placement with respect to a shared pool of configurable computing resources.

Claims

1. A computer-implemented method for managing a set of wear-leveling data with respect to a set of physical cores of a set of compute nodes in a simultaneous multi-threading environment, the method comprising:

monitoring, in the simultaneous multi-threading environment using a set of processor utilization resource registers (PURRs), the set of physical cores of the set of compute nodes;

detecting, by monitoring the set of physical cores of the set of compute nodes using the set of PURRs, a set of thread events with respect to the set of physical cores of the set of compute nodes;

collecting a set of thread event data which indicates a historical utilization of the set of physical cores based on the set of thread events, wherein the set of thread event data includes a number of jobs assigned to the set of physical cores;

determining, based on analyzing the set of thread event data related to a frequency at which jobs are assigned to one or more hardware threads, the set of wear-leveling data, wherein the frequency indicates jobs per unit time;

determining, by a hypervisorcomputing resource management system based on the set of wear-leveling data, a placement arrangement for a set of assets with respect to the simultaneous multi-threading environment; and

placing, in the simultaneous multi-threading environment by the computing resource management system based on the placement arrangement, the set of assets.

6. The method of claim 1, further comprising:

receiving a workload;

ascertaining, based on the set of thread event data which indicates the historical utilization of the set of physical cores based on the set of thread events, that a first physical core has a lesser likelihood of a physical core error event in comparison to a second physical core; and

selecting, based on the set of thread event data in response to the first physical core having the lesser likelihood of a physical core error event in comparison to the second physical core and by comparing a first number of jobs of a first set of logical processors of the first physical core with a second number of jobs of a second set of logical processors of the second physical core, the first physical core of the set of physical cores to process the workload.

8. The method of claim 1, wherein detecting, by monitoring the set of physical cores of the set of compute nodes using the set of PURRs, the set of thread events with respect to the set of physical cores of the set of compute nodes includes:

scheduling, by a hypervisor in the computing resource management system, a job for the set of physical cores of the set of compute nodes;

tracking, using the set of PURRs, a count of j obs scheduled on a given hardware thread by the hypervisor;

tracking, using the set of PURRs, a job intensity of the job for the set of physical cores of the set of compute nodes; and

measuring, using the set of PURRs with respect to utilization of a set of logical processors of the set of physical cores, a set of physical processing temporal units which indicates a time for the set of physical cores to complete the job.

9. The method of claim 8, wherein measuring, using the set of PURRs with respect to utilization of the set of logical processors of the set of physical cores, the set of physical processing temporal units includes:

incrementing, in response to the hypervisor in the computing resource management system scheduling the job on the set of logical processors of the set of physical cores, a value of the set of physical processing temporal units to modify a processing speed of one or more physical cores of the set of physical cores, wherein the job is indicated to be resource intensive; and

tracking, using the set of PURRs, usage of a set of processing modules for the job which is indicated to be resource intensive, wherein tracking the usage of the set of processing modules includes tracking a usage frequency which indicates a rate at which an access request is processed by a processor module.

21. The method of claim 1, wherein determining the set of wear-leveling data includes:

analyzing the set of thread event data related to usage intensity which indicates an extent to which a processor module is taxed by an access request; and

weighting, when determining the set of wear-leveling data, the access request based on the usage intensity of the access request.

23. The method of claim 1, further comprising:

assigning, by the computing resource management system based on the historical utilization of the set of physical cores related to the number of jobs assigned to the set of physical cores of the set of thread event data a job.

24. The method of claim 23, further comprising:

receiving the job; and

selecting, by the computing resource management system based on the frequency indicated in jobs per unit time at which jobs are assigned to the one or more hardware threads a subset of the set of physical cores to process the job.

26. The method of claim 1, further comprising:

receiving a workload;

selecting, by comparing a first number of jobs of a first set of logical processors of a first physical core with a second number of j obs of a second set of logical processors of a second physical core, the first physical core of the set of physical cores to process the workload, wherein the first and second sets of logical processors represent hardware threads of the first and second physical cores;

calculating, in the simultaneous multi-threading environment, an average job count per a given time;

collecting a rate at which an access request is processed by a processor module, wherein the rate indicates an absolute number of access requests per unit time;

determining, based on the rate at which the access request is processed by a processor module, the set of wear-leveling data;

determining, based on an extent to which the processor module is taxed by the access request, the set of wear-leveling data;

incrementing, in response to the hypervisor scheduling a job on a set of logical processors of the set of physical cores, a value of a set of physical processing temporal units to modify a processing speed of one or more physical cores of the set of physical cores, wherein the job is indicated to be resource intensive in the simultaneous multi-threading environment;

identifying the set of wear-leveling data based on an actual measured usage of the set of physical cores including the rate at which the access request is processed by the processor module; and

sorting, to prioritize usage of hardware processors that have lower wear levels, the set of wear-leveling data based on a lesser utilized hardware processor factor which indicates a lesser likelihood of a hardware processor error event.

27. The method of claim 1, further comprising:

providing, by the set of PURRs, an actual count of physical processing time units utilized by a logical processor; and

determining, by the hypervisorcomputing resource management system using the actual count of physical processing time units utilized by the logical processor, the placement arrangement.

28. The method of claim 6, further comprising:

ascertaining that the set of thread event data indicates that the second physical core has a history of more frequent error events than the first physical core, wherein the set of thread event data indicates at least one abnormal job termination error and at least one lack of resources error.

29. The method of claim 9, further comprising:

weighting, when determining the set of wear-leveling data, the access request based on intensity of the access request.

30. The method of claim 21, further comprising:

receiving a read intensive access request;

identifying a first read frequency for a first processing module, wherein the first read frequency indicates a measure of a number of times the first processing module has performed a read operation per a given time period;

identifying a second read frequency for a second processing module, wherein the second read frequency indicates a measure of a number of times the second processing module has performed a read operation per the given time period;

comparing the first read frequency of the first processing module with the second read frequency of the second processing module;

determining that the first read frequency of the first processing module exceeds the second read frequency of the second processing module; and

processing, based on first read frequency of the first processing module exceeding the second read frequency of the second processing module, the read intensive access request using the second processing module.

31. The method of claim 30, further comprising:

receiving a write intensive access request;

identifying a first write frequency for the first processing module, wherein the first write frequency indicates a measure of a number of times the first processing module has performed a write operation per the given time period;

identifying a second write frequency for the second processing module, wherein the second write frequency indicates a measure of a number of times the second processing module has performed a write operation per the given time period;

comparing the first write frequency of the first processing module with the second write frequency of the second processing module;

determining that the second write frequency of the second processing module exceeds the first write frequency of the first processing module; and

processing, based on second write frequency of the second processing module exceeding the first write frequency of the first processing module, the write intensive access request using the first processing module.

32. The method of claim 1, further comprising:

maintaining at least a portion of the set of wear-leveling data in a hardware hypervisor which is accessible by a cloud hypervisor;

establishing the set of wear-leveling data in a hardware utilization database;

updating, by the cloud hypervisor, the hardware utilization database;

determining the placement arrangement for a set of virtual machines based on the set of wear-leveling data in the hardware utilization database; and

placing, based on the placement arrangement, the set of virtual machines.

33. The method of claim 1, further comprising:

analyzing utilization data for each physical core of the set of physical cores;

determining aggregate PURR counts for each physical core of the set of physical cores; and

determining, based on the aggregate PURR counts for each physical core of the set of physical cores, the set of wear-leveling data.

34. The method of claim 1, wherein the set of thread events includes error events, and further comprising:

determining, with respect to simultaneous access by different threads in the simultaneous multi-threading environment, a thread safety level which indicates a degree to which a given job is verified for operation in the simultaneous multi-threading environment.

35. The method of claim 1, wherein the set of thread events includes job allocations and job completion events, and further comprising:

calculating an average job completion time of a particular hardware thread to complete a job over a period of time;

detecting, over an extended time period, a change in the average job completion time of the particular hardware thread, wherein an increase in the average job completion time of the particular hardware thread indicates wear of the particular hardware thread; and

determining, using the change in average job completion time of the particular hardware thread, the set of wear-leveling data.

36. The method of claim 1, further comprising:

establishing a utilization history graph for the one or more hardware threads; and

identifying, based on the utilization history graph for the one or more hardware threads, a set of trends.

2-5. -5. (canceled)

7. (canceled)

10-12. -12. (canceled)

13. A system for managing a set of wear-leveling data with respect to a set of physical cores of a set of compute nodes in a simultaneous multi-threading environment, the system comprising:

a memory having a set of computer readable computer instructions, and

a processor for executing the set of computer readable instructions, the set of computer readable instructions including;

monitoring, in the simultaneous multi-threading environment using a set of processor utilization resource registers (PURRs), the set of physical cores of the set of compute nodes;

detecting, by monitoring the set of physical cores of the set of compute nodes using the set of PURRs, a set of thread events with respect to the set of physical cores of the set of compute nodes;

collecting a set of thread event data which indicates a historical utilization of the set of physical cores based on the set of thread events, wherein the set of thread event data includes a number of jobs assigned to the set of physical cores;

determining, based on analyzing the set of thread event data related to a frequency at which jobs are assigned to one or more hardware threads, the set of wear-leveling data, wherein the frequency indicates jobs per unit time;

determining, by a computing resource management system based on the set of wear-leveling data, a placement arrangement for a set of assets with respect to the simultaneous multi-threading environment; and

placing, in the simultaneous multi-threading environment by the computing resource management system based on the placement arrangement, the set of assets.

14-17. -17. (canceled)

18. A computer program product for managing a set of wear-leveling data with respect to a set of physical cores of a set of compute nodes in a simultaneous multi-threading environment, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, wherein the computer readable storage medium is not a transitory signal per se, the program instructions executable by a processor to cause the processor to perform a method comprising:

monitoring, in the simultaneous multi-threading environment using a set of processor utilization resource registers (PURRs), the set of physical cores of the set of compute nodes;

detecting, by monitoring the set of physical cores of the set of compute nodes using the set of PURRs, a set of thread events with respect to the set of physical cores of the set of compute nodes;

collecting a set of thread event data which indicates a historical utilization of the set of physical cores based on the set of thread events, wherein the set of thread event data includes a number of jobs assigned to the set of physical cores;

determining, based on analyzing the set of thread event data related to a frequency at which jobs are assigned to one or more hardware threads, the set of wear-leveling data, wherein the frequency indicates jobs per unit time;

determining, by a hypervisorcomputing resource management system based on the set of wear-leveling data, a placement arrangement for a set of assets with respect to the simultaneous multi-threading environment; and

placing, in the simultaneous multi-threading environment by the computing resource management system based on the placement arrangement, the set of assets.

19-20. -20. (canceled)

22. (canceled)

25. (canceled)