System and method to dynamically allocate varying processing capacity entitlements based on workload importance

• US 10,108,459 B2
• Filed: 09/12/2016
• Issued: 10/23/2018
• Est. Priority Date: 09/12/2016
• Status: Active Grant

• Current Assignee: BMC Software Incorporated (KKR & Co., Inc.)
• Original Assignee: BMC Software Incorporated (KKR & Co., Inc.)
• Inventors: Williams, Edward, DeGrange, Steven, Rama, Hemanth, Perini, Robert
• Primary Examiner(s): GHAFFARI, ABU Z
• Application Number:
  US15/262,723
• Publication Number:
  US 20180074859A1
• Time in Patent Office: 771 Days
• Field of Search: None
• CPC Class Codes: G06F 2209/5011 Pool
• CPC Class Codes: G06F 2209/5021 Priority
• CPC Class Codes: G06F 2209/5022 Workload threshold
• CPC Class Codes: G06F 2209/504 Resource capping
• CPC Class Codes: G06F 9/45533 Hypervisors; Virtual machine monitors
• CPC Class Codes: G06F 9/5038 considering the execution order of a plurality of tasks, e.g. taking priority or time dependency constraints into consideration scheduling strategies G06F9/4881 and subgroups
• CPC Class Codes: G06F 9/505 considering the load
• CPC Class Codes: G06F 9/5061 Partitioning or combining of resources
• CPC Class Codes: G06F 9/5077 Logical partitioning of resources; Management or configuration of virtualized resources specific details on emulation or internal functioning of virtual machines G06F9/455
• CPC Class Codes: G06F 9/5083 Techniques for rebalancing the load in a distributed system
• CPC Class Codes: Y02D 10/00 Energy efficient computing, e.g. low power processors, power management or thermal management

Abstract

A mainframe computing system hosts a plurality of logical partitions, each having a static entitlement of processing capacity. The mainframe computer system has a workload manager that schedules work requested by the logical partitions and tracks consumption of the processing capacity by the logical partitions, and a capping policy that is stored in non-transitory memory and which identifies a subset of the logical partitions. The mainframe computer system further includes a capping master that is configured to allocate dynamically varying entitlements of processing capacity to the subset of the logical partitions based on the high-importance work percentages of computing workloads running on the logical partitions to encourage completion of high-importance work over completion of low-importance work. The capping master limits the allocated dynamically entitlement amount in millions of service units per hour (MSU) for each system usage entity to be no greater than the static entitlement of the system usage entity.

Claims

1. A method for using dynamic entitlements as capping limits for regulating processing capacity consumption by system usage entities of a mainframe computer system, the processing capacity consumption being measured in millions of service units per hour (MSUs), each of the system usage entities having a pre-defined static entitlement of processing capacity for running a computing workload thereon, the method comprising:

for each of the system usage entities, determining consumption of processing capacity by the different portions of work corresponding to different pre-assigned importance levels in the computing workload running thereon;

determining a four-hour rolling average (4HRA) of processing capacity consumption by each of the system usage entities;

allocating a dynamic entitlement MSU amount for each of the system usage entities to be the greater of (1) all the MSUs that the system usage entity needs to complete high importance work, and (2) a predetermined fraction of the 4HRA of the system usage entity;

limiting the allocated dynamic entitlement MSU amount for each system usage entity to be less than or at most equal to the static entitlement for the system usage entity;

combining in an excess MSU pool unused MSUs of the static entitlements of all of the system usage entities;

allocating additional MSU amounts out of the excess MSU pool to the dynamic entitlements of the system usage entities, one-by-one, in one or more iterative loops, by;

for each of the system usage entities, determining a high-importance workload percentage (HIWP) as a percentage of standard central processing unit (CPU) consumption of service class periods having high importance over the CPU consumption of all service class periods,

determining a percentage of the 4HRA MSUs consumed by each system usage entity for high importance work as a product of HIWP and 4HRA, and

when the determined percentage of the 4HRA MSUs consumed is greater than the dynamic entitlement MSU amount for the system usage entity, in a first iterative loop, allocating additional MSU amounts out of the excess MSU pool to increase the dynamic entitlement for the system usage entity by approximately the determined percentage of the 4HRA MSUs consumed; and

using the allocated dynamic entitlements as capping limits for regulating processing capacity consumption by each of the system usage entities when running their respective computing workloads.

2. The method of claim 1, wherein the predetermined fraction of the 4HRA is one half.

3. The method of claim 1 further comprising:

for each of the system usage entities, determining a low-importance workload percentage (LIWP) as the percentage of standard CPU consumption of service class periods having low importance over the consumption of all service class periods; and

determining an adjusted low-importance workload percentage (ALIWP) by adjusting the LIWP upward or downward according to a priority associated with the system usage entity,

wherein allocating additional MSU amounts out of the excess MSU pool includes, in each iterative loop, beginning allocating additional MSU amounts with the system usage entity having the lowest ALIWP, proceeding to the system usage entity having the next lowest ALIWP, and ending with the system usage entity having the highest ALIWP.

4. The method of claim 3, wherein determining a low-importance workload percentage (LIWP) includes determining the percentage using an IBM Workload Manager service (IWMRCOLL) available on the mainframe computer system for workload activity collection.

5. The method of claim 1 further comprising:

when there still are unused MSUs left in the excess MSU pool after the first iterative loop, in a second iterative loop, allocating additional MSU amounts out of the excess MSU pool to increase the dynamic entitlement for a the system usage entity by an amount based on the percentage of the 4HRA MSU consumption of the system usage entity.

6. The method of claim 5, wherein the amount based on the percentage of the 4HRA MSU consumption of the system usage entity is (1% 4HRA+1) MSUs rounded to the next highest integer.

7. The method of claim 5 further comprising:

when there still are unused MSUs left in the excess MSU pool after the second iterative loop, in a third iterative loop, allocating additional MSU amounts out of the excess MSU pool to increase the dynamic entitlement for the system usage entity by a fixed number of MSUs.

8. The method of claim 1, wherein using the allocated dynamic entitlements as the capping limits in conjunction with a capping product for regulating processing capacity consumption by the system usage entities when running their respective computing workloads includes using the allocated dynamic entitlements as capping limits below a preset maximum capacity limit for a combined processing capacity consumption of the system usage entities.

9. A capping product for regulating processing capacity consumption by system usage entities of a mainframe computer system, the processing capacity consumption being measured in millions of service units per hour (MSUs), each of the system usage entities having a static entitlement of processing capacity for running a computing workload thereon, the capping product comprising a non-transitory computer-readable medium storing instructions which when executed by one or more processors cause the computer system to:

for each of the system usage entities, determine consumption of processing capacity by the different portions of work corresponding to different pre-assigned importance levels in the computing workload running thereon;

determine a four-hour rolling average (4HRA) of processing capacity consumption by each of the system usage entities;

allocate a dynamic entitlement MSU amount for each of the system usage entities to be the greater of (1) all the MSUs that the system usage entity needs to complete high importance work, and (2) a predetermined fraction of the 4HRA of the system usage entity;

limit the allocated dynamic entitlement MSU amount for each system usage entity to be less than or at most equal to the static entitlement of the system usage entity;

combine in an excess MSU pool unused MSUs of the static entitlements of all the system usage entities;

allocate additional MSU amounts out of the excess MSU pool, to the dynamic entitlements of the system usage entities, one-by-one, in one or more iterative loops, by;

for each of the system usage entity, determining a high-importance workload percentage (HIWP) as a percentage of standard central processing unit (CPU) consumption of service class periods having high importance over the CPU consumption of all service class periods,

determining a percentage of the 4HRA MSUs consumed by each system usage entity for high importance work as a product of HIWP and 4HRA, and

when the determined percentage of the 4HRA MSUs consumed is greater than dynamic entitlement MSU amount for the system usage entity, in a first iterative loop, allocating additional MSU amounts out of the excess MSU pool to increase the dynamic entitlement for the system usage entity by approximately the determined percentage of the 4HRA MSUs consumed; and

use the allocated dynamic entitlements as capping limits for regulating processing capacity consumption by each of the system usage entities when running their respective computing workloads.

10. The capping product of claim 9, wherein the predetermined fraction of the 4HRA is one half.

11. The capping product of claim 9, wherein the instructions when executed by the one or more processors further cause the computer system to:

for each system usage entity, determine a low-importance workload percentage (LIWP) as the percentage of standard CPU consumption of service class periods having low importance over the consumption of all service class periods; and

determine an adjusted low-importance workload percentage (ALIWP) by adjusting the LIWP upward or downward according to a priority associated with the system usage entity,

wherein allocating additional MSU amounts out of the excess MSU pool includes, in each iterative loop, beginning allocating additional MSU amounts with the system usage entity having the lowest ALIWP, proceeding to the system usage entity having the next lowest ALIWP, and ending with the system usage entity having the highest ALIWP.

12. The capping product of claim 11, wherein the instructions when executed by the one or more processors further cause the computer system to determine the percentage of high-importance work by using an IBM Workload Manager service (IWMRCOLL) available on the mainframe computer system.

13. The capping product of claim 9, wherein the instructions when executed by the one or more processors further cause the computer system to:

when there still are unused MSUs left in the excess MSU pool after the first iterative loop, in a second iterative loop, allocate additional MSU amounts out of the excess MSU pool to increase the dynamic entitlement for the system usage entity by an amount based on a percentage of the 4HRA MSU consumption of the system usage entity.

14. The capping product of claim 13, wherein the amount based on the percentage of the 4HRA MSU consumption of the system usage entity is (1% 4HRA+1) MSUs rounded to the next highest integer.

15. The capping product of claim 13, wherein the instructions when executed by the one or more processors further cause the computer system to:

when there still are unused MSUs left in the excess MSU pool after the second iterative loop, in a third iterative loop, allocating additional MSU amounts out of the excess MSU pool to increase the dynamic entitlement for the system usage entity by a fixed number of MSUs.

16. The capping product of claim 13, wherein the instructions when executed by the one or more processors further cause the computer system to:

regulate processing capacity consumption by the system usage entities when running their respective computing workloads by using the allocated dynamic entitlements as capping limits below a preset maximum capacity limit for a combined processing capacity consumption of the system usage entities.

17. A mainframe computing system comprising:

a central processor complex;

a plurality of system usage entities, each of the system usage entities having a static entitlement of processing capacity for running a computing workload thereon;

a workload manager that schedules work requested by the plurality of system usage entities on the central processor complex and tracks, by system usage entity, consumption of the processing capacity by a computing workload running on the system usage entity, the processing capacity consumption being measured in millions of service units per hour (MSUs);

a dynamic capping policy for the central processor complex stored in non-transitory memory, the dynamic capping policy identifying;

a maximum processing capacity limit,

a subset of the plurality of system usage entities, and

for each of the identified system usage entities, information from which to determine a dynamically varying entitlement value of processing capacity for the system usage entity on which a computing workload is being executed; and

a dynamic capping master configured to;

for each of the identified system usage entities, determine consumption of processing capacity by the different portions of work corresponding to different pre-assigned importance levels in the computing workload running thereon;

determine a four-hour rolling average (4HRA) of processing capacity consumption by each of the identified system usage entities;

allocate a dynamic entitlement MSU amount for each of the identified system usage entities to be the greater of (1) all the MSUs that the system usage entity needs to complete high importance work, and (2) a predetermined fraction of the 4HRA of the system usage entity;

limit the allocated dynamic entitlement MSU amount for each system usage entity to be less than or at most equal to the static entitlement of the system usage entity;

combine in an excess MSU pool unused MSUs of the static entitlements of all the system usage entities;

allocate additional MSU amounts out of the excess MSU pool to the dynamic entitlements of the system usage entities, one-by-one, in one or more iterative loops, by;

for each of the system usage entity, determining a high-importance workload percentage (HIWP) as a percentage of standard central processing unit (CPU) consumption of service class periods having high importance over the CPU consumption of all service class periods,

determining a percentage of the 4HRA MSUs consumed by each system usage entity for high importance work as a product of HIWP and 4HRA, and

when determined percentage of the 4HRA MSUs consumed is greater than dynamic entitlement MSU amount for the system usage entity, in a first iterative loop, allocating additional MSU amounts out of the excess MSU pool to increase the dynamic entitlement for the system usage entity by approximately the determined percentage of the 4HRA MSUs consumed, and

use the allocated dynamic entitlements as capping limits for regulating processing capacity consumption by each of the system usage entities when running their respective computing workloads.

18. The mainframe computing system of claim 17, wherein the predetermined fraction of the 4HRA is one half.

19. The mainframe computing system of claim 17, wherein the dynamic capping master is configured to:

for each of the system usage entities, determine a low-importance workload percentage (LIWP) as the percentage of standard CPU consumption of service class periods having low importance over the consumption of all service class periods; and

determine an adjusted low-importance workload percentage (ALIWP) by adjusting the LIWP upward or downward according to a priority associated with the system usage entity,

wherein allocating additional MSU amounts out of the excess MSU pool includes, in each iterative loop, beginning allocating additional MSU amounts with the system usage entity having the lowest ALIWP, proceeding to the system usage entity having the next lowest ALIWP, and ending with the system usage entity having the highest ALIWP.

20. The mainframe computing system of claim 19, wherein the dynamic capping master is configured to:

determine the low-importance workload percentage (LIWP) using an IBM Workload Manager service (IWMRCOLL) available on the mainframe computer system for workload activity collection.

21. The mainframe computing system of claim 17, wherein the dynamic capping master is configured to:

when there still are unused MSUs left in the excess MSU pool after the first iterative loop, in a second iterative loop, allocate additional MSU amounts out of the excess MSU pool to increase the dynamic entitlement for a system usage entity by an amount based on a percentage of the 4HRA MSU consumption of the system usage entity.

22. The mainframe computing system of claim 21, wherein the amount based on the percentage of the 4HRA MSU consumption of the system usage entity is (1% 4HRA+1) MSUs rounded to the next highest integer.

23. The mainframe computing system of claim 21, wherein the dynamic capping master is configured to:

when there still are unused MSUs left in the excess MSU pool after the second iterative loop, in a third iterative loop, allocate additional MSU amounts out of the excess MSU pool to increase the dynamic entitlement for a system usage entity by a fixed number of MSUs.

24. The mainframe computing system of claim 17, wherein the dynamic capping master is configured to:

regulate processing capacity consumption by the system usage entities when running their respective computing workloads by using the allocated dynamic entitlements as capping limits below a preset maximum capacity limit for a combined processing capacity consumption of the system usage entities.