Power capping in a composable computing system

1. A computer program product comprising a non-transitory computer readable medium with program instructions embodied therein, the program instructions being configured to be executable by a processor to cause the processor to perform operations comprising:

monitoring a level of a performance metric for a workload instance being executed by a composed system within a pool of composable resources in a composable computing system, wherein the composed system includes a compute resource and an associated hardware resource selected from a data storage resource, a memory resource and/or a graphic processing resource;

identifying a service level agreement for the workload instance, wherein the service level agreement includes a minimum level of the performance metric that the composed system must provide to support the workload instance;

imposing a compute resource power cap on the compute resource; and

imposing an associated hardware resource power cap on the associated hardware resource by sending a power capping command to a baseboard management controller on a server including the associated hardware resource.

2. The computer program product of claim 1 , the operations further comprising:

incrementally reducing the associated hardware resource power cap on the associated hardware resource in response to the monitored level of the performance metric being greater than a setpoint amount above the minimum level of the performance metric established by the service level agreement for the workload instance.

3. The computer program product of claim 2 , further comprising:

incrementally increasing the associated hardware resource power cap on the associated hardware resource in response to the monitored level of the performance metric being less than the minimum level of the performance metric established by the service level agreement for the workload instance.

4. The computer program product of claim 1 , the operations further comprising:

incrementally reducing the associated hardware resource power cap on the associated hardware resource until the monitored level of the performance metric is less than a setpoint amount above the minimum level of the performance metric established by the service level agreement for the workload instance.

5. The computer program product of claim 4 , wherein the associated hardware resource power cap is incrementally reduced by a fixed amount of power.

6. The computer program product of claim 4 , wherein the operation of incrementally reducing the associated hardware resource power cap on the associated hardware resource includes sending a further associated hardware resource power capping command to the baseboard management controller of the associated hardware resource.

7. The computer program product of claim 4 , further comprising:

incrementally increasing the associated hardware resource power cap on the associated hardware resource until the monitored level of the performance metric is greater than the minimum level of the performance metric established by the service level agreement for the workload instance.

8. The computer program product of claim 1 , wherein the associated hardware resource power cap limits the associated hardware resource to a setpoint amount of power.

9. The computer program product of claim 1 , wherein the associated hardware resource power cap identifies a power state for operation of the associated hardware resource.

10. The computer program product of claim 1 , wherein the power capping command is an Intelligent Platform Management Interface command and the power capping command is sent to the baseboard management controller of the associated hardware resource via an Intelligent Platform Management Interface.

11. The computer program product of claim 1 , wherein the power capping command is a Common Information Model command.

12. The computer program product of claim 1 , further comprising:

identifying an excess capacity of the associated hardware resource that is temporarily unusable under the associated hardware resource power cap imposed on the associated hardware resource; and

temporarily releasing the identified excess capacity of the associated hardware resource from the composed system back to the composable computing system resource pool.

13. The computer program product of claim 12 , further comprising:

determining that the associated hardware resource power cap on the associated hardware resource of the composed node has been eliminated or increased; and

requesting return of the temporarily released excess capacity of the associated hardware resource to the composed system.

14. The computer program product of claim 1 , further comprising:

determining a level of the associated hardware resource power cap using a machine learning algorithm.

15. The computer program product of claim 14 , wherein the machine learning algorithm is a Reinforcement Learning algorithm.

16. A computer program product comprising a non-transitory computer readable medium with program instructions embodied therein, the program instructions being configured to be executable by a processor to cause the processor to perform operations comprising:

monitoring a level of a performance metric for a workload instance being executed by a composed system within a pool of composable resources in a composable computing system, wherein the composed system includes a compute resource and a plurality of associated hardware resources selected from a data storage resource, a memory resource and/or a graphic processing resource;

identifying a service level agreement for the workload instance, wherein the service level agreement includes a minimum level of the performance metric that the composed system must provide to support the workload instance;

imposing a compute resource power cap on the compute resource; and

imposing, for each of the plurality of associated hardware resources, an associated hardware resource power cap on the associated hardware resource by sending a power capping command to a baseboard management controller on a server including the associated hardware resource.

17. The computer program product of claim 16 , the operations further comprising:

incrementally reducing, for each of the plurality of associated hardware resources, the associated hardware resource power cap on the associated hardware resource in response to the monitored level of the performance metric being greater than a setpoint amount above the minimum level of the performance metric established by the service level agreement for the workload instance.

18. The computer program product of claim 16 , the operations further comprising:

incrementally reducing, for each of the plurality of associated hardware resources, the associated hardware resource power cap on the associated hardware resource until the monitored level of the performance metric is less than a setpoint amount above the minimum level of the performance metric established by the service level agreement for the workload instance.

19. The computer program product of claim 18 , wherein the associated hardware resource power cap on a first one of the plurality of associated hardware resources is incrementally reduced until the monitored level of the performance metric is less than a setpoint amount above the minimum level of the performance metric established by the service level agreement for the workload instance prior to incrementally reducing the associated hardware resource power cap on a second one of the plurality of associated hardware resources until the monitored level of the performance metric is less than a setpoint amount above the minimum level of the performance metric established by the service level agreement for the workload instance.

20. The computer program product of claim 18 , further comprising:

incrementally increasing, for each of the plurality of associated hardware resources, the associated hardware resource power cap on the associated hardware resource until the monitored level of the performance metric is greater than the minimum level of the performance metric established by the service level agreement for the workload instance.