Workload management controller using dynamic statistical control

1. A computer system comprising:

a workload management controller that detects and tracks resource consumption volatility patterns and automatically and dynamically adjusts resource headroom according to the resource-consumption volatility patterns, the controller being a hardware controller or a combination of software and hardware executing the software, the workload management controller including a workload monitor that determines volatility of at least one measurable metric by calculating a standard deviation of the at least one measurable metric.

2. The computer system according to claim 1 further comprising:

the workload management controller further comprising an initialization logic that specifies minimum and maximum resources to be applied to ones of a plurality of workloads and a goal based on a measurable metric, and specifies an initial headroom amount for reducing a frequency of occurrences of workload demand exceeding allocated resources.

3. The computer system according to claim 1 , wherein the at least one measurable metric is selected from a group consisting of central processing unit (CPU) utilization, response time, number of users, workload queue length, memory consumption, input/output device usage, network input/output traffic volume, and disk input/output volume.

4. The computer system according to claim 1 further comprising:

the workload management controller further comprising a workload adjuster that automatically adjusts headroom values as volatility of the at least one measurable metric increases or decreases during normal workload operation.

5. The computer system according to claim 1 wherein the workload management controller further determines and tracks volatility of a plurality of measurable metric variables, computing statistical indices for the variables, and iteratively changing entitlements based on the computed statistical indices.

6. The computer system according to claim 1 further comprising:

the workload management controller further comprising the workload monitor that determines volatility of the at least one measurable metric comprising calculating a standard deviation based on short-term or long-term historical data, or a combination of short-term and long-term historical data.

7. The computer system according to claim 1 wherein the workload management controller detects and tracks resource consumption volatility patterns for at least one resource selected from a group consisting of central processing units (CPUs), memory, disk storage, disk input/output (I/O) interfaces, virtual machines (VMs), virtual partitions (vPar), and physical partitions (nPar).

8. A computer-implemented workload management method comprising:

a computer detecting and tracking resource consumption volatility patterns, the tracking including determining a volatility of at least one measurable metric by calculating a standard deviation of the at least one measurable metric; and

the computer automatically and dynamically adjusting resource headroom according to the resource-consumption volatility patterns.

9. The computer-implemented workload management method according to claim 8 further comprising:

initializing workload management control comprising:

specifying minimum and maximum resources to be applied to ones of a plurality of workloads;

specifying a goal based on a measurable metric; and

specifying an initial headroom amount for reducing a frequency of occurrences of workload demand exceeding allocated resources.

10. The computer-implemented workload management method according to claim 8

wherein the at least one measurable metric is selected from a group consisting of central processing unit (CPU) utilization, response time, number of users, workload queue length, memory consumption, input/output device usage, network input/output traffic volume, and disk input/output volume.

11. The computer-implemented workload management method according to claim 8 further comprising:

collecting the at least one measurable metric at selected time intervals;

determining whether workload meets a predetermined goal; and

determining changes in entitlements to address deviations.

12. The computer-implemented workload management method according to claim 8 further comprising:

analyzing at least one measurable metric;

determining volatility of the at least one measurable metric; and

automatically determining headroom values.

13. The computer-implemented workload management method according to claim 12 further comprising:

automatically adjusting headroom values as volatility of the at least one measurable metric increases or decreases during normal workload operation.

14. The computer-implemented workload management method according to claim 12 further comprising:

determining likelihood of a spike in load during a subsequent time interval.

15. The computer-implemented workload management method according to claim 12 further comprising:

determining and tracking volatility for a plurality of measurable metric variables;

computing statistical indices for the variables; and

iteratively changing entitlements based on the computed statistical indices.

16. The computer-implemented workload management method according to claim 12 further comprising:

determining volatility of the at least one measurable metric comprising calculating a standard deviation based on short-term or long-term historical data, or a combination of short-term and long-term historical data.

17. The computer-implemented workload management method according to claim 8 further comprising:

detecting and tracking resource consumption volatility patterns for at least one resource selected from a group consisting of central processing units (CPUs), memory, disk storage, disk input/output (I/O) interfaces, virtual machines (VMs), virtual partitions (vPar), and physical partitions (nPar).

18. An article of manufacture comprising a non-transitory controller-usable medium having a computer readable program code embodied therein for workload management control, the computer readable program code including:

a code configured to, when executed by a processor, cause the controller to detect and track resource consumption volatility patterns at least in part by determining volatility of at least one measurable metric by calculating a standard deviation of the at least one measurable metric; and

a code configured to, when executed by a processor, cause the controller to automatically and dynamically adjust resource headroom according to the resource-consumption volatility patterns.

19. A system comprising non-transitory computer-readable media encoded with code configured to, when executed by a processor,

track utilization by workloads of hardware resources of a computer system to yield utilization data;

calculate respective utilization volatilities for respective workloads at least in part using said utilization data by calculating a standard deviation based on said utilization data;

determine respective projected amounts of said hardware resources expected to be consumed by respective workloads; and

allocate respective actual amounts of said resources to respective workloads, respective actual amounts including respective projected amounts plus respective headrooms, respective headrooms being determined as a function of respective utilization volatilities.

20. The system as recited in claim 19 further comprising said processor.

21. A computer-implemented method comprising:

tracking utilization by workloads of hardware resources of a computer system to yield utilization data;

calculating respective utilization volatilities for respective workloads using said utilization data, said calculating including determining a standard deviation based on said utilization data;

determining respective projected amounts of said hardware resources expected to be consumed by respective workloads; and

allocating respective actual amounts of said resources to respective workloads, respective actual amounts including respective projected amounts plus respective headrooms, respective headrooms being determined as a function of respective utilization volatilities.