Granted Patent B1

US 6,970,805 · App. 10/744,995 · Granted Nov 29, 2005

Abstract

Claims (60)

1. A computer-implemented method for determining an optimal workload level for a data processing system, comprising:

collecting samples of system operating characteristics, each sample including a workload level, a throughput level, and one or more performance metrics having values used in detection of a bottleneck condition, wherein a bottleneck condition exists if the workload level, the throughput level, and one or more of the performance metrics in a sample have values equal to a predetermined set of values defined as a bottleneck condition;

identifying a first workload level that corresponds to the sample having a greatest throughput level; and

selecting a workload level from among workload levels of one or more samples as the optimal workload level, wherein in each of the one or more samples the workload level is less than the first workload level, the throughput level is less than the greatest throughput level, and no bottleneck condition is indicated by the performance metrics.

2. The method of claim 1 , further comprising:

selecting a subset of the samples as a degraded performance data set, wherein the degraded performance data set includes samples with workload levels greater than the first workload level, and if the degraded performance data set is not empty selecting a degraded performance sample from the degraded performance data set;

selecting a subset of the samples as an optimal performance data set, wherein the optimal performance data set includes samples with workload levels less than the first workload level; and

selecting a sample from the optimal performance data set as a function of a workload level of the degraded performance sample, wherein the workload level of the sample is the optimal workload level.

3. The method of claim 2 , further comprising:

determining an average degraded workload level as an average of workload levels of samples in the degraded performance data set; and

analyzing samples in the degraded performance data set having workload levels within a range from a selected amount greater to a selected amount less than the average degraded workload level in order from a least workload level to a greatest workload level and establishing a first-encountered sample with a bottleneck condition as the degraded performance sample.

4. The method of claim 2 , further comprising:

determining an acceptable workload level in the optimal performance data set as a function of the workload level of the degraded performance sample; and

analyzing samples in the optimal performance data set having workload levels within a range from a selected amount greater to a selected amount less than the acceptable workload level in order from a greatest workload level to a least workload level and establishing the workload level of a first-encountered sample without a bottleneck condition as the optimal workload level.

5. The method of claim 4 , wherein determining an acceptable workload level comprises finding a sample in the optimal performance data set with a workload level that is approximately equal to the workload level of the degraded performance sample.

6. The method of claim 2 , further comprising:

if the degraded performance data set is empty, determining an acceptable workload level in the optimal performance data set as an average of a least workload level in the optimal performance data set and the first workload level; and

analyzing samples in the optimal performance data set with workload levels within a range from a selected amount greater to a selected amount less than the acceptable workload level in order from a greatest workload level to a least workload level and establishing the workload level of a first-encountered sample without a bottleneck condition as the optimal workload level.

7. The method of claim 1 , further comprising:

establishing a baseline performance level;

repeating the steps of identifying a first workload level and selecting the optimal workload level;

determining a candidate baseline performance level as a function of the workload level, throughput level, and one or more performance metrics from the sample of the optimal workload level;

replacing the baseline performance level with the candidate performance level if the candidate performance level is greater than the baseline performance level.

8. A system for determining an optimal workload level for a data processing system, comprising:

a sample collector configured to collect samples of system operating characteristics, each sample including a workload level, a throughput level, and one or more performance metrics having values used in detection of a bottleneck condition, wherein a bottleneck condition exists if the workload level, the throughput level, and one or more of the performance metrics in a sample have values equal to a predetermined set of values defined as a bottleneck condition;

an analyzer configured to identify a first workload level that corresponds to the sample having a greatest throughput level and select a workload level from among workload levels of one or more samples as the optimal workload level, wherein in each of the one or more samples the workload level is less than the first workload level, the throughput level is less than the greatest throughput level, and no bottleneck condition is indicated by the performance metrics.

9. The system of claim 8 , wherein the analyzer is further configured to select a subset of the samples as a degraded performance data set, wherein the degraded performance data set includes samples with workload levels greater than the first workload level, and if the degraded performance data set is not empty selecting a degraded performance sample from the degraded performance data set, select a subset of the samples as an optimal performance data set, wherein the optimal performance data set includes samples with workload levels less than the first workload level, and select a sample from the optimal performance data set as a function of a workload level of the degraded performance sample, wherein the workload level of the sample is the optimal workload level.

10. The system of claim 9 , wherein the analyzer is further configured to determine an average degraded workload level as an average of workload levels of samples in the degraded performance data set, analyze samples in the degraded performance data set having workload levels within a range from a selected amount greater to a selected amount less than the average degraded workload level in order from a least workload level to a greatest workload level, and establish a first-encountered sample with a bottleneck condition as the degraded performance sample.

11. The system of claim 9 , wherein the analyzer is further configured to determine an acceptable workload level in the optimal performance data set as a function of the workload level of the degraded performance sample, analyze samples in the optimal performance data set having workload levels within a range from a selected amount greater to a selected amount less than the acceptable workload level in order from a greatest workload level to a least workload level, and establish the workload level of a first-encountered sample without a bottleneck condition as the optimal workload level.

12. The system claim 11 , wherein the analyzer is further configured to find a sample in the optimal performance data set with a workload level that is approximately equal to the workload level of the degraded performance sample in determining an acceptable workload level.

13. The system of claim 9 , wherein the analyzer is further configured to, if the degraded performance data set is empty, determine an acceptable workload level in the optimal performance data set as an average of a least workload level in the optimal performance data set and the first workload level, analyze samples in the optimal performance data set with workload levels within a range from a selected amount greater to a selected amount less than the acceptable workload level in order from a greatest workload level to a least workload level, and establish the workload level of a first-encountered sample without a bottleneck condition as the optimal workload level.

14. The system of claim 8 , wherein the analyzer is further configured to establish a baseline performance level, repeat the steps of identifying a first workload level and selecting the optimal workload level, determine a candidate baseline performance level as a function of the workload level, throughput level, and one or more performance metrics from the sample of the optimal workload level, and replace the baseline performance level with the candidate performance level if the candidate performance level is greater than the baseline performance level.

15. An article of manufacture for determining an optimal workload level for a data processing system, comprising:

a computer-readable medium configured with instructions for causing a processor to perform the steps of,

collecting samples of system operating characteristics, each sample including a workload level, a throughput level, and one or more performance metrics having values used in detection of a bottleneck condition, wherein a bottleneck condition exists if the workload level, the throughput level, and one or more of the performance metrics in a sample have values equal to a predetermined set of values defined as a bottleneck condition;

identifying a first workload level that corresponds to the sample having a greatest throughput level; and

selecting a workload level from among workload levels of one or more samples as the optimal workload level, wherein in each of the one or more samples the workload level is less than the first workload level, the throughput level is less than the greatest throughput level, and no bottleneck condition is indicated by the performance metrics.

16. The article of manufacture of claim 15 , wherein the computer-readable medium is further configured with instructions for causing a computer to perform the steps of:

selecting a subset of the samples as a degraded performance data set, wherein the degraded performance data set includes samples with workload levels greater than the first workload level, and if the degraded performance data set is not empty selecting a degraded performance sample from the degraded performance data set;

selecting a subset of the samples as an optimal performance data set, wherein the optimal performance data set includes samples with workload levels less than the first workload level; and

selecting a sample from the optimal performance data set as a function of a workload level of the degraded performance sample, wherein the workload level of the sample is the optimal workload level.

17. The article of manufacture of claim 16 , wherein the computer-readable medium is further configured with instructions for causing a computer to perform the steps of:

determining an average degraded workload level as an average of workload levels of samples in the degraded performance data set; and

analyzing samples in the degraded performance data set having workload levels within a range from a selected amount greater to a selected amount less than the average degraded workload level in order from a least workload level to a greatest workload level and establishing a first-encountered sample with a bottleneck condition as the degraded performance sample.

18. The article of manufacture of claim 16 , wherein the computer-readable medium is further configured with instructions for causing a computer to perform the steps of:

determining an acceptable workload level in the optimal performance data set as a function of the workload level of the degraded performance sample; and

analyzing samples in the optimal performance data set having workload levels within a range from a selected amount greater to a selected amount less than the acceptable workload level in order from a greatest workload level to a least workload level and establishing the workload level of a first-encountered sample without a bottleneck condition as the optimal workload level.

19. The article of manufacture of claim 18 , wherein the computer-readable medium is further configured with instructions for causing a computer to perform the step of finding a sample in the optimal performance data set with a workload level that is approximately equal to the workload level of the degraded performance sample in determining an acceptable workload level comprises.

20. The article of manufacture of claim 16 , wherein the computer-readable medium is further configured with instructions for causing a computer to perform the steps of:

if the degraded performance data set is empty, determining an acceptable workload level in the optimal performance data set as an average of a least workload level in the optimal performance data set and the first workload level; and

analyzing samples in the optimal performance data set with workload levels within a range from a selected amount greater to a selected amount less than the acceptable workload level in order from a greatest workload level to a least workload level and establishing the workload level of a first-encountered sample without a bottleneck condition as the optimal workload level.

21. The article of manufacture of claim 15 , wherein the computer-readable medium is further configured with instructions for causing a computer to perform the steps of:

establishing a baseline performance level;

repeating the steps of identifying a first workload level and selecting the optimal workload level; and

determining a candidate baseline performance level as a function of the workload level, throughput level, and one or more performance metrics from the sample of the optimal workload level; and

replacing the baseline performance level with the candidate performance level if the candidate performance level is greater than the baseline performance level.

22. An apparatus for determining an optimal workload level for a data processing system, comprising:

means for collecting samples of system operating characteristics, each sample including a workload level, a throughput level, and one or more performance metrics having values used in detection of a bottleneck condition, wherein a bottleneck condition exists if the workload level, the throughput level, and one or more of the performance metrics in a sample have values equal to a predetermined set of values defined as a bottleneck condition;

means for identifying a first workload level that corresponds to the sample having a greatest throughput level; and

means for selecting a workload level from among workload levels of one or more samples as the optimal workload level, wherein in each of the one or more samples the workload level is less than the first workload level, the throughput level is less than the greatest throughput level, and no bottleneck condition is indicated by the performance metrics.

Assignments (11)