초록 ▼

One embodiment of the present invention provides a system that characterizes computer system workloads. During operation, the system collects metrics for a number of workloads of interest as the workloads of interest execute on a computer system. Next, the system uses the collected metrics to build

One embodiment of the present invention provides a system that characterizes computer system workloads. During operation, the system collects metrics for a number of workloads of interest as the workloads of interest execute on a computer system. Next, the system uses the collected metrics to build a statistical regression model, wherein the statistical regression model uses a performance indicator as a response, and uses the metrics as predictors. The system then defines a distance metric between workloads, wherein the distance between two workloads is a function of the differences between metric values for the two workloads. Furthermore, these differences are weighted by corresponding coefficients for the metric values in the statistical regression model.

대표청구항 ▼

What is claimed is: 1. A method for characterizing computer system workloads, comprising: collecting metrics for a number of workloads of interest as the workloads of interest execute on a computer system and storing the collected metrics in a file; using the metrics to build a statistical regressi

What is claimed is: 1. A method for characterizing computer system workloads, comprising: collecting metrics for a number of workloads of interest as the workloads of interest execute on a computer system and storing the collected metrics in a file; using the metrics to build a statistical regression model, wherein the statistical regression model uses a performance indicator as a response, and uses the metrics as predictors; defining a distance metric between workloads, wherein the distance between two workloads is a function of the differences between metric values for the two workloads, wherein the differences are weighted by corresponding coefficients for the metric values in the statistical regression model; using the distance metric to analyze similarities between workloads; and displaying results of the analysis to a user. 2. The method of claim 1, wherein prior to building the statistical regression model, the method further comprises performing a linear analysis on the collected metrics to eliminate metrics that highly correlated with other metrics, thereby reducing the number of metrics that need to be considered. 3. The method of claim 1, wherein the performance indicator for the statistical regression model includes an instruction count for the number of instructions executed on the computer system. 4. The method of claim 3, wherein the statistical regression model represents the instruction count as an additive function of the metrics plus noise, wherein the noise represents effects not captured by the metrics. 5. The method of claim 1, wherein after building the statistical regression model, metrics that do not explain the performance indicator are not subsequently used. 6. The method of claim 1, further comprising: using the distance metric to cluster a set of workloads; and identifying one or more representative workloads for each cluster. 7. The method of claim 1, wherein using the distance metric to analyze similarities between workloads involves identifying a set of representative benchmarks by using the distance metric to minimize a maximum distance between representative benchmarks and a set of workloads that the benchmarks are meant to cover. 8. The method of claim 1, further comprising using the distance metric to classify a customer's workload at a customer's site. 9. The method of claim 1, wherein using the distance metric to analyze similarities between workloads involves selecting a benchmark that approximates a customer's workload; and wherein the method further comprises selecting an architecture for the customer which is optimized for the selected benchmark, and is hence likely to perform well on the customer's workload. 10. The method of claim 1, wherein collecting the metrics involves collecting the metrics from hardware counters on the computer system. 11. A computer-readable storage medium storing instructions that when executed by a computer cause the computer to perform a method for characterizing computer system workloads, the method comprising: collecting metrics for a number of workloads of interest as the workloads of interest execute on a computer system and storing the collected metrics in a file; using the metrics to build a statistical regression model, wherein the statistical regression model uses a performance indicator as a response, and uses the metrics as predictors; defining a distance metric between workloads, wherein the distance between two workloads is a function of the differences between metric values for the two workloads, wherein the differences are weighted by corresponding coefficients for the metric values in the statistical regression model; using the distance metric to analyze similarities between workloads; and displaying results of the analysis to a user. 12. The computer-readable storage medium of claim 11, wherein prior to building the statistical regression model, the method further comprises performing a linear analysis on the collected metrics to eliminate metrics that highly correlated with other metrics, thereby reducing the number of metrics that need to be considered. 13. The computer-readable storage medium of claim 11, wherein the performance indicator for the statistical regression model includes an instruction count for the number of instructions executed on the computer system. 14. The computer-readable storage medium of claim 13, wherein the statistical regression model represents the instruction count as an additive function of the metrics plus noise, wherein the noise represents effects not captured by the metrics. 15. The computer-readable storage medium of claim 11, wherein after building the statistical regression model, metrics that do not explain the performance indicator are not subsequently used. 16. The computer-readable storage medium of claim 11, wherein the method further comprises: using the distance metric to cluster a set of workloads; and identifying one or more representative workloads for each cluster. 17. The computer-readable storage medium of claim 11, wherein using the distance metric to analyze similarities between workloads involves identifying a set of representative benchmarks by using the distance metric to minimize a maximum distance between representative benchmarks and a set of workloads that the benchmarks are meant to cover. 18. The computer-readable storage medium of claim 11, wherein the method further comprises using the distance metric to classify a customer's workload at a customer's site. 19. The computer-readable storage medium of claim 11, wherein using the distance metric to analyze similarities between workloads involves selecting a benchmark that approximates a customer's workload; and wherein the method further comprises selecting an architecture for the customer which is optimized for the selected benchmark, and is hence likely to perform well on the customer's workload. 20. The computer-readable storage medium of claim 11, wherein collecting the metrics involves collecting the metrics from hardware counters on the computer system. 21. A system that characterizes computer system workloads, comprising: a processor; a memory; a receiving mechanism within the processor and the memory configured to receive metrics for a number of workloads of interest as the workloads of interest execute on a computer system and to store the collected metrics in a file; a modeling mechanism within the processor and the memory configured to use the metrics to build a statistical regression model, wherein the statistical regression model uses a performance indicator as a response, and uses the metrics as predictors; a distance-metric generator within the processor and the memory configured to defining a distance metric between workloads, wherein the distance between two workloads is a function of the differences between metric values for the two workloads, wherein the differences are weighted by corresponding coefficients for the metric values in the statistical regression model; an analysis mechanism configured to use the distance metric to analyze similarities between workloads; and an output mechanism configured to display results of the analysis to a user.