Scalable Analysis Techniques for Microprocessor Performance Counter Metrics

@inproceedings{citeulike:2649707, abstract = {Contemporary microprocessors provide a rich set of integrated performance counters that allow application developers and system architects alike the opportunity to gather important information about workload behaviors. Current techniques for analyzing data produced from these counters use raw counts, ratios, and visualization techniques help users make decisions about their application performance. While these techniques are appropriate for analyzing data from one process, they do not scale easily to new levels demanded by contemporary computing systems. Very simply, this paper addresses these concerns by evaluating several multivariate statistical techniques on these datasets. We find that several techniques, such as statistical clustering, can automatically extract important features from the data. These derived results can, in turn, be fed directly back to an application developer, or used as input to a more comprehensive performance analysis environment, such as a visualization or an expert system.}, author = {Ahn, D. H. and Vetter, J. S. }, booktitle = {Supercomputing, ACM/IEEE 2002 Conference}, citeulike-article-id = {2649707}, doi = {10.1109/SC.2002.10066}, journal = {Supercomputing, ACM/IEEE 2002 Conference}, keywords = {performance\_analysis}, pages = {3}, posted-at = {2008-04-10 16:13:56}, priority = {2}, title = {Scalable Analysis Techniques for Microprocessor Performance Counter Metrics}, url = {http://dx.doi.org/10.1109/SC.2002.10066}, year = {2002} }

TY - CONF ID - citeulike:2649707 L3 - citeulike-article-id:2649707 TI - Scalable Analysis Techniques for Microprocessor Performance Counter Metrics T2 - Supercomputing, ACM/IEEE 2002 Conference JF - Supercomputing, ACM/IEEE 2002 Conference SP - 3 EP - 3 N2 - Contemporary microprocessors provide a rich set of integrated performance counters that allow application developers and system architects alike the opportunity to gather important information about workload behaviors. Current techniques for analyzing data produced from these counters use raw counts, ratios, and visualization techniques help users make decisions about their application performance. While these techniques are appropriate for analyzing data from one process, they do not scale easily to new levels demanded by contemporary computing systems. Very simply, this paper addresses these concerns by evaluating several multivariate statistical techniques on these datasets. We find that several techniques, such as statistical clustering, can automatically extract important features from the data. These derived results can, in turn, be fed directly back to an application developer, or used as input to a more comprehensive performance analysis environment, such as a visualization or an expert system. KW - performance_analysis AU - Ahn, DH AU - Vetter, JS PY - 2002/// UR - http://dx.doi.org/10.1109/SC.2002.10066 ER -