Design and implementation of a parallel performance data management framework

Empirical performance evaluation of parallel systems and applications can generate significant amounts of per-formance data and analysis results from multiple ex-periments as performance is investigated and problems diagnosed. Hence, the management of performance in-formation is a core component of performance analy-sis tools. To better support tool integration, portability, and reuse, there is a strong motivation to develop per-formance data management technology that can provide a common foundation for performance data storage, ac-cess, merging, and analysis. This paper presents the de-sign and implementation of the Performance Data Man-agement Framework (PerfDMF). PerfDMF addresses objectives of performance tool integration, interopera-tion, and reuse by providing common data storage, ac-cess, and analysis infrastructure for parallel performance profiles. PerfDMF includes an extensible parallel profile data schema and relational database schema, a profile query and analysis programming interface, and an ex-tendible toolkit for profile import/export and standard analysis. We describe the PerfDMF objectives and ar-chitecture, give detailed explanation of the major compo-nents, and show examples of PerfDMF application. 1