Design and Implementation of a Parallel Performance Data Management Framework

Abstract

Empirical performance evaluation of parallel systems and applications can generate significant amounts of performance data and analysis results from multiple experiments as performance is investigated and problems diagnosed. Hence, the management of performance information is a core component of performance analysis tools. To better support tool integration, portability, and reuse, there is a strong motivation to develop performance data management technology that can provide a common foundation for performance data storage, access, merging, and analysis. This paper presents the design and implementation of the Performance Data Management Framework (PerfDMF). PerfDMF addresses objectives of performance tool integration, interoperation, and reuse by providing common data storage, access, 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 extendible toolkit for profile import/export and standard analysis. We describe the PerfDMF objectives and architecture, give detailed explanation of the major components, and show examples of PerfDMF application.

The Tau Parallel Performance System

“…ParaProf ) is TAU's parallel profile analysis and visualization tool. Also distributed with TAU is the PerfDMF (Huck et al 2005) tool providing multi-experiment parallel profile management. Given the wealth of third-party trace analysis and visualization tools, TAU does not implement its own.…”

“…However, the management of performance data from multiple experiments can be logistically difficult, impeding the effective analysis and understanding of performance outcomes. The Performance Data Management Framework (PerfDMF) (Huck et al 2005) provides a common foundation for parsing, storing, querying, and analyzing performance data from multiple experiments, application versions, profiling tools and/or platforms. The PerfDMF design architecture is presented below.…”

Parametric Studies in Eclipse with TAU and PerfExplorer

“…For that reason, we extended the Parallel Tools Platform (PTP) plugin for Eclipse to include a parametric study framework for the TAU plugin in Eclipse. Users can set up a desired study, launch the experiment, and the framework will automatically compile and execute the application with the specified configuration combinations, storing the performance result after each run in a PerfDMF [5] repository. When the experiment is complete, the multiexperiment analysis and data mining tool PerfExplorer [6] is launched, and the automated comparative analysis results are produced.…”

“…TAU includes utilities to convert both its profile and trace output to a diverse array of other performance data formats, allowing performance analysis and visualization in many third party performance analysis programs. Performance profiles are automatically uploaded to a PerfDMF [5] repository for analysis.…”

Performance Profiling and Analysis of DoD Applications Using PAPI and TAU

“…Alternatively, users can use the schema to implement their own local code profiling database. The TAU Performance Data Management Framework (PerfDMF) tools [4] can be used to upload data to the database and the analysis tools described above can be used to retrieve and analyze data stored in the database.…”