Performance Engineering Technology for Scientific Component Software
Large-scale, complex scientific applications are beginning to benefit from the use of component software design methodology and technology for software development. Integral to the success of component-based applications is the ability to achieve high-performing code solutions through the use of performance engineering tools for both intra-component and inter-component analysis and optimization. Our work on this project aimed to develop performance engineering technology for scientific component software in association with the DOE CCTTSS SciDAC project (active during the contract period) and the broader Common Component Architecture (CCA) community. Our specific implementation objectives were to extend the TAU performance system and Program Database Toolkit (PDT) to support performance instrumentation, measurement, and analysis of CCA components and frameworks, and to develop performance measurement and monitoring infrastructure that could be integrated in CCA applications. These objectives have been met in the completion of all project milestones and in the transfer of the technology into the continuing CCA activities as part of the DOE TASCS SciDAC2 effort. In addition to these achievements, over the past three years, we have been an active member of the CCA Forum, attending all meetings and serving in several working groups, such as the CCA Toolkit working group, the CQoS working group, and the Tutorial working group. We have contributed significantly to CCA tutorials since SC'04, hosted two CCA meetings, participated in the annual ACTS workshops, and were co-authors on the recent CCA journal paper [24]. There are four main areas where our project has delivered results: component performance instrumentation and measurement, component performance modeling and optimization, performance database and data mining, and online performance monitoring. This final report outlines the achievements in these areas for the entire project period. The submitted progress reports for the first two years describe those year's achievements in detail. We discuss progress in the last project period in this document. Deployment of our work in CCA components, frameworks, and applications is an important metric of success. We also summarize the project's accomplishments in this regard at the end of the report. A list of project publications is also given.
- Research Organization:
- University of Oregon, Eugene, OR
- Sponsoring Organization:
- USDOE - Office of Public Affairs (PA); USDOE Office of Science (SC)
- DOE Contract Number:
- FG02-03ER25561
- OSTI ID:
- 909872
- Report Number(s):
- DOE/ER25561-Final; TRN: US200816%%658
- Country of Publication:
- United States
- Language:
- English
Similar Records
Center for Technology for Advanced Scientific Componet Software (TASCS)
Knowledge-Based Parallel Performance Technology for Scientific Application Competitiveness Final Report