Abstract
With the rapid development of high performance computing, reduction of the gap between software and hardware is an urgent issue to be resolved. Performance monitor and analysis techniques, as the important way to improve parallel application efficiency, still face the challenge of scalability as well as online analysis. The research designs and realizes the implementation of a novel prototype system, PAPMSA, for parallel application performance monitor and analysis. The system mainly adopts the mechanism of dynamic binary instrumentation to acquire performance data, which will then converge into the database through the scalable transmission architecture. The system can be visually displayed via web with online and post-mortem analysis. In addition, the scalable system can be applied to analyze large-scale application. This paper introduces the realization mechanisms of PAPMAS in detail with respect to the data acquiring, data transmission and analysis, etc, and meanwhile evaluates and verifies the PAPMAS' applicability via PPerfMark test benchmarks.
| Original language | English |
|---|---|
| Pages (from-to) | 64-72 |
| Number of pages | 9 |
| Journal | International Journal of Advancements in Computing Technology |
| Volume | 3 |
| Issue number | 8 |
| DOIs | |
| State | Published - Sep 2011 |
Keywords
- Instrumentation
- Online analysis
- Performance bottleneck
- Profiling
- Scalability
- Tracing
Fingerprint
Dive into the research topics of 'PAPMAS: A novel prototype system for parallel application performance monitor and analysis'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver