Abstract
Evaluating, analyzing and predicting the performance of a parallel system is challenging due to the complex inter-play between the application characteristics and architectural features. The overheads in a parallel system that limit its scalability have to be identified and separated in order to enable performance-conscious parallel application design and the development of high-performance parallel machines. We have developed an evaluation framework that uses a combination of experimentation, simulation and analytical modeling to quantify these parallel system overheads. At the heart of this framework is an execution-driven simulation testbed called SPASM which uses a suite of real applications as the workload. We discuss our experiences in using this simulator in a wide range of architectural projects in this paper.
| Original language | English (US) |
|---|---|
| Pages (from-to) | 1021-1028 |
| Number of pages | 8 |
| Journal | Winter Simulation Conference Proceedings |
| DOIs | |
| State | Published - 1997 |
| Event | Proceedings of the 1997 Winter Simulation Conference - Atlanta, GA, USA Duration: Dec 7 1997 → Dec 10 1997 |
All Science Journal Classification (ASJC) codes
- Software
- Modeling and Simulation
- Safety, Risk, Reliability and Quality
- Chemical Health and Safety
- Applied Mathematics
Fingerprint
Dive into the research topics of 'Execution-driven simulators for parallel systems design'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver