A Hierarchical Statistic Methodology for Advanced Memory System Evaluation
- Los Alamos National laboratory
Advances in technology have resulted in a widening of the gap between computing speed and memory access time. Data access time has become increasingly important for computer system design. Various hierarchical memory architectures have been developed. The performance of these advanced memory systems, however, varies with applications and problem sizes. How to reach an optimal cost/performance design eludes researchers still. In this study, the authors introduce an evaluation methodology for advanced memory systems. This methodology is based on statistical factorial analysis and performance scalability analysis. It is two fold: it first determines the impact of memory systems and application programs toward overall performance; it also identifies the bottleneck in a memory hierarchy and provides cost/performance comparisons via scalability analysis. Different memory systems can be compared in terms of mean performance or scalability over a range of codes and problem sizes. Experimental testing has been performed extensively on the Department of Energy's Accelerated Strategic Computing Initiative (ASCI) machines and benchmarks available at the Los Alamos National Laboratory to validate this newly proposed methodology. Experimental and analytical results show this methodology is simple and effective. It is a practical tool for memory system evaluation and design. Its extension to general architectural evaluation and parallel computer systems are possible and should be further explored.
- Research Organization:
- Los Alamos National Lab., NM (US)
- Sponsoring Organization:
- USDOE Office of Defense Programs (DP) (US)
- DOE Contract Number:
- W-7405-ENG-36
- OSTI ID:
- 760037
- Report Number(s):
- LA-UR-98-4176
- Country of Publication:
- United States
- Language:
- English
Similar Records
Performance characterization and validation of ASCI applications: A memory centric view
The Tradeoffs of Fused Memory Hierarchies in Heterogeneous Computing Architectures