Estimation of Execution times on Heterogeneous Supercomputer Architectures
- 1 August 1993
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 1 (01903918) , 219-226
- https://doi.org/10.1109/icpp.1993.80
Abstract
For managing tasks efficiently in a Distributed Het erogeneous Supercomputing System (DHSS) , we re quire a thorough understanding of applications and their intelligent scheduling within the system. For this purpose, an accurate estimation of the execu tion time of applications on various architectures is needed. In this paper we present a framework to ad dress this issue. We propose two techniques, called augmented code profiling and augmented analytical benchmarking, to characterize applications and archi tectures in a DHSS, respectively. These techniques are based on code profiling and analytical benchmarking, respectively and provide a detailed architectural-dependent characterization of DHSS applications.Keywords
This publication has 13 references indexed in Scilit:
- Augmenting the Optimal Selection Theory for SuperconcurrencyPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2005
- On the assignment problem of arbitrary process systems to heterogeneous distributed computer systemsIEEE Transactions on Computers, 1992
- Performance of synchronous parallel algorithms with regular structuresIEEE Transactions on Parallel and Distributed Systems, 1991
- Benchmark characterizationComputer, 1991
- PAWS: a performance evaluation tool for parallel computing systemsComputer, 1991
- Gigabit network testbedsComputer, 1990
- Hypertool: a programming aid for message-passing systemsIEEE Transactions on Parallel and Distributed Systems, 1990
- Characterizing computer performance with a single numberCommunications of the ACM, 1988
- Computer benchmarking: Paths and pitfalls: The most popular way of rating computer performance can confuse as well as inform; avoid misunderstanding by asking just what the benchmark is measuringIEEE Spectrum, 1987
- The Effects of Problem Partitioning, Allocation, and Granularity on the Performance of Multiple-Processor SystemsIEEE Transactions on Computers, 1987