Performance evaluation model for multicomputer systems

Authors:
O. B. Efremides;M. P. Bekakos;D. J. Evans
Affiliations:
Department of Electrical & Computer Engineering, School of Engineering, Democritus University of Thrace, Greece;Department of Electrical & Computer Engineering, School of Engineering, Democritus University of Thrace, Greece;Department of Computing Nottingham Trent University, UK
Venue:
Neural, Parallel & Scientific Computations
Year:
2002

Citing 5
Cited 0

A static performance estimator to guide data partitioning decisions

PPOPP '91 Proceedings of the third ACM SIGPLAN symposium on Principles and practice of parallel programming
A methodology for performance analysis of parallel computations with looping constructs

Journal of Parallel and Distributed Computing
Theoretical performance-based cost-effectiveness of multicomputers

The Computer Journal - Special issue on safety and security parallel computation
Evaluating the communication performance of MPPs using synthetic sparse matrix multiplication workloads

ICS '93 Proceedings of the 7th international conference on Supercomputing
Distributed memory systems performance evaluation using synthetic parametric workloads

Neural, Parallel & Scientific Computations

Quantified Score

Hi-index	0.00

Visualization

Abstract

The primary constraining factor of the overall performance of a multicomputer system is communication delay. In this work a performance evaluation framework for a distributed memory parallel computer system is investigated. The evaluation is carried out using a synthetic dynamic workload developed to evaluate both the communication and the computational cost. A sparse matrix, specifying the workload, is produced according to a set of parameters and it is utilized by the sparse matrix - dense vector multiplication algorithm. The procedure is tested evaluating the performance of gigacubes of the Parcytec GCel3/512 message passing parallel system.