Sensitivity of Performance Prediction of Message Passing Programs

Authors:
Sergi Girona;Jesús Labarta
Affiliations:
CEPBA-UPC, Technical University of Catalonia, Barcelona, Spain;CEPBA-UPC, Technical University of Catalonia, Barcelona, Spain
Venue:
The Journal of Supercomputing
Year:
2000

Citing 8
Cited 2

Queueing models of parallel applications: the Glamis methodology

Proceedings of the 7th international conference on Computer performance evaluation : modelling techniques and tools: modelling techniques and tools
Performance measurement, visualization and modeling of parallel and distributed programs using the AIMS toolkit

Software—Practice & Experience
Analyzing scheduling policies using Dimemas

Parallel Computing - Special double issue on environment and tools for parallel scientific computing
Assessing the Performance of the New IBM SP2 Communication Subsystem

IEEE Parallel & Distributed Technology: Systems & Technology
Heterogeneous System Performance Prediction and Analysis Using PS

IEEE Concurrency
Estimating and Optimizing Performance for Parallel Programs

Computer
Exploiting Parallelism Through Directives on the Nano-Threads Programming Model

LCPC '97 Proceedings of the 10th International Workshop on Languages and Compilers for Parallel Computing
DiP: A Parallel Program Development Environment

Euro-Par '96 Proceedings of the Second International Euro-Par Conference on Parallel Processing-Volume II

Load balancing using dynamic cache allocation

Proceedings of the 7th ACM international conference on Computing frontiers
System for transport protocol evaluation with automatic calculation of statistical accuracy and distributed and parallel execution

Proceedings of the 4th International ICST Conference on Simulation Tools and Techniques

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper discusses the issues related to the accuracy of performance prediction tools for message passing programs. We present the results of two sets of experiments to quantify the effect of the instrumentation overhead and variance in the accuracy of Dimemas. The results show that this performance prediction tool can be used with a high level of confidence as the effect of instrumentation overhead on the predicted performance is minimal. We also show that it is possible to carry out instrumentation runs in highly loaded multi-user environments and still be able to accurately analyze the performance of the application as if it had run alone.