Home // ICSEA 2011, The Sixth International Conference on Software Engineering Advances // View article

Performance Simulation of a System's Parallelization

Authors:
Markus Meyer
Helge Janicke
Peter Trapp
Christian Facchi
Marcel Busch

Keywords: software performance optimization; performance simulation; parallelization; dynamic performance stubs

Abstract:
A new approach to simulate the parallelization of a software function is presented in this paper. The parallelization’s effects onto the system’s performance prior to a costly realization of the parallelization are evaluated, leading to a more gain-oriented performance optimization. The presented approach defines a methodology to transform a single-threaded software function into a multi-threaded simulation. CPU stubs, simulating both, the performance and functional behavior, are applied to simulate the expected time slices. The proposed technique can estimate the expected performance gain for the whole system. A proof of concept is used to evaluate the proposed methodology and the simulation results are compared to a known parallel implementation of the algorithm. Initial results show our approach can be used to simulate the performance behavior of a parallelized system with high accuracy. In addition, the number of threads that result in the highest performance gain of the system is determined.

Pages: 233 to 238

Copyright: Copyright (c) IARIA, 2011

Publication date: October 23, 2011

Published in: conference

ISSN: 2308-4235

ISBN: 978-1-61208-165-6

Location: Barcelona, Spain

Dates: from October 23, 2011 to October 29, 2011