Ferenci, Steve
Fujimoto, Richard
Ammar, Mostafa
Perumulla, Kalyan
Riley, George
Abstract
A technique called updateable simulations is proposed to reduce the
time to complete multiple executions of a discrete event simulation
program. This technique updates the results of a prior simulation
run rather than reexecute the entire simulation to take in to account
variations in the underlying simulation model. A framework for
creating updateable simulations is presented. This framework is
applied to the problem of simulating a set of cascaded ATM multiplexers
and a network of ATM switches. Performance measurements of sequential
and parallel implementations of these simulations on a shared memory
multiprocessor are presented, demonstrating that updateable simulations
can yield substantial reductions in the time required to complete
multiple simulation runs if there is much similarity among the runs.
|
Keywords
shared computation
rerun
statistics
event reuse
Notes
Related Papers
Bibtex
@inproceedings{ferenci.fujimoto_update02,
author ="teve Ferenci, Richard Fujimoto, Mostafa Ammar, Kalyan Perumulla, George Riley",
title ="Updateable Simulations",
pages ="",
volume ="",
booktitle ="Proceedings of the 2002 Parallel and Distributed Simulation Conference",
year ="2002",
month ="May",
documentURL ="http://www.cc.gatech.edu/fac/Mostafa.Ammar/NEW.html"
}