AESOP home

Publications

Distributed Solution of Large Markov Models Using Asynchronous Iterations and Graph Partitioning

Nicholas J. Dingle, William J. Knottenbelt

National Workshop Paper
18th UK Performance Engineering Workshop (UKPEW 2002), Glasgow July 10-11, 2002
June, 2002
pp.27–34
Abstract

We present a distributed approach for the steady state solution of large Markov models. We use asynchronous iterations to minimise processor idle time and graph partitioning techniques to minimise inter-processor communication. We demonstrate the scalability of our approach by solving a benchmark model for a number of large state space sizes on both a network of commodity PCs and a distributed memory parallel computer. The performance of our approach is contrasted with published results for an out-of-core solver.

PDF of full publication (203.6 kilobytes)
(need help viewing PDF files?)

Information from pubs.doc.ic.ac.uk/async-ukpew.