Hypergraph partitioning for the parallel computation of continuous Petri nets

Abstract

Continuous Petri net can be used for performance analysis or static analysis. The analysis is based on solving the associated ordinary differential equations. However, large equation groups will give us overhead computing. To solve this issue, this paper presents a method to compute these differential equations in parallel. We first map the Petri net to a hypergraph, and then partition the hypergraph with minimal inter-processor communication and good load balance; Based on the partition result, we divide the differential equations into several blocks; Finally we design parallel computing algorithm to compute these equations. Software hMETIS and SUNDIALS have been used to partition the hypergraph and to support the parallel computing, respectively. Gas Station problem and Dining Philosopher problem have been used to demonstrate the benefit of our method. © 2011 Springer-Verlag Berlin Heidelberg.