Concurrency preserving partitioning algorithm for parallel simulation of hierarchical, modular discrete event models

This paper presents a concurrency preserving partitioning algorithm for the optimistic parallel simulation of hierarchical, modular discrete event models. The proposed algorithm pursues the following three goals to achieve the overall objective of the minimum simulation time: (1) balance the computational loads of partitions, (2) maximize the parallel execution of independent models, and (3) minimize inter-processor communication. To estimate the parallelism inherent in models, the proposed algorithm utilizes the inherent hierarchical structural information of DEVS models. This paper describes how the proposed algorithm works through an example partitioning process.