When distributed simulation is considered, several optimization problems can be identified, such as: partitioning, load balancing and communication among processes. Aiming at reducing the impact of these factors in the simulation running time, a genetic algorithm (AG-Partition) is proposed in order to provide partitions to distributed simulations aiming at reduces the interprocess communication. This paper presents the algorithm, its implementation and its efficiency and performance considering hypothetical models. The results indicate that its usage can provide several benefits depending on the factors used in the simulation.
|