This project is focused on the development, analysis, and empirical testing of automated methods for load-balancing parallelized discrete-event simulations that use certain "conservative" synchronization protocols, on distributed memory parallel architectures. Both static and dynamic methods of load balancing will be investigated. The algorithms will be evaluated by their performance on large simulation models of queueing networks, Petri nets, and communication networks, executed on an Intel iPSC/2 distributed memory multiprocessor. Automated techniques for load balancing parallel simulations is essential if parallel simulation is to move out of research labs, and become widely used.
