That Discrete Event Dynamic Systems (DEDS) constitute an important topic of systems theory and operation research is no longer in question, and while examples of DEDS ranging from airports to communication networks to logistics, service, and manufacturing systems abound in our daily lives, the fact remains that performance of only a small percentage of such systems can be analyzed in closed form. Simulation remains the general purpose performance evaluation tool of choice. However, simulation is notoriously time consuming, particularly if parametric system performance studies are involved. Furthermore, traditional performance optimization algorithms are inherently iterative and sequential.Attempts to parallelize or distribute the computation on modern massively parallel computers have not been successful. The result is that simulation in practice has not been used effectively as a design tool and often as a last resort to validate ad hoc designs. Researchers advocate a fundamental mind set change on the way discrete event simulation languages are constructed. The approach parallelizes the performance evaluation experiments instead of the algorithms. By taking advantages of recent theoretical developments in perforrnance sensitivity analysis and stochastic optimization techniques, the creation of a general purpose discrete event simulation language that has the potential of several orders of magnitude improvement in the efficiency of performance evaluation of discrete event systems is visualized.Commercial Applications:The proposed R&D will result in a user-friendly software package for general discrete event simulation with several orders of magnitude speed improvement and enable the solution of performance evaluation and optimization problems previously thought to be beyond reach.
