Tuning a Distributed Simulator Using an Evolutionary Algorithm

Use of an evolutionary algorithm to tune constant values that help control the performance of a distributed discrete event simulator is investigated. Simulator constants are parameters that are often set empirically; the simulator is then used with those constant values indefinitely or for some set of simulations. However, when the simulator is ported to a new environment, the constant values may have to be reset. Using an evolutionary algorithm to guide the setting of these constants can expedite the recalibration process. The empirical results demonstrate the usefulness of this approach.