Marked Event Method In Discrete Event Simulation

In this paper, we develop a new method for sensitivity analysis of discrete event dynamic systems. The method is based on the observation that event lists under nominal system parameters can be shared by simulations with slightly perturbed parameters. Thus the simulation on different system configurations may be more efficient. The resulting algorithm is exact for Markovian systems, and approximate for non-Markovian systems. Experimental results are given to study the new algorithms.