An approximation approach of the standard clock method for general discrete-event simulation

Abstract

The standard clock (SC) method is an efficient approach for discrete-event simulation. Its basic ideas are quite different from traditional approaches. SC has neither an event list nor event lifetimes, its applicability is limited, however, to exponential distributions and a class of nonexponential distributions. In this paper we provide an efficient approach to general distributions. Shifted exponential and hyperexponential distributions are used as second-order approximations to simulation input distributions. Numerical testing demonstrates that they serve as good approximations and preserve the advantages of SC. In addition, an nth order method is presented that provides arbitrarily good approximations. The idea of event insertion extends SC use to further applications and improves simulation efficiency on SIMD machines. >