Coverage Analysis of Sequential Regenerative Simulation

TR-COSC 02/99Regenerative simulation (RS) is a method of stochastic steady-state simulation in which output data are collected and analysed within regenerative cycles (RCs). Since data collected during consecutive RCs are independent and identically distributed, there is no problem with the initial transient period in simulated processes, which is a perennial issue of concern in all other types of steady-state simulation. In this report, we address the issue of experimental analysis of the quality of sequential regenerative simulation in the sense of the coverage of the nal con dence intervals of mean values. The ultimate purpose of this study is to determine the best version of RS to be implemented in Akaroa2 [1], a fully automated controller of distributed stochastic simulation in computer network environments