Simulation Estimation of Dynamic Properties inQueueing

The papers in the series are intended for internal use and are distributed by the author. Copies may be ordered from the library of Depertment of Computer Science. Abstract| In simulation studies the focus has traditionally been on time-independent characteristics of queueing systems , particularly on the steady-state mean. The systems, even if they are in stochastic equilibrium, have also time-dependent or dynamic properties which are of great practical importance. Typically, the dynamic properties are related to the variation of traac or load during some period of time. In this paper we propose a way to examine dynamic properties of queueing systems through simulation. Our approach is based on the following idea: The original traac rates are classiied into low and heavy values. The resulting binary sequence of indicators characterize the traac level of the system. The variation in the traac level is summarized through estimating the two histograms of level durations. The rationale of translating the problem into simultaneous estimation of two histograms is that the estimates for the cell proportions are based on cumulative statistics. Therefore , we can use the Bonferroni inequality and the existing methods for assessing the precisions of estimated proportions. Keywords| dynamic properties of queueing models, simulation estimation of proportions, simultaneous assessment of precisions.