Simulation methods of queues: An overview (Q1116204)

The paper gives an overview of those aspects of simulation methodology that exploit the stochastic structure of queueing systems. The main tool used for describing is a generalized semi-Markov process. The following main topics are covered: Basic methodology for the analysis of simulation output with estimation of transient and steady-state quantities; diffusion approximations for queues in heavy traffic with insight into the simulation run-length required to estimate steady-state quantities as the traffic intensity tends to unity; the ideas of variance reduction and efficiency improvement which permit the simulator to improve upon conventional estimation methods by incorporating knowledge of the stochastic system under consideration; seven techniques exploiting queueing-related stochastic structure. Finally, two methods (based on perturbation analysis and likelihood ratios) for estimating gradients of queueing performance parameters are discussed.