Biased initial distribution for simulation of queues with a superposition of periodic and bursty sources

摘要

In this paper, we focus on queues with a superposition of periodic and bursty sources where the initial states of the sources are randomly and independently selected. The dynamics of the queue are described as a reducible Markov chain. For such queues, we consider the tail probability of the queue length as the performance measure and try to estimate it by simulations. To do simulation efficiently, we develop a sampling technique which replaces the initial distribution of the reducible Markov chain with a new biased one. Simulation results show that the sampling technique is useful to reduce the variance of the estimates.