Some considerations in simulating an M/M/1 queue

In spite (or perhaps due) to its simplicity, the question of simulating the M/M/1 queue has attracted much interest. It has served as a benchmark as various properties related to the simulation, or to the simulated performance measures, are known for this queue. In this paper we report on some experience that we obtained from this experiment that we believe will be of importance in simulating many other queueing types as well as networks of queues. In particular, we present some results on improving precisions or on reducing simulation time by processing the simulation results with the bootstrap method, and by using the quantile method to obtain con dence intervals. As a secondary objective we have made a comparison of ns-2 and ns-3 for the particular case of the M/M/1 queue.