Abstract
This chapter describes state-of-the-art techniques in rare event simulation for queuing systems, the rare events under consideration being overflow probabilities, probabilities of extremely long delays, etc. We first consider a number of generic examples (and counterexamples) that are very useful in the queuing context. Then we systematically assess importance sampling for the cases of light-tailed input (where large-deviations arguments play a crucial role) and heavy-tailed input (where the change of measure is typically state-dependent). Other issues dealt with are: results under the many-sources scaling, estimation of the tail of the sojourn time distribution for processor-sharing queues, tandem and intree networks, and loss networks.
Authors
José Blanchet, Michel Mandjes
Publication date
2009/3/20
Journal
Rare Event Simulation Using Monte Carlo Methods
Pages
87-124
Publisher
John Wiley & Sons, Ltd
