Efficient estimation of the probability of extreme cell delays

Abstract

We wish to estimate the probability of extreme cell delays for a particular virtual connection (VC) or virtual path (VP) in an Asynchronous Transfer Mode (ATM) network. Since extreme cell delays are rare events we cannot efficiently estimate the probability of extreme cell delays using crude Monte-Carlo methods. We increase (exponentially twist) the workload of a related Markov additive chain, which we call the encumbrance, and use the A-cycle sampling technique to analyse switches with finite buffers multiplexing cells from Bernoulli interrupted sources with constant bit rate during bursts.