End-to-end performance in ATM networks

We are concerned with the end-to-end performance of an ATM network. The focus is on the characterization of the smoothing effect on the individual traffic streams as they traverse the network nodes. We develop an analytical model and a method of analysis for a generic node. For Markov modulated on-off streams, it is found that the on period gets lengthened as the stream passes through each node. The characterization of the lengthening effect allows a departure stream to be approximated by another Markov modulated on-off stream with modified parameters. The method permits a recursive computation of performance at all the nodes in a tandem network representing an end-to-end connection. The accuracy of the approximate analysis is validated by comparing it with simulation results.<>