Modeling traffic and statistical gains for multimedia applications

This work addresses the problems of traffic modeling, estimation of packet delay and statistical multiplexing gain for ATM networks. Particular attention is given to variable bit rate (VBR) traffic generated in videoconferencing and entertainment applications. The focus of this work is to understand the impact of traffic correlation and burstiness on quality of service. Traffic models are derived from statistical analyses of VBR traffic measurements. It is shown that correlation in video traffic can be modeled by autoregressive processes with time-varying parameters. The autoregressive model allows statistical multiplexing gains to be estimated. The magnitude of the gain is shown to vary between two to three for the applications considered