ABSTRACT
Models of multimedia trafc offered to the network or to a component of the network are critical in providing high quality of service (QoS). Trafc models are used as the input to analytical or simulation studies of resource allocation strategies. We may view trafc at the application or packet level, where an application-level view may simply describe the proled trafc as “a videoconference between three parties,” while the packet-level view is based on a stochastic model that mimics the arrival process of packets associated with this application reasonably well. Clearly, in order to quantify trafc, packet-level representation of applications will be used. An important feature of multimedia trafc at the packet level having a signicant impact on performance is trafc correlation. The complexity of trafc in a multimedia network is a natural consequence of integrating, over a single communication channel, a diverse range of trafc sources such as video, voice, and data that signicantly differ in their trafc patterns as well as their performance requirements. Specically, “bursty” trafc patterns generated by data sources and variable bit rate (VBR) real-time applications such as compressed video and audio tend to exhibit certain degrees of correlation between arrivals, and show LRD in time (Sahinoglu and Tekinay 1999).
