Statistical analysis and simulation study of VBR broadcast-video traffic

Traffic generated by video sources is expected to be an important class of traffic in future networks, such as asynchronous transfer mode (ATM) networks. Statistical models characterizing traffic generated by variable-bit-rate (VBR) video coders are needed to address packet video per formance issues such as predicting cell loss rates in ATM networks. For video teleconference traffic, it has been shown that traffic is sufficiently accurately characterized by a multi-state Markov chain model that can be derived from three traffic parameters (mean, correlation, and variance). In this talk, we describe modeling results for sequences with frequent scene changes (the previously studied video teleconferences have very little scene variation) such as entertainment television, news, and sports. We analyze 11 long sequences. Our conclusion is that different sequences require different details regarding distributions of cells per frame and that a single model based on a few physically meaningful parameters and applicable to all sequences does not seem to be available.