Simple and efficient models for variable bit rate MPEG video traffic

For the performance analysis of ATM networks carrying Variable Bit Rate (VBR) MPEG video sequences there is a need for appropriate video traffic models. In this paper, we discuss Markov chain models for this specific type of traffic with respect to several statistical properties and their ability to predict cell losses at buffers of ATM multiplexers. We intentionally selected a simple model class to check for its appropriateness to model a complex traffic type. The results of this paper show that Markov chain models can be used efficiently for MPEG video traffic under the following conditions. The Group of Picturesʹ (GOP) generation process should be modeled rather than the frame process. If the performance measures depend on the correlation properties of the video traffic the scene process should also be considered in the course of the model development.