The impact of VBR MPEG video traffic on ATM multiplexer performance and its evaluation

The Broadband Integrated Services Digital network (B-ISDN) will be based on the Asynchronous Transfer Mode (ATM) that allows statistical multiplexing of variable bit rate (VBR) sources to make efficient use of the network resources. The ability of the network elements to apply statistical multiplexing while providing a guaranteed quality of service (QoS) to the users is depending on the statistical properties of the ATM traffic streams. Thus it is essential to investigate the multiplexing behaviour of different ATM traffic streams with respect to the ATM QoS parameters cell loss, cell delay and cell delay variation using realistic source models. Since compressed video data will have a major share in future broadband traffic and the MPEG coding algorithm up to now has achieved wide popularity, we mill concentrate on VER MPEG video data streams. The flexible hierarchical source model presented in [Enss 95], that is capable of modelling VBR MPEG video and other sources that exhibit long range dependence (LRD) and/or periodic properties, is used to investigate in detail the multiplexing behaviour of such sources. The focus of this paper is twofold. First, the consequences of the usage of LRD traffic streams in discrete event simulations in terms of the necessary simulation duration and the convergence to a steady state behaviour are investigated. Therefore, we will investigate how the convergence rate of the confidence interval size of the empirical mean value of discrete stochastic processes depends on the Hurst parameter. Second, it will be shown that both, the LRD and periodic traffic components have a significant impact on the multiplexer performance in terms of the cell loss probability and the cell delay but mainly act on different time scales. The actually noticeable influence on the quality of service (QoS), that the VER MPEG video data streams perceive, depends on the multiplexer buffer size and the stochastic properties of the video data, streams at the different time scales.