Modeling And Characteristic Analysis Of Video Traffic On ATM

In B-ISDN based on ATM a variety of traffic data is divided into cells of constant-length and carries out various processions such as transmission and exchange on networks. In the same network a variety of traffic types in integrated fashion is the advantage that ATM brings on, but its cost is that is must be used to handle with various bursty traffic types on networks. A burst is stochastic properties of various traffic cell streams. So how to characterize the performance of cell streams on networks is a problem. It is necessary to model to reflect the statistical properties of video sources accurately in order to study video streams transmission.We study video models in the context of statistical multiplexing. In this paper, we discuss three types of video source models: a autoregressive model, a Markov-modulated fluid flow model, and a Markov-modulated Poisson process model (MMPP).The autoregressive model is useful in providing a simple representation of coded video source statistics for simulations of network behavior with video traffic applied. However because the autoregressive model for video cannot readily be applied to the analytical determination of buffer size or the equivalent delay determinations in broadband, cell-, or packet-based networks, Other models such as the fluid flow model or MMPP models must be introduced to analyze the impact of video traffic on networks. The approach to match first and second moments of composite Markov chain model and the measured statistic is used to the first models to analyze the characteristic of loss probability, while the histogram approximation is used to MMPP model.Nevertheless none model can adapt to all the regions of the buffer size. So we analysis the small and large buffer region, using the terminology "cell" and "burst" region, respectively. The histogram approximation of MMPP model is used to cell region, and the fluid flow model is used to burst region. We shall use adoubly-stochastic process to model the traffic arrivals. We will introduce a simple linked technique that enable us to combine the curves obtained in the two region quite simply, thereby obtaining a smooth loss probability curve that shows the variation with buffer size.