ABSTRACT
Analytical performance results are used in the proper engineering decisions, such as buffer and link dimensioning, in the deployment of computer networks including Asynchronous Transfer Mode networks. This chapter presents a survey of some of the most important results in the area of source characterization. It introduces the multiple definitions and quantitative measures of burstiness as well as the concept of the ON/OFF model. The chapter reviews the phase processes for building Markovian models of bursty traffic, and provides the approximation of the superposition of an arbitrary number of ON/OFF sources by a Markov Modulated Poisson Processes. It addresses the issue of representing more complex forms of correlation, which is particularly suited for video traffic, by Auto-Regressive Models. The chapter examines schemes for deriving quantized rates to enhance the tractability of models related to video traffic. It highlights the results on Long Range Dependence and self-similar models for both data and video traffic.
