Abstract
Multimedia data traffic on the Internet has a highly self-similar structure. Self-similarity can be expressed by using the ?Hurst parameter?. In the literature there are three basic methods proposed to evaluate the Hurst parameter, respectively: R/S analysis, variance time analysis, and periodogram analysis. This paper analyzes and compares these techniques by using a set of MPEG videos. We propose the best method for two different tasks: traffic admission control and traffic shaping. We also present a detailed description of the algorithm for various situations. Finally, we analyze the behavior of the H parameter when various flows of self-similar traffic are multiplexed together. We also present a possible scheme for MPEG-video servers to maintain bounded the aggregated value of H