Internet Traffic and Queuing Analysis
speaker: Mine Caglar (Koc University)abstract: The Internet brings together heterogeneous interconnecting domains. Accurate queueing models of IP traffic over the entire network domain are necessary. Most of the current queuing analysis still rely on classical Poisson distributed traffic whereas the Internet traffic has been shown to be self-similar and long-range dependent. Recently, efforts have gone to develop accurate queuing models in the presence of self-similarity. This lecture starts with a survey of the Internet traffic models such as onoff processes, Markov modulated Poisson processes and infinite source Poisson processes, in particular, the telecom process. Queuing analysis with such models will be described as current research area. We present a novel analytical model which takes into account multiple classes of self-similar traffic based on a GM1 queueing system with priority. On the other hand, many recent applications in the Internet such as file sharing and content distribution are based on epidemic protocols and are implemented on a larger network. Recent research on the properties of traffic generated by epidemic protocols and their delay analysis will also be presented.
<< Go back