Inference and Performance Problems Related to High Variability Phenomena in Measured Data Network Traffic

点击翻译按钮获取中文摘要

A cooperative arrangement between Cornell University and AT&T Labs-Research allows Cornell's Professor Sidney Resnick to visit AT&T Labs-Research to collaborate with Dr. Walter Willinger and others at AT&T Labs-Research on problems related to (i) providing an in-depth understanding of the dynamic nature of today's data network traffic and (ii) exploiting the newly-gained insights for the economic design and effective and efficient management of modern high-speed communications networks. The work focuses on the dual themes of how to construct and fit models that can account for empirically observed data network traffic characteristics such as high-variability, heavy-tails, long-range dependence, self-similar behavior. Phenomena exhibiting heavy tails and/or long-range dependence, although departing in marked fashion from classical assumptions of Gaussian distributions, finite variances and short-range interactions, have been frequently noted in a broad array of fields such as insurance, economics and finance. These phenomenahave recently attracted renewed interest due to their ubiquitous presence in traffic measurements from today's data networks and especially because they often play havoc with established network and traffic engineering methodologies that are based on conventional (i.e., telephony-based) wisdom. As a result, this networking application opens up a variety of new and fundamental research topics at the intersection of applied probability and statistics and provides unique opportunities for close collaborations between mathematicians and engineers to work on technically challenging problems that are, at the same time, highly relevant in practice. This GOALI project is jointly supported by the MPS Office of Multidisciplinary Activities (OMA) and the Division of Mathematical Sciences (DMS).