NeTS:Small:Dynamic Coupling and Flow-Level Performance in Data Networks: From Theory to Practice

点击翻译按钮获取中文摘要

This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5).The performance of network users' best effort flows, e.g., file transfer delays and web browsing responsiveness, depends on the resources they are allocated over time. When varying traffic loads share these resources and/or wireless nodes' transmission capacities depend on each other through interference, the allocated resources and thus flows' performance are coupled through the traffic and interference dynamics. This project investigates such performance coupling in data networks and, in turn, how it affects protocol and network design. This is a significant problem, as most data networks share these characteristics, and we currently have no robust tools to effectively predict and thus optimize performance. Expected results include new theory, approximations and performance bounds enabling the analysis of such systems which are also applicable to other domains. Expected results also include using these tools to investigate: (1) the benefits of, and capacity allocation, in networks supporting multipath transport and/or shared wireless access networks; and (2), the development of improved protocols and algorithms that are sensitive to such performance coupling, e.g., preliminary work suggests the state-of-the-art base station association policies can be substantially improved by factoring the performance coupling resulting form interference. The work's impact will be assured through broad dissemination in the research community, and by leveraging industry partners in evaluating practical applications and technology transfer opportunities.