University of Southern California NetPD Lab of Department of Electrical Engineering The USC Andrew and Erna Viterbi School of Engineering USC
Site Home

Project Title: Transport and scheduling challenges in multi-hop wireless networks


Funding: Cisco Systems


PI: Konstantinos Psounis
Duration: Apr 2008 - ongoing



Description

We are interested in the fundamental performance limits of static multi-hop wireless (mesh) networks, an emerging networking architecture that provides community networking, distributed sensing, support for medicalapplications, and Internet access in locations like airports, convention centers, education facilities, and hospitals.

Mesh networks pose a challenge that does not exist in traditional wireline networks like the Internet: neighboring links interfere with each-other in complex ways that have detrimental effects on performance. One way to address this issue is to design medium access schemes that carefully schedule the various competing links. But, it is well know that such schedulers require a lot of computation and a centralized implementation. As a result, random access schedulers have become the de factostandard used in all deployments. Achieving efficiency over random access schedulers requires the transport protocol to be aware of the existing complex interference. With this in mind, we argue for the need to design new mechanisms for transporting data that view congestion and rate allocation as a neighborhood, not a single link affair.

Even with the best rate allocation scheme, one cannot exceed the performance limits imposed by interference and scheduling inefficiencies. An important open problem in the research community has been to characterize the fundamental performance limits of today's de facto standard random access scheduler, IEEE 802.11, in a multi-hop setting. We are working on an analytical methodology that gives a formal answer to this problem. Preliminary results indicate that Interestingly enough under commonly used topologies the performance loss of 802.11 over perfect schedulers is surprisingly low.

Publications