Speaker: Sachin Katti, ICSI, University of California, Berkeley
Location: Warren Weaver Hall 1302
Date: March 27, 2009, 11:30 a.m.
Host: Michael Overton
Wireless is becoming the preferred mode of network access. The performance of wireless networks in practice, however, is hampered due to the harsh characteristics of the wireless medium: its shared broadcast nature, interference, and high error rate. Traditionally, network designers have viewed these characteristics as problematic, and tried to work around them. In this talk, I will show how we can turn these challenges into opportunities that we exploit to significantly improve performance.
To do so, we use a simple yet fundamental shift in network design. We allow routers to "mix" (i.e., code) packets' content before forwarding them. We built three systems, COPE, ANC and MIXIT, which exploit this network coding functionality via novel algorithms to provide large practical gains. In this talk, I will discuss COPE and ANC; COPE exploits wireless broadcast, while ANC exploits strategic interference to improve throughput.
This work bridges and contributes to two unrelated areas: network coding and wireless mesh network design. It lays down the algorithmic framework for using network coding in modern wireless networks, by designing algorithms which work with the common case of unicast flows in dynamic and unknown environments. It also provides the first implementation, deployment and experimental evaluation of network coding. For wireless mesh networks, it shows how the framework of network coding allows us to profitably harness the inherent wireless characteristics. This union ultimately allows us to deliver a several-fold increase in wireless throughput.
Sachin Katti is currently a postdoctoral scholar at U.C.Berkeley. He recently received his PhD in EECS from MIT in September, 2008. Before coming to MIT, he received his B.Tech. in Electrical Engineering from the Indian Institute of Technology, Bombay, in 2003. His dissertation research focuses on redesigning wireless networks with network coding as the central unifying design paradigm. The dissertation won the George Sprowls Award for Best Doctoral Dissertation in EECS at MIT and has been nominated for the ACM Doctoral Dissertation Award. His work on network coding was also awarded a MIT Deshpande Center Grant. His research interests are in networks, wireless communications, applied coding theory and security.
Refreshments will be offered starting 15 minutes prior to the scheduled start of the talk.