SRINIVASAN SESHAN
Associate Professor, Computer Science
www

My primary interests are in the broad areas of network protocols and distributed network applications. In the past, I have worked on topics such as transport/routing protocol interactions with wireless networks, sensor networking, fast protocol stack implementations, RAID system design, performance prediction for Internet transfers, firewall design, and improvements to the TCP protocol. The following three projects are examples of my current research efforts.

Chaotic Wireless Networks. Today, one can find 802.11 hardware and other personal wireless technology employed at homes, shopping malls, coffee shops and airports. Present-day wireless network deployments bear two important properties: 1) they are unplanned, with most access points (APs) deployed by users in a spontaneous manner, resulting in highly variable AP densities; and 2) they are unmanaged, since manually configuring and managing a wireless network is very complicated. We refer to such wireless deployments as being chaotic. Unfortunately, as wireless networks become more common and more densely packed, these chaotic deployments will suffer from serious contention, poor performance, and security problems. As part of this project, we are developing new protocols and algorithms to address these problems.

Measurement-Based Analysis of Internet Performance. This research includes an extensive measurement study to discover, classify and characterize non-access bottleneck links in terms of their location, latency and available capacity. Using a combination of simulation and analysis, our research shows that the structure, operation and evolution of the Internet together have an undesirable impact on the growth of congestion at particular hot-spots. In addition, we are developing novel routing-based mechanisms that end-networks can use to improve the performance and reliability of their Internet connectivity.

Large-Scale Multiplayer Games. Despite their commercial importance, networked games have been largely overlooked by the research community. Our goal is to enable the implementation of distributed game servers that support 100000's of players in a large-persistent game world while operating on either a federated collection of servers or a peer-to-peer collection of clients. The current core of our design is a distributed content-based publish-subscribe system called Mercury. Previous publish-subscribe systems have either offered a very restrictive query language or had poor scalability properties. The Mercury system is unique in that it supports range queries over multiple attributes as its subscription language, and yet it provides a scalable and efficient lookup protocol. This makes it ideally suited to the spatial 3-dimensional searches that are often used in the implementation of game servers.