Dr. Daniel Zappala The goal of the Internet Research Laboratory, under the direction of Dr. Daniel Zappala, is to pursue research that promises to improve the Internet and the way in which we use it to communicate.
Dr. Zappala and his students explore ways in which the reach of the internet can be extended, the infrastructure of the Internet can be improved, and the services offered by the Internet can be enhanced.
One area that the lab researches is wireless networking. Wireless networking is becoming increasingly important in today's society. People typically move around their environment throughout the day, be it in their office building or in their city, which makes desk-bound wired connectivity an inconvenience. In the lab, students are designing ways to improve the speed and efficiency of wireless mesh networks, creating novel routing and transport protocols. To test the ideas, the students are currently building a wireless mesh network in the computer science building, where they will experiment with new networks that incorporate multiple wireless transmitters on each router, along with network coding techniques that can dramatically improve network efficiency.
Another research area in the lab is peer-to-peer networking. Peer-to-peer networking has become a popular method for downloading large files, such as songs, movies, and software. When using a program such as BitTorrent, a user who wants a large file can download parts of the file from a network of peers on the Internet. To make this system work when there aren't many peers available, the owner of the content typically runs a primary seed that has the entire file and is wiling to serve blocks to anyone who needs them. Any other users who have already downloaded the file may also become seeds, though there is little incentive for them to help out in this way.
One of the problems with systems like BitTorrent is that the primary seed typically bears much of the burden of transferring the file to users. Because the primary peer has the entire file, is usually available all the time, and generally has good bandwidth available, it can serve most of the content. This can make it difficult to host popular content, as BitTorrent will tend to use up all available bandwidth.
Students in the lab are designing a new BitTorrent seed that monitors the content available in the network of peers. If it determines that some blocks of a file are readily available from the peers, it won't serve them itself, thus reducing its burden. This requires sophisticated monitoring and some intelligence to decide which blocks to serve.
The lab is also looking at online multi-player games. Current multi-player games require enormous server bandwidth. As a result, many games fail to survive if they do not attract enough subscribers. Games require large amounts of bandwidth because they use a client-server architecture, in which the game must deliver every move from every player to all the other players. Some researchers have developed peer-to-peer architectures, but these do not provide enough control to eliminate cheating and ensure smooth game play.
In response to the existing problems, students in the lab are creating a hybrid game architecture that combines the best of client-server and peer-to-peer architectures.
Using a client-server component allows the game developer to charge for access to the game, in return for eliminating cheating and providing the infrastructure required for law-latency gaming. Using peer-to-peer networking reduces many of the costs associated with hosting a game, by using peers to pass along positional moves that do not affect critical game state.