Speaker: Tom Anderson
University of Washington
Friday, January 12, 2007
1:00 - 2:00 pm
EBU3b 1202

ABSTRACT
An emerging paradigm in the design of large scale distributed systems is to explicitly consider incentives as part of the design. In this talk, I outline two case studies of the role of incentives in distributed systems, in BGP and BitTorrent. In the case of BGP, we were able to show that detour routing inefficiencies in the Internet can be largely solved by better aligning incentives among ISPs. In the case of BitTorrent, a system explicitly designed to take incentives into account, we were able to show that its incentive mechanism can be easily subverted, that all users have an incentive to subvert the mechanism, and that the likely end result will be worse performance for all. In both cases I will outline solutions which are efficient, robust to incentives, and easy to implement.

BIO
Tom Anderson's research concerns the practical issues in constructing robust, secure, and efficient computer systems. He is a professor of Computer Science and Engineering at the University of Washington, where he also received his Ph.D. (1991) and M.S. (1989). His current research project is RIP (Re-architecting the Internet Protocols), a method for fixing the myriad problems with the Internet by re-thinking its design from first principles.

Tom's Website