Games and networks

Abstract

Modern networks are the product of, and arena for, the complex interactions between selfish entities. This talk surveys recent work (with Alex Fabrikant, Eli Maneva, Milena Mihail, Amin Saberi, and Scott Shenker) on various instances in which the theory of games offers interesting insights to networks. We study the Nash equilibria of a simple and novel network creation game in which nodes/players add edges, at a cost, to improve communication delays. We point out that the heavy tails in the degree distribution of the Internet topology can be the result of a trade-off between connection costs and quality of service for each arriving node. We study an interesting class of games called network congestion games, and prove positive and negative complexity results on the problem of computing pure Nash equilibria in such games. And we show that shortest path auctions, which are known to involve huge overpaymentsin the worst case, are “frugal” in expectation in several random graph models appropriate for the Internet.