Peer-to-peer networks use the Internet to unite an evolving set of
computers for a common goal. For example, one such goal is locating a
copy of a file or service. A necessary step toward solving this
problem is designing an overlay network to connect the participating
computers. Because the participating computers use the Internet
almost as a black-box, they have little or no knowledge of the global
structure they are joining. Nevertheless, we would like to construct
a network with low diameter, a simple routing scheme, and tolerance to
faulty and, ideally, even malicious nodes. For efficiency, the
overlay should use short paths and find nearby copies of objects when
possible. My talk will describe ways to solve these problems.