ISP-Friendly Peer Matching Algorithms

In peer-to-peer (P2P) systems, a receiving peer needs to be matched with multiple sending peers, because peers have limited capacity and reliability. We study the following peer-matching problem: given a set of potential senders for an object requested by a receiver, find the subset of senders that will minimize the load on the backbone network and achieve the best performance for the receiver. We propose two new algorithms to solve this problem: ISPfriendly and network-based. Our ISP-friendly algorithm leverages information about the Autonomous System (AS) graph and different types of relationship among ASes, and finds the closest senders in terms of AS proximity, i.e., the number of AS hops that the data from senders have to cross to reach the receiver. Our network-based algorithm finds the closest senders in terms of network proximity, i.e., number of IP hops. Our initial results indicate that a significant reduction in the number of ASes that P2P traffic has to traverse from senders to receivers is achievable.