YQiao-SOSP03-WIP

Abstract

Elders Know Best: Lifespan-Based Ideas in P2P Systems Yi Qiao and Fabi ´an E. Bustamante I. INTRODUCTION The transiency of peer population and its implications on peer-to-peer (P2P) applications are increasingly calling the attention of the research community. As undesirable as un- avoidable,peerstransiencycouldnegatemanyoftheappealing features of the P2P approach. We are exploring new P2P protocols and strategies that, by considering peers’ lifespan a key attribute, can greatly boost thestability,efficiencyandscalabilityofthesesystems.Aspart of our work, we performed a thorough study of peers’ lifespan in a current and widely deployed P2P network. Through active probing of over 500,000 peers, we found that peer lifespan distribution can be modeled by a Pareto distribution of the form ¸Tk(k < 0) which, in this context, means that a peer’s expected remaining lifetime is directly proportional to its current age. II. PEERLIFESPAN AND P2P PROTOCOLS This last observation is the basis for a number of new protocols, first proposed in [1], for unstructured and loosely structured (ultra-peers) systems. In most P2P protocols there are a number of instances during a peer’s life where it must choose among “acquaintances”, such as when deciding whom to “befriend” (i.e. establishing a open connection) and when responding to a thir party’s request for recommendation. We have shown elsewhere [1], [2] that even simple lifespan- based P2P protocols can offer significant advantages, not only reduc