Description
Peer-to-peer systems were designed from the outset to handle a high rate of churn: node joins, graceful leaves, and failures. Nevertheless, there is a price to church which may manifest itself as dropped messages, increased user-experienced latency, or increased bandwidth use, potentially limiting the scenarios in which a P2P system is deployable. Even in a reasonably stable managed infrastructure like PlanetLab, there can be a high rate of effective node failure due to nodes becoming extremely slow suddenly and unpredictably. In this paper, we study how to reduce the churn rate itself.