Still Alive: Extending Keep-Alive Intervals in P2P Overlay Networks

摘要

Detecting node failures within Peer-to-Peer networks is an inherent trade-off between timely de-tection and consuming bandwidth on network main-tenance. In the absence of user-driven messages, the majority of P2P networks rely upon the exchange of periodic keep-alive messages to maintain connections and network topology. We investigate three novel al-gorithms which prioritise keep-alive messages to nodes that are more likely to have failed. In doing so, these algorithms significantly reduce the expected delay be-tween failures occurring and their subsequent detec-tion in comparison to the standard approach, whilst consuming similar levels of bandwidth. Our algorithms build upon several studies that have shown that older peers are more likely to remain in the network than their short-lived counterparts. Each of our algorithms increase the interval between successive keep-alive messages as peers age in the system, based upon the distribution of peer session times and the current age of peers. We extensively describe the details of each algorithm, before comparing them to the standard pe-riodic approach using simulations based upon mea-sured network data. Furthermore, we show that these algorithms are complimentary to existing gossip-based mechanisms and investigate alternate methods of as-certaining a node's age so that our algorithms can be robustly deployed in untrustworthy environments.