A Gossip-Based Reliable Multicast for Large-Scale High-Throughput Applications

Gr oup-based reliable multicast is an important building block for distributed applications. For large systems, however, traditional appr oaches do not scale well due to centr alize drecovery mechanisms and excessive message overhead. In this paper, we present a reliable probabilistic multicast, rpbcast, that is a hybrid of the centr alize dand gossip-b ased approaches. In p articular,rpbcast extends previous work by supporting high packet r ates and many active senders. Rpbcast uses gossip as the primary retransmission mechanism and only contacts loggers if gossips fail. Large gr oups of active senders are supported using negative gossip that sp eci es those messages a receiver is missing instead of those messages it received. Moreover, we show that negative gossip allows pull based recovery and converges faster than push based recovery. Rpbcast also applies hashing techniques to reduc e message overhead and approximate group membership for garbage collection. We describe the key features of rpbcast and present simulation results.