科研成果详情

This paper presents an efficient and novel reliable group multicast protocol for asynchronous distributed systems. Based on a logical-clock loken ring approach, the protocol guarantees the total ordering and atomicity of multicast messages and is highly efficient. The recovery algorithms of the protocol can handle communication faults, process crash failures and network partition. We describe the detailed design, a correctness proof and the performance analysis of the protocol. For a group of n processes, the protocol requires only up to (n - 1) control messages to reach a consistent view of the group in a fault-free case. The protocol has been implemented on a local area network of workstations. The performance of the implementation is also presented. © 1998 Elsevier Science B.V.