A Single Phase Protocol for Total andCausal Ordering of Group Operations inDistributed

The total and causal ordering of group operations in distributed systems is at the core of solutions to many problems in distributed computing. Existing approaches are based on either a centralized coordinator to assign a sequence number for each operation or on a distributed two-phase protocol to determine the total ordering. This paper proposes a distributed single-phase protocol which determines the total ordering of group operations at the moment when operations are generated, without the need for a second phase of determining the sequence number. Therefore, the latency time can be kept in the range of 0; d], where d is the maximum message propagation delay between any pair of sites in the system. Furthermore, in contrast to the arbitrary total ordering imposed by two-phase protocols, the total ordering imposed by the proposed protocol is consistent with the causal ordering among group operations, which is required by many distributed applications. With an optimized acknowledging scheme integrated , the number of acknowledgment (overhead) messages involved in each group operation can be kept in the range of 0; N ? 1] in multicasting networks, or 0; (N ? 1) 2 ] in point-to-point networks. Moreover, the proposed protocol is self-adaptive to the group operation traac: the heavier the group operation communications traac, the shorter the average latency time and the less overhead messages in the system. Both theoretical analysis and software simulation have been conducted to verify the correctness of the proposal protocol.