An efficient and reliable group multicast protocol

Maintaining a consistent state is crucial for cooperative work in autonomous and decentralized systems (ADS). This is achieved through reliable communication protocols in a group of autonomous components. Many existing protocols are quite expensive and not very suitable for the ADS. The paper presents a logical token ring based group multicast protocol for the service of decentralized action manage systems. Unlike previous token based algorithms, we assign a virtual identity to each of the processes executing in the autonomous components in an increasing order so as to form a logical ring. The protocol ensures total ordering and atomicity of multicast messages in the presence of the component failures. The total order is achieved by placing a sequence number in each multicast message. A token process implicitly multicasts one such ordered message and transfers the token to the next process on the ring, hence, reducing the synchronous overhead and preventing possible token lose problem. With a minimal number of control messages, the protocol guarantees that all nonfaulty processes belonging to the same group will observe a consistent view of ordered events affecting all cooperative components in the group. The efficient recovery algorithms can handle the member failures, dynamic groups, group partitions and merges