Causal Order Delivery in a Multicast Environment: An Improved Algorithm

Abstract

Causal order delivery of messages is required for many distributed applications. One of the problems with causal order delivery algorithms is the need to attach the dependency information with each message to ensure the causal ordering of delivery. This introduces a large amount of overhead and limits the scalability of the algorithm. To reduce the amount of dependency information, an algorithm was proposed by Prakash to attach only direct dependency information to the messages. The algorithm is optimal in a broadcast environment. However, in a multicast environment, it fails to eliminate indirect dependencies in some cases. In this paper, an improved causal order delivery algorithm that eliminates indirect dependencies in a broadcast as well as a multicast environment is proposed. The new algorithm is analyzed against Prakash's algorithm in terms of optimality on eliminating indirect dependencies. Simulation studies were also carried out to compare the performance of the two algorithms. The results show that the new algorithm incurs less communication overhead than Prakash's algorithm in all the cases.