Scalable and efficient update dissemination for distributed interactive applications

Abstract

Distributed interactive applications such as multiplayer games will become increasingly popular in wide area distributed systems. To provide the response time desired by users despite high and unpredictable communication latency in such systems, shared objects will be replicated or cached by clients that participate in the applications. Any updates to the shared objects will have to be disseminated to clients that actually use the objects to maintain consistency. We address the problem of efficient and scalable update dissemination in an environment where client interests can change dynamically and the number of multicast channels available for update dissemination is limited. We present a heuristic based algorithm that can group objects and clients in a way that it handles limited bandwidth resources. We show that our algorithm can produce better results than several algorithms that have been developed in the past for update dissemination.