A cluster-tree protocol for reliable multicasting

Abstract

Distributed simulations and conferences require a reliable multicast protocol that guarantees safe data delivery in a reasonably short time. Such high-quality service demands substantial network resources. As these applications grow in use, scalability becomes an important issue. One way to achieve scalability is through clustering. The overall load is distributed among clusters so that large multicast groups avoid overloading the network. We propose a protocol for reliable multicasting, based on a cluster structure. We prove that in the cluster, the leader is an ancestor of all cluster members with respect to the multicast routing tree. This relationship yields an efficient acknowledgement structure. We also describe an acknowledgement algorithm based on a pulsing mechanism and prove that it has constant latency for acknowledging data. Finally, we show that the protocol is scalable by proving that it generates a constant load for all nodes.