Multicasting to multiple groups over broadcast channels

Multicasting is a communication mode in which a given source communicates with a subset of the entire network user population. Previous work in this area concentrated on the multicast problem of a single source that always communicates with the same destination group. In this paper we investigate a more natural case of multicast communication where a single source communicates with several different destination groups. Specially, we focus on the design and analysis of multicast data link protocols for this environment. Straightforward implementations of such protocols are inappropriate in the case of a large destination population, as a source will have to store a large amount of state information even if it maintains only a single variable per destination. In most typical applications, though. The total destination population is large, the number of destinations that any given source is in conversation with, is typically small. We propose a framework for adapting protocols so that memory requirement does not grow with the total destination population but depends upon the number of destinations actually in communication with the source. The savings in memory are achieved by slightly increasing the amount of communication. We address the performance of such a protocol in an environment of a broadcast channel. We analyze several strategies and control techniques and demonstrate the tradeoff between throughput and the amount of memory.<
>