An implementation of reliable broadcast using an unreliable multicast facility

The authors consider the problem of reliable broadcast in a point-to-point asynchronous network. Such a network consists of host computers and a communication subnetwork. The latter, in turn, is a collection of switches (special-purpose computers that have the ability to store and forward messages), interconnected by point-to-point bidirectional communication links. The subnetwork is unreliable, i.e. a message is never guaranteed to be delivered in a finite interval of time. Broadcast is to be implemented among the hosts of the network. The basic mode of operation for the proposed broadcast mechanism is the posting by the source of multiply addressed data messages to be delivered to all participating hosts, combined with a distributed redelivery algorithm that ensures that those messages that were lost in transit do eventually arrive at all intended destinations. At the heart of the algorithm is a set of structures called priority lists. The use of the priority lists to encode topological information and the performance of the algorithm are discussed