Probabilistic multicast: generalizing the multicast paradigm to improve scalability

The article considers the multicast-to-some generalization of the traditional multicast paradigm. In this form of communication, a multicast group is still associated with each multicast message. It is, however, only necessary for any randomly determined subset of the multicast group to receive a multicast message. This subset is not defined a priori nor is it necessarily the same from one multicast to the next for the same group. It explores the use of this "probabilistic" multicast communication paradigm to address scalability issues in the response collection application which arises in some distributed computing applications. It also shows how this form of multicast can be used to reduce the discarding of responses to the multicast message due to buffer overflow at the multicast source.<<ETX>>