Scalable Reliable Multicast Protocols have been the subject of much research in recent years. We propose a new protocol that groups receivers for error recovery into fixed-size groups, thus reducing their processing requirement to O(1). This means that, regardless of the size of the multicast session, the processing requirements for receivers remains constant. The processing requirement on the sender is tunable according to its processing capabilities and/or the expected multicast session size. The concept of Local Recovery is then applied which further improves the processing requirements, especially on the sender. The processing requirements on both the sender and the receiver are analytically studied, and compared with another same-class protocol.
Elias G. Khalaf, S. Sitharama Iyengar