Reliable probabilistic communication in large-scale information dissemination systems
Anne-Marie Kermarrec, Laurent Massoulié, Ayalvadi J. Ganesh · 2000
Reliable group communication is important for large-scale distributed applications such as information dissemination systems. The challenging issue in this context remains scalability. The computation time and amount of data dedicated to the reliability mechanism should remain manageable as the number of nodes in a system grows, and no bottleneck should emerge. Probabilistic algorithms has proven their ability to fill this gap. In this paper, we present the theoretical analysis and evaluation of a scalable reliable group communication protocol for wide-area dissemination systems. The scalability of the protocol relies on its probabilistic flavor. The protocol provides probabilistic guarantee of delivery and thus makes do with a lightweight recovery protocol. A distributed membership service is described and simulation results show that the protocol exhibits very stable behavior in the presence of transient and/or permanent failures. 1