Fault-Tolerant Distributed Publish/Subscribe Using Self-Stabilization

We propose a new fault-tolerant (self-stabilizing) distributed framework for the publish/subscribe scheme in a P2P network, where the participating nodes (peers) only need the state information of its immediate neighbors. The proposed system is guaranteed to recover from an unlimited number of transient faults (e.g., state corruption). We provide a complexity analysis of the recovery (stabilization) time of the system after arbitrary failures in the network. Also, we show that the system always remains alive and starvation free. Nodes can form multiple logical layers depending on the topics of interest; the framework provides flexibility to dynamically configure the layers to provide time space tradeoff.