A Failure Detection Service Based on Epidemic Dissemination for Peer-to-Peer Networks

Failure detectors were first proposed as an abstraction that makes it possible to solve consensus in asynchronous systems. A failure detector is a distributed oracle that provides information about the state of processes of a distributed system. This work presents a failure detection service based on a gossip strategy. The service was implemented on the JXTA platform. A simulator was also implemented so the detector could be evaluated for a larger number of processes. Experimental results show that increasing the frequency in which gossip messages are sent gives better results than increasing the fanout. Other results include CPU and memory usage, fault and recovery detection time, mistake rate and how the detector performs when used in a simple election algorithm.