A novel fault diagnosis algorithm for K-connected distributed clusters

摘要

In this paper, we propose an on-line two phase (TPD) fault diagnosis algorithm for distributed clusters that follows an arbitrary network topology with connectivity k. Intermediate nodes communicate heartbeat messages between different source destination pairs. The algorithm addresses a realistic fault model considering crash and value faults in the cluster nodes. The algorithm is shown to produce a time complexity of O(l) and message complexity of O(n. e) respectively. The algorithm has been simulated using discrete event simulation techniques and the results show that the algorithm is feasible for large distributed clusters.