FAULT-DIAGNOSIS IN FULLY DISTRIBUTED SYSTEMS

The problem of diagnosis of failures, at the system level, in large, truly distributed networks of processing elements is considered. A model proposed earlier, in which nodes of the network perform tests upon certain neighboring nodes and pass the results to all the neighboring nodes identifying faults in t he system, is extended by relaxing two assumptions related to f aults on communication links and the flow of diagnostic information in the network. Specific problems considered include: i) Establishment of conditions under which each individual node of a fully distributed system can correctly diagnose failures of other nodes and/or internode communication facilities, based only on results of tests performed by the node itself upon neighboring nodes, and on potentially invalid, second-hand information received via normal communication paths. algorithm to accomplish this diagnosis. ii) Development of a distributed

# Distributed systems and fault tolerance # Computer Science # Interconnection Networks and Systems # Engineering and Test Systems # Engineering

Jon G. Kuhl, Sudhakar M. Reddy (2005). "FAULT-DIAGNOSIS IN FULLY DISTRIBUTED SYSTEMS". IEEE International Symposium on Fault-Tolerant Computing. https://doi.org/10.1109/ftcsh.1995.532651