Error detection mechanisms for massively parallel multiprocessors

A survey on the most important methods for error detection in multiprocessor systems is presented. A detailed comparison between watchdog processor and master-checker based fault tolerance is given. The fault coverage, hardware and run-time overhead are discussed, based on the experiences gained in the development of the MEMSY fault-tolerant multiprocessor system. The cumulative effects resulting from the simultaneous use of different hardware-near and high level fault-tolerance mechanisms are shown