An integrated error-recovery scheme for multicomputers
Linda S. DeBrunner, P. Ghali · 2002
The paper examines the problem of recovering from processing node failures in multicomputers. A general-purpose, application-transparent integrated error-recovery scheme is presented to recover the multicomputer from processing node failures in the absence of concurrent fault detection and diagnosis facilities. In the scheme, a distributed system-level fault diagnosis algorithm and error recovery algorithms cooperate to obtain a set of consistent and error-free checkpoints. The integration of fault diagnosis and error recovery algorithms permits the implementation of an effective and comprehensive fault tolerance scheme for a wide variety of distributed systems and multicomputer networks.>