Mutual-Aid: Diskless Checkpointing Scheme for Tolerating Double Faults
Jane-Ferng Chiu, Wei-Hua Hao · 2008
Tolerating double faults is an important issue for diskless checkpointing due to the size and increase of the executing time. This is why Mutual-Aid checkpointing has become the first scheme to achieve the goal. Mutual-aid checkpointing combines the advantages of neighbor-based, with parity-based diskless approaches. This also tolerates all double processor faults by bitwising exclusive-or snapshots from its neighbor processors in its virtual assistant ring. In view of the fact that checkpointing and recovery of mutual-aid are so simple and efficient, this increases the performance, reduces application running time, and allows more frequent checkpoints. Moreover, it could be employed towards a very largescale and high performance computing field because of its distributed methods as well as localized operations. The degree of fault tolerance has achieved higher success than other schemes.