A commit protocol for checkpointing transactions

A commit protocol is described for checkpointing distributed transactions. Commit protocols are used by distributed transaction management systems to ensure that the multiple nodes participating in a distributed transaction will commit or abort together. This commit protocol is different from others in that a process executing on behalf of a transaction can be interrupted and restarted at some previous snapshot of its state (a checkpoint). The commit protocol guarantees that processes working on behalf of a distributed transaction will be consistent, which implies that the work performed by a restarted process between the time of the checkpoint and the time of the interruption will be undone automatically. The undoing includes any local state changes during that period of time, and any state changes in other processes due to communication with the restarted process in that period. The use of a commit protocol for recovery purposes allows normal execution to be resumed before recovery is completed. Recovery will be carried out in parallel, and the commit protocol guarantees that it is performed eventually. A novel approach of reusing portions of a transaction reduced lost work