CCUML: a checkpointing protocol for distributed system processes
Sarmistha Neogy, A. Sinha, Pradip Kumar Das · 2004
This paper presents a checkpointing protocol CCUML-coordinated checkpointing with unacknowledged message logging. A checkpoint initiator initiates taking of checkpoints at the end of each checkpoint interval. Processes take local checkpoints only after being notified by the initiator. However, there is no central initiator, but each process takes turn to act as the initiator at each checkpoint initiation. The guaranty that no message would be lost in case of failure, has been brought about by maintaining a log of unacknowledged messages along with the latest checkpoint in a process. Since only unacknowledged messages are logged, the overhead is negligible. Thus, the distributed checkpointing protocol described here always ensures a consistent set of checkpoints from which processes can resume during recovery after a fault.