Читать книгу From Traditional Fault Tolerance to Blockchain - Wenbing Zhao - Страница 46

In this coordinated checkpointing protocol due to Tamir and Sequin [22], one of the processes is designated as the coordinator and the remaining processes are participants. The coordinator must know all other processes in the system. The coordinator uses a two-phase commit protocol to ensure that not only the checkpoints taken at individual processes are consistent with each other, the global checkpointing operation is carried out atomically, that is, either all processes successfully create a new set of checkpoints or they abandon the current round and revert back to their previous set of checkpoints. The objective of the first phase is to create a quiescent point of the distributed system, thereby ensuring the consistency of the individual checkpoints. The second phase is to ensure the atomic switchover from the old checkpoint to the new one. When a participant fails to respond to the coordinator in a timely fashion, the coordinator aborts the checkpointing round.