Based on the Markov chain model for process states and the Laplace transform, a novel optimal checkpoint period solving model for log-based checkpointing and rollback recovery fault-tolerant schemes is proposed. The model takes the difference between the rollback recovery and the failure-free speed into consideration completely, and allows the failure event occurrence in the periods of checkpointing and rollback recovery of the process. The expected execution time of the checkpoint interval is evaluated through solving the transition probability and the weight of the Markov chain of the process state. Finally, the optimal checkpoint period of the process is obtained through minimizing the fault tolerant overhead ratio of the system. The proposed model is consistent with others ff it is degenerative and the results show that the proposal ensures a low fault tolerant overhead.%基于Markov链进程状态模型和拉普拉斯变换,提出了一种日志检查点回卷恢复容错策略的最佳检查点周期求解模型,该模型充分考虑了日志检查点回卷恢复策略中进程回卷恢复与正常运行期间执行速度存在的差别,同时允许进程检查点和回卷恢复期间发生故障事件.通过求解进程状态Markov链转移概率和权重,得到完成检查点间隔的期望执行用时,最后通过系统最小容错负载率得出进程的最佳检查点周期.该模型退化后与现有其它求解模型相一致,结果表明该模型能确保相对较低的容错开销.
展开▼
