关键词: 容错, 动态分组, 两级检查点, 回卷恢复, 单阶段提交算法

Abstract:

In order to reduce the overhead of time and space for setting checkpoints,a two-level checkpoint algorithm is proposed,in which the cooperative checkpoint algorithm is adopted for the group level and the single-stage checkpoint algorithm is used for the system level.The algorithm,which is based on the distributed dynamic grou-ping,eliminates orphan messages by sender grouping and implements the change from the conventional two-stage submission algorithm to the single-stage one.Experimental results indicate that the proposed algorithm is of high practicability.The execution time of it is comparatively low and the time complexity for obtaining checkpoints reduces from O（n2） to O（n）.

Key words: fault tolerance, dynamic group, two-level checkpointing algorithm, rollback recovery, single phase commit algorithm