实现分布 /并行系统容错的基础是单进程检查点设置和卷回恢复技术 ,而对进程活动文件状态进行保存和恢复则是这种技术的重要方面 .本文提出的延迟写策略 ,实现了对用户文件的检查点设置 ,有效地解决了在发生故障时用户文件内容与进程全局状态的不一致问题 .它对用户通明 ,并且通过优化设置内存缓冲区大小、时延隐藏等手段 ,使得这种策略在空间开销、正常运行时间、恢复时间等性能指标上优于其它方法 . To achieve distributed / parallel system fault tolerance is based on the single-process checkpoint set-up and rollback recovery technology, and save and restore process activity file status is an important aspect of this technology.This delay proposed write strategy to achieve the user File checkpoint set effectively solve the user file content and process in the event of a failure of the overall state of the state of inconsistency.It is user-transparent, and by optimizing the memory buffer size, delay hiding and other means of making such a strategy in Space overhead, uptime, recovery time and other performance indicators superior to other methods.