备查点间隔对程序执行时间影响的分析

Analysis of Effect of Checkpointing Intervalon Program Execution Time

软件错误按其表现形式可分为柔性错误与固性错误两大类．本文根据以往软件测试的结果指出，对柔性错误的防范与容错应被视为软件容错的主要目标．文中简要介绍了对柔性错误具有良好容错效果的备查点与卷回机制，并在此基础上以一个顺序程序的容错模型来讨论备查点的设置间隔对容错程序执行时间的影响，然后通过对该模型的计算机模拟与理论分析得出几点具有实际意义的结论．本文工作可为备查点插入策略（程序员插入、编译器插入和硬件插入）的优化提供有力支持．

n this paper,software faults are classified into two classes based on their arising patterns,that is,soft faults and solid faults. We point out that tolerating of soft faults is the main goal of software fault tolerance. Then,we introduce briefly the checkpointing & rollback mechanism which is commonly considered to be efficient for tolerating soft faults,and discuss the effect of setting of the checkpointing interval on the executing time of fault tolerating sequential programs based on a simple fault tolerant model. Besides,we give a computing method of the best checkpointing interval,also get some helpful conclusions through simulation and theoretical analysis. The computing method of the best checkpointing interval presented in this paper can provide a strong support to various checkpoint inserting strategies.