节点文献
一种具有最小阻塞消息发送时间的检查点策略
A Checkpointing Strategy with the Minimum Blocking Message-sending Time
【Author】 Wang Yan, Xu Zhenpeng, Cao Liujuan (College of Computer Science and Technology, Harbin Engineering University, Harbin, Heilongjiang 150001)
【机构】 哈尔滨工程大学计算机学院;
【摘要】 基于时钟的自适应检查点策略是一种简单有效的协同检查点策略。现有的基于时钟计时器的检查点策略采用阻塞进程消息发送或两段自适应检查点策略来解决不一致消息的问题。文中讨论了不一致消息引发的系统不一致问题,并基于其特性提出一种具有最小阻塞消息发送时间的检查点策略。该策略不会产生多米诺效应且不需要记录任何进程间消息,缩短了进程阻塞消息发送时间,且检查点生成过程中使进程得以同步,因此算法具有较好性能。
【Abstract】 The time-based checkpointing strategy is an efficient coordinated checkpointing scheme. The existing time-based checkpointing schemes deal with inconsistent problems when failure occurs in some special period through blocking message-sending or two-phase time-based strategy. In this paper, the problem that a checkpointing strategy will result in the system’s inconsistency is investigated, then a checkpointing strategy with the minimum blocking message-sending time is proposed based on its feature. The strategy can avoid domino effect and log message between processes, which makes the blocking message-sending time shorter, and synchronizes processes in the process of checkpointing.
【Key words】 distributed system; fault tolerant; checkpointing; rollback recovery;
- 【会议录名称】 第四届中国测试学术会议论文集
- 【会议名称】第四届中国测试学术会议
- 【会议时间】2006-08
- 【会议地点】中国河北秦皇岛北戴河
- 【分类号】TP316.4
- 【主办单位】中国计算机学会容错计算专业委员会