一种基于信道不可靠环境的协调式检查点协议

在分布式计算环境中经常使用检查点／恢复策略来进行容错。文中主要研究在信道不可靠的环境中通过协调使相互通信的各进程所做的检查点保持全局一致性的方法。通过分析中途消息与信道可靠性之闯的关系以及已有检查点协议对于中途消息处理方法，提出了一种应用于信道不可靠环境下的协调式检查点方法，其消息复杂度为O（N）且不引入其他的计算负担，只通过一次同步即可达到全局一致性状态，相比于以往的协调式检查点协议大大减小了时间开销，提高了在不可靠信道环境中做全局一致检查点的效率。

A Coordinated Checkpointing Protocol Based on Unreliable Channels

In distributed computing environments,checkpoint/rollback recovery mechanism is widely used to achieve fault tolerance.Focuses on a globally consistent checkpoint protocol developed for distributed processes that communicate with each other through unreliable channels.After analysing the methods adopted by existing global checkpointing protocols dealing with in-transmit messages,bring forward a novel method that is applicable in unreliable communication channel.The message complexity involved is within O(N),with no other related computation overhead.It can achieve global consistency through one synchronization,thus greatly reduces the time overhead comparing with the existing methods.Tests indicate that it performs over existing mechanism of channel clearing or message counting.