实时分布式系统失败结点判定及优先列表重构

Research on Diagnosing the Node Failure and Reconstructing the Preferred List in a Distributed Real-time System

实时分布式系统中,如果结点n自从最后一次收到结点i的广播信息以后超过时间T(i)out还没有收到结点i的广播信息,就认为结点i失败,然后调整各正常的结点的优先列表,以保证每个结点被一个也只能是一个结点选择做自己的第k个优先结点,以减少任务的失败率.确定最佳的超时值T(i)out及时检测一个结点的失败,并调整结点的优先列表以处理结点的失败.

In a distributed real-time system, if node n has not heard from node i for T^((i))_(out) since its receipt of the latest broadcast from node i, it will be considered that node i failed; then the preferred list is modified in such a way that each node will be selected as the k^(th) preferred node of one and only one other node. This paper determines the 'best' timeout period T^((i))_(out) to diagnose the failure of a node by the other nodes and proposes an algorithm to modify the preferred list in such a way that node failures are tolerated.