分布式系统中失败结点的判定与优先列表的重构新方法

A New Method of Diagnosing Node Failures and Reconstructing the Preferential list in Distributed Systems

实时分布式系统中,如果结点n自从上一次收到结点i的广播信息后经过超时值时没有再收到结点i的广播信息,就认为结点i失败。这时需要调整各正常结点的优先列表,以保证每个结点被一个也只能是一个结点选择为自已的第k个优先结点,以减少任务的失败率。本文提出了一种确定最佳超时值以及检测结点失败并且调整结点的优先列表以处理结点失败的新方法。

In a distributed real-time system, if node n has not heard from node i for the timeout period since its receipt of the latest broadcast from node i , it will consider node i fails. In this case, it needs to modify the preferential list for all nodes to ensure each node in the preferential list is selected as the k th preferential node by one and only one node else.This paper proposes a new method for determining the optimal timeout value and diagnosing the failures of a node, and an algorithm for modifying the preferential list to tackle node failures.