分布式数据库死锁检测协议

为解决分布式数据中,由于文件加锁的冲突而引起死锁的问题。这类死锁产生的充要条件是在它的请求图中有一有向回路,本文中提出了两个死锁检测协议。协议1适用于一般分布式数据库,它只要求进行一次通讯,且只要求传送一个表的信息;协议2 是针对巨型系统的,是一个层次协议。

On Protocols for Deadlock Detection in Distributed Database Systems

Deadlocks may occur due to conflicts in data file lockings in distributed databases. The necessary and sufficient conditions for the occurrence of this class of deadlocks is that there is a directed cycle in its demand graph, However, due to the inherent communication delay in a distributed system, it is difficult to construct a consistent demand graph for the whole system. In this paper, two deadlock detection protocols were discussed. The first protocol uses a single communication phass and no more information, is needed to be sent. Based on the first protocol, a hierarchical protocol was developed.