一种新的分布式死锁探测与解除方法

The Distributed Deadlock Detecting and Releasing Method Based on Bidirectional Probing through the Request Path

为了有效地解决分布式环境下的死锁探测与解除问题,本文提出了一种基于请求路径双向探测的方法,此方法较以前各类方法具有更好的分布性、可靠性和更短的死锁探测时间。

In order to effectively solve the deadlock detecting and releasing problems in a distributed system, a new method based on bidirectional probing is developed. By using this method, two detecting signals, probB and probF, will be produced whenever a transaction with a higher priority requests a resource held by a transaction with a lower priority. probF will be sent forward and probB backward along the request path. Whenever two signals probB and probF meet somewhere on the path, deadlock will be reported in the system and an algorithm will come to release it. The method proposed is of higher reliability, higher parallelism and with shorter deadlock detecting delay than all currently used methods.