A dynamic priority based algorithm for distributed deadlock detection and resolution

In this paper we develop a distributed algorithm, using probes stamped with dynamic priorities, for detecting deadlocks in the single resource model. The correctness of the algorithm is proven. A dynamic priority is a pair (p, c) where p is a process and c is the counter value in p when p initiates a deadlock detection computation. The pair is called a detector when it is referenced in a process or a resource manager It is maintained in a process and a resource manager updated dynamically upon certain events, and used as the priority of a probe in a deadlock detection computation. Introduction of a dynamic priority to the probes, combined with a dynamically updated detector in a process and resource manager make the algorithm more efficient and less complex to implement than the existing ones. Our algorithm has an O(1) space complexity for each process. In certain cases, the algorithm uses n messages to detect a deadlock and has an O(n) detection delay where n is the number of processes and resource managers that are involved in the deadlock.