The Problem of Mutual Exclusion: A New Distributed Solution

In both centralized and distributed systems, processes cooperate and compete with each other to access the system resources. Some of these resources must be used exclusively. It is then required that only one process access the shared resource at a given time. This is referred to as the problem of mutual exclusion. Several synchronization mechanisms have been proposed to solve this problem. In this thesis, an effort has been made to compile most of the existing mutual exclusion solutions for both shared memory and message-passing based systems. A new distributed algorithm, which uses a dynamic information structure, is presented to solve the problem of mutual exclusion. It is proved to be free from both deadlock and starvation. This solution is shown to be economical in terms of the number of message exchanges required per critical section execution. Procedures for recovery from both site and link failures are also given