Performance analysis of synchronization for two communicating processes

Synchronization of processes is one of the major performance bottlenecks in a distributed system. The synchronization is usually achieved via message passing. There are two basic types of overhead in such a synchronization: the rate of message exchange, and the blocking probabilities of processes. In this paper we consider two processes synchronizing via message passing and study their performance behavior on the basis of the above-mentioned overheads. A number of protocols for message exchange are analyzed. The model gives rise to a three-dimensional Markov chain. An algorithm to solve the model and numerical results are presented to compare the various protocols.