Mathematical modeling and simulation of an algorithm for optimistic concurrency control in centralized database systems

This paper presents the analysis of an algorithm for optimistic concurrency control in centralized database systems. The effectiveness of optimistic concurrency control in a centralized database system depends on the probability of transaction conflict being low. When conflict does occur, it is resolved by aborting the older transaction and allowing resubmission. The concern here is to examine the probability that a transaction is able to avoid conflict and commit. Specifically, a set of probabilistic formulas which provide an analysis of this probability is examined. Furthermore, a report of two simulation studies, the design and implementation of a GPSS simulation and the design of a simulation for the IBM token-ring network, is given.