Implementation of a speculative concurrency controller

Abstract

Many scheduling algorithms for databases have been developed in order to improve transaction responsiveness. Some developed algorithms meet special needs of real-time databases. These algorithms however, deal with temporal execution aspects, often disregarding semantics of data usage which could further improve performance. One such algorithm is speculative concurrency control, which uses execution redundancy by forking to improve performance. A simplified version of this algorithm has been implemented as a concurrency controller for an object oriented real-time database. The paper describes the experiences with the implementation, explains the reasons for only implementing a subset, and indicates required extensions to implement the complete algorithm. Also provided is a theoretical evaluation of the algorithm performance and scalability.