Supporting Transaction Predictability in Replicated DRTDBS

Abstract

The design and implementation of replicated distributed real time database system (RDRTDBS) must meet two rigorous requirements; deadline of real time transactions (RTTs) and preserving of the mutual consistency of replicated data. Previous researches in RDRTDBS have been concentrated mainly on designing of replica update techniques (RUTs) for soft and firm RTTs with sole correctness criteria of serializability and epsilon serializability. No work has been reported for predictable processing of real time transaction (RTT) with guaranteeing the mutual consistency of replicated data. Therefore, this paper first addresses the factors of predictability and mutual consistency in RDRTDBS and then briefly discusses the features and requirements of RDRTDBS and presents a processing plan that supports predictable execution of hard, soft and firm RTT along with maintaining the mutual consistency. The simulation results demonstrate that the proposed processing scheme enhances the performance of RDRTDBS beyond that offered by the existing RUTs.