Predicting processing time of real time transaction in replicated DRTDBS via middleware

Abstract

The performance of Replicated Distributed Real Time Database System depends on the processing time of Real Time Transaction (RTT) and mutual consistency between replicated data item. Existing research works are kernel based and primarily concentrates on maintaining mutual consistency between replicated real time and non-real time data item. However, research towards predicting the processing time of RTT is still in preliminary stage. Predicting the processing time of RTT involves different factors such as resources/data conflicts between different RTTs, the data requirements of RTT, conventional recovery algorithm, interaction with indeterministic sub-system, and mutual consistency between replicated data. The objective of this paper is to predicts the processing time of RTT with maintaining mutual consistency between replicated data item via middleware. An existing middleware [15] is extended with prediction layer that efficiently predicts the processing time of RTTs and maintains replica consistency. This proposed solution is well suited for different types of real time applications where simultaneously mutual consistency and timeliness demand of RTT must be satisfied. The experimental results show that our proposed approach is beneficial for the RDRTDBS.