An Improved Transaction Scheduling Policy for Mixed Real-Time Nested Transaction Models

Abstract

In this paper, we study the problem of scheduling transactions in the distributed mixed real-time/non-real-time databases systems, in which both real-time and non-real-time flat/nested transactions exist simultaneously in a multiprocessor environment. Feedback adjustment policy (FAP) is proposed to attempt minimizing the number of missed real-time transactions, and reducing the impact on the performance of non-real-time transactions. FAP allocates the processors to both of real-time transactions and non-real-time transactions dynamically based on the real-time ratio for overload adjustment measured statistically by MissRatio. Simulation Results are shown to deliver good performance when an application requires a mixed real-time/nonreal- time transaction model.