Dynamic multiprocessor scheduling for supporting real-time constraints

A real-time transaction carries the constraint that it must be completed before its assigned deadline. For some real-time applications, a successfully completed transaction may contribute a value to the system to reflect its profit. Satisfying both constraints of maximizing the totally obtained profits and minimizing the number of missed transactions simultaneously under various system conditions is a challenge. In this paper, we present a dynamic scheduling policy named Dynamic Processor Allocation (DPA) for scheduling value-based transactions in a multiprocessor real-time database system. The DPA policy allocates the processors to both of high-value transactions and urgent transactions dynamically by utilizing the statistical information of the system. Through simulation experiments, DPA is shown to deliver good performance in both maximizing the totally obtained profits and minimizing the number of missed transactions under various system environments. Hence, it resolves the drawbacks of the existing scheduling policies which can deliver good performance only at normal loads or at high loads.