Co-Scheduling of Hybrid Transactions on Multiprocessor Real-Time Database Systems

Abstract

Deriving deadlines and periods for update transactions and scheduling hybrid transactions have been recognized as an important issue in real-time database systems. Despite years of research, all existing techniques focus on assigning deadlines and periods for update transactions in uniprocessor systems and, thus, cannot be applied in multiprocessor systems. Two partitioned scheduling methods for hybrid transactions in multiprocessor systems are proposed in this paper, by developing existing scheduling methods for uniprocessor systems to multiprocessor environments. An effective strategy is proposed to improve the efficiencies of the proposed methods, under which the improvement of efficiency is at the price of the acceptance ratio of transaction sets. For addressing the low acceptance ratio of the improvement strategy, we propose a novel scheduling method. A set of experiments is conducted to test the performances of the proposed methods. To the best of our knowledge, this is the first paper for co-scheduling of hybrid transactions on multiprocessor platforms.