Improved Decomposition-Based Global EDF Scheduling of DAGs

Abstract

Recently, an increasing number of real-time systems are implemented on multicore systems. To fully utilize the computation power of multicore systems, the scheduling problem of the real-time parallel task model is receiving more attention. Different types of scheduling algorithms and analysis techniques have been proposed for parallel real-time tasks modeled as directed acyclic graphs (DAG). In this paper, we study the scheduling problem for DAGs under the decomposition paradigm. We propose a new schedulability test and corresponding decomposition strategy. We show that this new decomposition approach strictly dominates the latest decomposition-based approach. Simulations are conducted to evaluate the real-time performance of our proposed scheduling algorithm, against the state-of-the-art scheduling and analysis methods of different types. Experimental results show that our method consistently outperforms other global methods under different parameter settings.