An Improved Energy-Efficient DAG Scheduling Algorithm for Uniform Multiprocessor Systems

A significant number of embedded systems are battery-powered, including smartphones, automotive electronics, and drones. As a result, energy management is crucial in the design and application of embedded systems. With the continuous advancement of technology, the design of embedded systems is becoming increasingly complex. Balancing system performance with efficiency has become a critical challenge. This work addresses the energy-aware scheduling problem for real-time applications with precedence constraints in uniform multiprocessor systems with Dynamic Voltage and Frequency Scaling. We propose an energy-efficient algorithm called the Out-Degree-Based Energy-Efficient Real-Time DAG Scheduler (OERDS). OERDS effectively exploits the out-degree characteristics of tasks and employs a pre-enhancement approach that selects the appropriate execution frequency for each task-processor pair in the systems. We evaluate the performance of the OERDS using real-world benchmark applications. Experimental results demonstrate that OERDS outperforms other approaches in terms of energy efficiency optimization.