Deep Reinforcement Learning for Energy-Efficient Workflow Scheduling in Edge Computing

Workflow scheduling in dynamic edge computing environments faces challenges in minimizing completion time and energy consumption due to unpredictable workloads and limited resources. We propose DQN-Edge, an efficient scheduling method using an attention-based Deep Q-Network (DQN) to learn optimal task prioritization and resource allocation policies. DQN-Edge’s two-phase approach first prioritizes tasks using a modified upward ranking algorithm considering critical path dependencies, then employs a DQN with a context-aware attention mechanism to balance time and energy rewards adaptively. Comprehensive evaluations using real-world scientific workflows show that DQN-Edge consistently outperforms state-of-the-art methods across various scenarios, maintaining high success rates while reducing completion time and energy consumption, even under high-load conditions. The experimental results demonstrate that DQN-Edge can significantly surpass existing methods, reducing makespan by an average of 51.6% and energy consumption by 54.3% compared to basic techniques, and achieving 20.2% and 24.7% improvements over the latest advanced methods, respectively.