Reliability-Aware Real-Time Task Mapping for Multi-Server Edge Computing Systems With DVFS

Edge computing enhances user Quality of Experience (QoE) by deploying computing and storage resources at the network edge, enabling efficient management and scheduling of offloaded tasks. Application execution performance is primarily determined by task mapping. However, multi-server edge networks face challenges due to potential failures in hardware or software, where both permanent faults in devices or links and transient faults during task execution can harm task completion. Additionally, real-time responsiveness and energy efficiency are critical for service quality and system sustainability. To address these limitations, we propose a Reliability-Aware Task Mapping (RATM) method integrating task duplication, DVFS, and communication link selection to optimize energy consumption while satisfying task reliability and real-time constraints. The joint design is formulated as a complex nonlinear optimization problem and is linearized to find the optimal solution. Experimental results demonstrate that the proposed method outperforms existing approaches regarding system energy efficiency, task reliability, and schedulability.