Efficiency-oriented Task Offloading with Quality Level Constraint in Green Edge Computing
Maosheng Zhu, Xi Li, Hong Ji, Heli Zhang · GLOBECOM 2022 - 2022 IEEE Global Communications Conference · 2022
Edge computing ensures close processing in proximity to mobile devices (MD) via task offloading, resulting in a timely manner to support diversified services with low latency tolerance. Nevertheless, for the proliferation of services, the massive connection between MDs and edge servers (ES) results in huge energy consumption, leading to sustainability issues for greenhouse gas emissions. Besides, quality levels within the same service (e.g., accuracy of object detection, and clarity of video) become subdivided, requiring tighter service environment constraint of ESs for such subdivided quality levels. In this paper, we investigate efficiency-oriented task offloading for services with subdivided quality levels within quality level constraint of ESs. Specifically, we first devise a unified quality-aware service model to abstract out service structure, i.e., the quality-relation, number, and dependency of tasks within it. Then, the offloading algorithm is proposed based on an optimized version of non-dominated sorting genetic algorithm-II (NSGA-II), in which due to the heterogeneity of services, we also embed a scheduling algorithm for services in advance to improve tasks parallelism for concurrent execution and NSGA-II adaptation. Additionally, compared with other algorithms, simulation results demonstrate that our proposed algorithm not only significantly improves the convergence, but also optimizes its energy and latency efficiencies.