Abstract: The development of mobile devices has led to a proliferation of various computationally intensive mobile applications.In response to the prominent demand of such applications on the computational power and energy consumption of mobile devices,the workflow tasks were optimized with approximate computation,and the Quality of Service(QoS)perspective was introduced to limit the degree of optimization processing.An Energy Aware workflow offloading strategy based on QoS(EA-QoS)was proposed.In the EA-QoS strategy,the edge nodes was sorted according to the transmission energy consumption per unit load,the nodes were grouped based on the energy consumption threshold and the QoS levels were assigned,so as to reduce the complexity of subsequent offloading algorithms.The tasks were stratified and the Heterogeneous Earliest Finish Time(HEFT)algorithm was used to generate the ready queues.The Non-dominated Sorting Genetic Algorithm(NSGA-III)was used to find the optimal scheduling solution scheme by integer mapping of task execution order,edge nodes and QoS levels.The simulation experimental results showed that the EA-QoS policy improved the energy consumption optimization of workflows by an average of 12.02% and 35.14% compared to NSGA-III and PSO respectively.

Key words: mobile edge computing, delay optimization, energy optimization, approximate computation, genetic algorithms, workflow

摘要： 移动设备的发展导致了各种计算密集型移动应用的激增,针对这类应用对移动设备计算能力和能耗的突出需求,利用近似计算优化处理工作流任务,引入服务质量(QoS)的角度限制优化处理程度,并提出一种基于服务质量的能耗感知工作流卸载策略(EA-QoS)。首先,EA-QoS对边缘节点按照单位负载传输能耗排序,对节点按能耗阈值进行分组并分配QoS等级。然后,对任务进行分层并利用HEFT算法生成就绪队列。最后,采用非支配排序遗传算法(NSGA-Ⅲ),在编码策略中通过对任务执行顺序、边缘节点和QoS等级进行整数映射,寻找任务最优调度方案。仿真实验结果表明,EA-QoS相较NSGA-Ⅲ和粒子群优化算法(PSO),工作流的能耗优化效果分别平均提升了12.02%和35.14%。

关键词: 移动边缘计算, 延时优化, 能耗优化, 近似计算, 遗传算法, 工作流