融合虚拟机分簇与休眠机制的MEC任务卸载策略

Task offloading strategy with clustered virtual machine and sleep mechanism in MEC

为了在低延时约束下降低能源消耗,研究一种移动边缘计算任务卸载策略。采用虚拟机分簇技术,在移动边缘计算服务器中设置始终保持活跃的主模块及可以适时休眠的备用模块。基于备用模块虚拟机周期性休眠机制,建立同步多重休假排队模型,利用拟生灭过程与矩阵几何解方法,给出任务平均延时及系统节能率。实验结果验证了该任务卸载策略的有效性。融合"教与学"过程改进鸽群算法,优化该任务卸载策略的系统参数,实现不同性能指标的合理均衡。

In order to reduce energy consumption,a task offloading strategy is proposed in mobile edge computing(MEC).By using virtual machine clustering technology,the virtual machines in MEC server are clustered into two modules.The virtual machines in the main module keep active all the time,while the virtual machines in the spare module can go to sleep adaptively.According to the sleep mechanism of the virtual machines in the spare group,a queuing model with synchronous multiple-vacation is established.By using the methods of Quasi-Birth-and-Death process and matrix-geometric solution,performance measures in term of the average delay of tasks and the energy saving rate of system are derived.Experimental results verify the validity of task offloading strategy in MEC.Integrating"teaching and learning"process into the pigeon algorithm,system parameters of the proposed task offloading strategy in MEC are optimized,and a reasonable balance between different performance measures is achieved.