基于概率预测与粒子群算法的桌面云能源策略

On the Energy Policy for Virtual Desktop Infrastructure Based on Probabilistic Prediction and Particle Swarm Optimization

桌面云终端尤其是移动设备在使用时与服务器的断开、连接操作十分频繁,导致基于传统超时策略的桌面云数据中心能效比较低,而一种基于概率预测与粒子群算法的能源策略使桌面云管理平台能够实时分析用户的历史行为数据,并制定出使虚拟机期望能耗和期望服务时延最低的能源策略.通过实际数据集仿真验证,该策略在保证最小服务时延的情况下,较基准策略节省约34%的能耗,运算速度提升26倍.

When using virtual desktop terminals,especially on mobile devices,the operation of connecting and disconnecting service is frequent,which results in low energy efficiency of VDI center based on traditional timeout policies.But a VDI energy policy based on probabilistic prediction and particle swarm optimization(PSO)enables the VDI management platform to analyze user s historical behavior data in real time and develop an energy policy that minimizes the expected energy consumption and service delay of the virtual machine(VM).Simulation results of actual data set show that while ensuring the minimum service delay,the proposed policy can save about 34%of energy consumption and improve the computing speed by 26 times compared with the benchmark policy.