一种均衡的移动云计算任务迁移决策算法

A TRADE-OFF TASK OFFLOADING DECISION ALGORITHM FOR MOBILE CLOUD COMPUTING

移动云计算可以通过任务迁移将计算复杂型应用从移动设备卸载至云端执行。然而,任务迁移涉及数据的无线传输,会导致传输延时和传输能耗。为了作出任务迁移的最佳决策,提出一种均衡的移动云计算任务迁移决策算法。算法将任务迁移决策问题建立为Lagrange乘子的非线性优化模型,模型同步考虑了任务迁移后的执行时间代价和执行能耗代价;为了更准确地求解迁移决策,设计一种考虑用户应用动态行为的统计回归模型进行任务执行时间的估算,从而获得时间能耗均衡性能的任务迁移决策。利用N皇后问题和面部识别应用两种任务类型对算法进行了仿真测试分析。结果表明,在平均执行时间、执行能耗、预测准确性等方面,所提算法较对比算法均表现出较好的优势。

Mobile cloud computing can offload the computation-complexity application from mobile devices to perform in the cloud through task migration.However,task offloading refers to wireless data transmission,which leads to transmission delay and transmission energy consumption.To make a best decision of task offloading,this paper proposed a trade-off mobile cloud computing task offloading decision algorithm.Our algorithm established the task offloading decision as a non-linear optimization model of Lagrange multiplier,which took into account the execution time cost and execution energy cost after task offloading simultaneously.In order to solve the offloading decision more accurately,we designed a statistical regression model considering the dynamic behavior of users to estimate the task execution time,so as to obtain the trade-off task offloading decision between time and energy.Two task types,N Queen problem and face recognition application,were used to simulate and analyze the algorithm.The results show that the proposed algorithm has better advantages than the comparative algorithm in terms of average execution time,execution energy consumption and prediction accuracy.