基于移动边缘云的计算密集型任务卸载决策算法

OFFLOADING DECISION ALGORITHM OF COMPUTATION-INTENSIVE TASKSBASED ON MOBILE EDGE CLOUD

为了实现移动边缘云环境中移动设备任务执行时间的优化,提出一种任务卸载决策算法。算法考虑了更为普遍的任务结构类型,包含了顺序任务结构和并发任务结构。其主要发现有两点:(1)一个任务卸载执行,则较大可能其邻居任务也会卸载执行;(2)应用的执行时间可以通过将任务卸载至边缘云端最大化边缘云端与移动设备端的并行性来降低任务完成时间。对于顺序任务构成的线性拓扑任务图,算法可以找到最优的卸载任务群集;而对于并发任务的通用拓扑任务图,算法可以实现更好负载均衡的任务卸载决策。仿真实验结果证明,与基准算法相比,所提算法得到的任务完成时间与理论最优解更为接近,任务完成效率更高。

In order to realize the optimization of the completion time of tasks execution in mobile edge cloud environment,this paper proposes a task offloading decision algorithm.The algorithm took into account more general task structure types,including sequential task structure and concurrent task structure.There were two main findings in this algorithm:if a task was offloaded,it was more likely that its neighbor tasks would be offloaded.Application execution time could be reduced by offloading tasks to the edge cloud to maximize the parallelism between the edge cloud and mobile devices.For the linear topological task graph composed of sequential tasks,the algorithm could find the optimal offloaded task cluster.For the common topological task graph of concurrent tasks,the algorithm could realize better load balancing task offload decision.Simulation results show that compared with the benchmark algorithm,the task completion time obtained by the proposed algorithm is closer to the theoretical optimal solution,and the task completion efficiency is higher.