Task Offloading And Scheduling For Stochastic Requests

It is well known that computing resource is very precious for mobile applications on mobile devices,and the quality of service(QoS)of the mobile applications is dominated by the limited computing resources.In order to improve the QoS of the mobile applications,mobile edge computing is playing an important role.However,it is a challenge to provide the optimal offloading strategy with the reliability,together with the efficient scheduling strategy for computing resources,to reduce the energy consumption of mobile devices and achieve the efficiency of the systems.Firstly,in this paper we study the impact of encryption and decryption of task data on offloading policy of task in dynamic network.We model the encryption and decryption of task as a computational model based on reliable service composition.In order to obtain the optimal offloading policy in the current network,we propose an online task offloading algorithm based on adaptive receding horizon.The optimal offloading policy in current network environment is obtained by balancing the two optimization objectives of reducing energy consumption and communication latency of tasks.Simulation results demonstrate that the proposed online algorithm not only meets the real-time requirement of the network,but also approximate the global optimal solution of offloading tasks based on the prior information of the network.For a large number of task computing requests in multi-user mobile edge computing system,we propose a mobile edge computing model based on M/M/1 queuing system to minimize the total time of offloading tasks in mobile edge computing system.We prove this optimization problem is NP-complete.In order to achieve load balancing of edge servers and minimize total response time of offloading tasks in polynomial time,we propose two heuristic algorithms greedy and tabu search,respectively.We also propose random algorithm to evaluate the performance of the above two heuristic algorithms.Simulation results demonstrate that the average response time of tasks in the proposed greedy algorithm is saved by 20%,in comparison to the random algorithm.Tabu search algorithm is decreased by 8.5%,compared to the greedy algorithm.