Real Time Offloading And Scheduling Of Multi-User Dependent Tasks In Mobile Edge Computing

Delay-sensitive and computationally intensive tasks such as online games and face recognition have emerged driven by the upcoming 5G.The long distance between the user and traditional clouds prevents traditional clouds from providing millisecond-level delay computing services for mass users.Therefore,new architectures,which effectively reduces the offloading delay by integrating cloud computing functions into mobile edge networks,are proposed,i.e.,referred to as mobile edge computing(MEC).A large number of studies related to offloading have emerged in recent years,but there are still many limitations.First,most existing work focus on independent tasks,rather than dependent tasks.Unreasonable scheduling is more likely to cause waste of computing resources due to complex dependencies between tasks.Second,some work pays more attention to single-user task offloading than multi-user task offloading.However,multi-user tasks may compete for resource due to the limited resources of the edge server.Finally,most existing work focus on the static offloading,challenges in dynamic scenarios are ignored,such as the dynamic arrival of tasks and the dynamic speed of computing resource.To address the above problems and challenges,multi-user dependent task offloading strategy is conducted in the dedicated and non-dedicated server,which aims at reducing the task response time and delay time.Computing resources are statically stable on a dedicated edge server.First,facing the problem of dynamic arrival of tasks,we propose an on-line sliding offloading and scheduling algorithm for dependency tasks based on the linked list,which overcomes the problem of difficult scheduling caused by task dependency constraints by assigning nonfixed execution time to task nodes.In addition,to further optimize the scheduling solution,two optimization algorithms are proposed: adaptive backtracking algorithms and neighborhood search algorithm based on delayed task.It improves task scheduling performance by rescheduling the delayed task.The speed of computational resource dynamics because of competition between tasks.A dependent task adaptive offloading and scheduling algorithm is proposed for the delayed task caused by the change of computing speed,which overcomes the problem of cascade delay by adjusting the key tasks processed in slow computing resources.To improve the robustness of the algorithm in the case of task surge,we propose an adaptive delay optimization admission control algorithm,which adaptively equalizes some tasks to the cloud computing center based on the delayed task queue,which alleviates the edge server load.A variety of experiments are conducted.The experimental results show that compared with the existing algorithms,our algorithm can reduce the task delay time and task response time on the dedicated and non-dedicated servers.