Task Offloading And Deployment For Mobile Edge Computing

In recent years,with the development of mobile Internet,mobile applications have become more and more demanding for computation resources.Due to the physical size limitation of mo-bile devices,computation capacity and battery capacity of mobile devices are extremely limited,and therefore local computing of mobile applications cannot meet the requirements of processing delay and energy consumption.At the same time,in the future 5G communication system,there are a large number of redundant computation resources at the network edge.In order to solve the con-tradiction between the increasingly high computation resource requirements of mobile applications and the relatively limited computation resources of mobile devices,mobile edge computing(MEC)has been proposed.MEC is the inheritance and development of mobile cloud computing(MCC).By pushing computation resources to the network edge,computation-intensive and latency-critical mobile applications can be enabled at the resource-limited mobile devices.Consequently,task of-floading and deployment are the core issues in MEC.Therefore,this paper has carried out in-depth research and exploration on the two aspects of coarse-grained task offloading and fine-grained task deployment.The related research is summarized as follows.1.Coarse-grained task offloading in MECThe research context of task offloading in MEC is grasped as a whole.By analyzing the delay and energy consumption of different task offloading strategies,the coarse-grained task offloading problem is modeled as combinatorial optimization.Firstly,we study the computation offloading problem of multi-user scenarios in heterogeneous networks,considering the channel interference of multiple users,and by transforming the problem into a pure strategy game of finite strategy space,a low complexity game-theoretic algorithm that can quickly converge to Nash equilibrium is proposed.Secondly,we study the computation offloading problem of multi-task scenarios in hybrid edge-cloud system,considering the multi-task time overlap.Then we use the optimization techniques to relax the problem and obtain an approximate optimal task offloading strategy.Finally,the simulation results show that the task offloading in heterogeneous network and hybrid edge-cloud system can significantly reduce the total system overhead.2.Fine-grained task deployment in MECSince the coarse-grained task offloading only considers the resource allocation of computa-tion and communication,and ignores the algorithmic structure of the task,we model the task as a fine-grained directed acyclic graph(DAG),and the edge network is modeled as heterogeneous processor network with limited computation and communication capacity.By analyzing the rela-tionship between task structure and network structure,we innovatively proposed distributed task deployment in MEC,which brought computation and communication convergence.Among them,the point weight of DAG indicates the computation complexity of the subtask,and the edge weight indicates the dependency relationship and data exchange between the subtasks,so the model can fully match the algorithmic structure of the task with the structure of the edge network.In order to obtain the minimum task completion time while satisfying the constraints of computation and communication resources and the constraints of the task algorithmic structure,we model the dis-tributed task deployment as a mixed integer program and design an efficient heuristic algorithm.In addition,we present the distributed deployment process of Gaussian elimination algorithm and fast Fourier transform in the edge network.Finally,numerical simulations demonstrate that fine-grained distributed task deployment in MEC can significantly reduce task completion time.