Research On Resource Allocation In Mobile Edge Networks

The prosperity of various mobile applications has greatly improved people's daily lives.As support for mobile applications,5G mobile network can improve users' experience by providing a fast response for the mobiles'request.The mobile-edge-computing servers deployed in the access network can make services closer to users,thus reducing the response delay.Mobile edge servers can not only reduce service response latency but also effectively reduce the congestion in the core network as the on-premises character by restricting requests and responses in the access network.In addition,mobile edge servers can also obtain local user information and network information,thus gaining the chance to integrate information technology and communication technology.With the new opportunities in 5G,we study resource problems in the mobile edge network.This thesis focuses on the two main functions of mobile edge computing,i.e.,caching and computing,and studies the cost optimization implementation scheme for mobile edge caching and mobile edge task offloading.Irn the mobile-edge caching problem,the storage process is divided into the dissemination stage and the fetch stage.In the mobile-edge task offloading problem,the processing in the server is divided into centralized processing and distributed processing.The main innovations are as follows:Firstly,in the fetch stage,this thesis studies the problem of minimizing the cost of selectina edge-network servers for caching under dense base station deployment.After the analysis of coverage criteria,a storage server selection algorithm based on the areedy strategy is proposed.The proposed algorithm can effectively reduce the sum cost of content storage and content fetching by combining the user degree and the coverage criteria.Moreover,the proposed algorithm can tolerate different centralities for selecting storage servers with a slight degradation.Secondly,in the dissemination stage?this thesis studies the construction of network coding subgraph for over-lapping multicasts.A unified model is established for traffic multiplication in single multicast and link overlapping in two multicasts,in which single multicast traffic multiplication is regarded as a special case.The schemes for constructing feasible coding subgraphs are proposed for both cases.The proposed algorithms implement a smooth transition of topology with traffic changes and reduce the cost increase by multiplexing existing links in topology.Thirdly,with the edge server adopting centralized processing,this thesis studies the problem of maximizing energy saving for multi-users in a single cell during task offloading.The task queue of sequential processing is introduced into the edge server,and the joint allocation of periods in task queue and channel provide a guarantee for completing users' tasks.This thesis proposes a greedy algorithm based on strategies of tasks calculation tim e adjustment and subtask power allocation,which can effectively help users save more energy.Finally,with the edge server adopting distributed processing,this thesis studies the problem of minimizing network cost of single-user task offloading based on the MapReduce framework.This thesis integrates MapReduce distributed processing frarmework into task offloading problem,establishes a complete model of delay and cost analysis.A selection method for distributed computation resource is designed based on exsiting MapReduce scheme,which provides a feasible way for distributed compation resource allocation.