Research On Resource Allocation Of Mobile Edge Computing

With the rapid iteration of wireless technology,5G has arrived,which demands greater bandwidth,lower delay for its bearer network.To satisfy these requirements,mobile edge computing is regarded as one of the key technologies of 5G network reconfiguration.Mobile edge computing abstracts server hardware resources at the edge layer close to the user layer into virtual machine resources with the help of virtualization technology to greatly improve the computing capacity of mobile devices,reduce service delay and realize service localization.However,resources of mobile edge computing are still limited.For example,the number of available virtual machines in the edge layer,the transmission links and computing resources in cloud layer based on the "cloud-edge-end" architecture.From the point of view of operators,they are finding ways to improve running efficiency of limited resources,make system more reliable and shorten the service latency.Aiming at the resource allocation problem of mobile edge computing,the research in this thesis is mainly as follows:（1）For the user layer and the edge computing layer,this thesis studies the resource allocation based on the unreliability of the service.In setting up the model,the decline of service capability because of deploying multiple virtual machines in the same server node and the reliability of computing network caused by virtual machines software or hardware failure are considered.On this basis,two offloading models are proposed,they have different processing methods for tasks that cannot be completed by VM failure.This thesis uses the method of Markov chain,defines the state space involved in different offloading models in the form of tuples and solves the steady-state probability in the form of matrices.Then,the evaluation index to measure the system performance is established and derived.The correctness of the model and evaluation index is verified by simulation experiments.Finally,this thesis discusses how system performance is affected by different kinds of resource allocation strategies.It provides an analysis model and theoretical support for resource allocation of unreliability mobile edge computing.（2）For the resource allocation in the cloud layer with the goal of reducing service delay,this thesis proposes a resource allocation algorithm that considers transmission link switching and resource borrowing based on load balancing.The computing tasks of the tenant in edge layer arrive randomly.In addition to virtual machine computing resources in cloud layer,the geographical location of the data node and the virtual machine is different,and transmission links between them cause different delays.Therefore,links are classified into high-speed and low-speed ones.The fixed transmission links and computing resources will cause waste due to mismatch with the task load of tenant.Therefore,tenant in edge layer can improve resource utilization by giving up the right to use high-speed links and lending computing resources according to their own load.So,this thesis builds an analysis model and proposes a resource allocation algorithm considering transmission link switching and resource borrowing.The main steps include:Firstly,by setting the threshold value,the busy status and resource borrowing status of tenant are determined.Then,the borrowed and lent resources are calculated.Finally,the links and computing resources are reasonably allocated according to the matching function.The simulation results show that this algorithm can reduce service latency.