Research On Service Migration Method In Mobile Edge Computing Environment

Mobile Edge Computation(MEC)shortens the distance to users geographically by deploying computing resources to the edge of the network,and can handle user requests nearby,avoiding long network transmissions,thereby improving service response speed.Because edge nodes are deployed at the edge of the network,the coverage of a single node is relatively limited,so the movement of users may cause the user to leave the coverage of the current node and enter the coverage of another node.When a user enters the coverage of another node from the coverage of o ne node,in order to ensure the quality of experience(Qo E)of the end user,service migration is required.In order to minimize the impact of service migration on the quality of user experience,it is necessary to complete the service migration as quickly as possible.To this end,this thesis has done the following work from three aspects: service migration methods,service image caching,and container migration.First,this thesis proposes a service migration method based on replay.This method caches the user input data during the migration while performing data migration,and reprocesses the input data after the data migration is completed,so as to achieve state synchronization.Compared with the current mainstream iterative-based method,especially for the migration of computationally intensive services,this method can complete the migration of user services in a shorter time,and also has a relatively shorter service interruption time.Second,this thesis uses machine learning methods to predict the requests of users with known activity patterns.According to the special environment of mobile edge computing,the model is optimized for the two elements of geographic location and time,so as to realize the prediction of user access requests.By predicting user access requirements,relevant application images can be deployed in advance,thereby effectively reducing the average time required for service migration in the region.Third,this thesis uses the relevant characteristics of Docker to design and implement rapid image migration and container rapid migration methods.This thesis fully combed the union file system used by Docker and the layered storage ideas of Docker,designed a rapid image migration method based on incremental synchronization,and also implemented a container migration method based on iteration and replay.The two container migration methods are compared.The mirror migration method based on incremental synchronization can effectively reduce the amount of data that needs to be transferred during mirror migration,and the recurring container migration can effectively shorten the time required for container migration when migrating computationally intensive containers.Aiming at the problem of service migration in the MEC field,this thesis proposes a replay-based service migration method and a mirror pre-caching method using machine learning methods from different levels and perspectives to achieve faster migration of user services between two edge nodes.At the same time,this thesis also applies the replay-based migration method to the current mainstream container products,realizing container migration in a shorter time;in addition,it also designs a mirror migration method based on incremental synchronization,which can be used when the registry server is unavailable.