Research On User Mobility Oriented Service Migration Strategy In Mobile Edge Computing

In recent years,the wide application of smart Internet of Things mobile terminal devices has provided great convenience for people’s life and work.5G network has enhanced the application service performance with the advantage of low latency and high bandwidth.As a key technology in 5G network,mobile edge computing can maximize the service needs of end users.In particular,the high uncertainty of users’ mobility and the limitation of geographical coverage of edge cloud services make users put forward higher requirements for real-time service quality,while edge cloud service migration can provide continuous services for users on the move.However,in the real scene,more users to run multiple services at the same time,these services may be different in delay limit and resource allocation,the migration system scalability,service time delay,data transmission overhead,the respect such as node energy consumption in the edge cloud services migration process have influenced on the overall quality of service,this increases the service migration of the complexity of the decision-making process.In order to guarantee high quality of users’ services and low overhead of edge cloud services when edge cloud service migrations,in this paper,the edge of the cloud service migration strategy for research,the main work is as follows:(1)In view of the low service efficiency caused by the optimal node selection process of the target edge cloud in the edge cloud migration strategy,so put forward a kind of service migration preprocessing method based on k-dimensional tree(kdTSM).Firstly,the edge cloud nodes of n-dimensional data features are constructed into a k-dimensional tree.Then,before the edge cloud node sends out the service migration request,the K nearest neighbor search is carried out in the established kd-tree to obtain the candidate target edge cloud node.Finally,during the migration of edge cloud service,real-time monitoring of users’ service demands is carried out,and the target migration node is obtained from the set of candidate edge cloud nodes,so as to avoid the time cost caused by global search for edge cloud nodes in each migration execution process,so improve the service response speed and improve service efficiency.(2)A model-free deep reinforcement learning service migration method(NMDQN)is proposed to solve the problems such as service delay and migration cost caused by unknown user movement and random edge cloud distribution in the process of edge cloud service migration.Firstly,the service migration problem model of the edge cloud is transformed into a deep reinforcement learning model.Then,optimization objectives concerning user service delay,node energy consumption,and service migration overhead are formulated for mobile edge cloud service migration strategy.Finally,the deep neural network and action decision are optimized to achieve the goal of fast migration under the expectation of global optimization in large-scale state space.(3)Through experimental verification,the kdTSM method has certain improvement in the target edge cloud search and service efficiency,to a certain extent,the NMDQN method improves the performance of service migration decision in terms of service delay and migration cost when the user movement is unknown.