Research On Edge Cloud Service Migration Strategy For Mobile Applications

With the further development of 5G technology and artificial intelligence,new applications such as virtual reality,face recognition,emotional recognition which need low latency,as well as cloud based games,which require a large amount of memory resources for servers,pose a challenge to the existing network.In order to deal with these challenges,mobile edge computing emerges as the times require.Mobile edge computing is a cloud computing platform that takes cloud computing technology as the core and deploys servers on gateway nodes such as small base stations,routers around users.Compared with traditional cloud services,edge cloud provides services with significantly lower latency and more stability.However,in order to allow each user to enjoy the convenience of edge cloud computing,we need to deploy a large number of edge cloud servers to completely cover different areas.Thus,a single edge cloud computing node can only contain limited computing and storage resources.In addition,due to the mobility of users,many delay-sensitive application services need to migrate with users.This frequent migration will obviously cause problems such as low efficiency in resource utilization and finally affect user experience.To solve the above problems,we research on the migration strategy of edge cloud services.First of all,edge computing node can't meet the low latency requirements of some services due to high load state.Therefore,according to the user's mobility,we employ the periodicity of the edge cloud computing node and elastic service to establish a system model which mitigate the high load state of edge computing nodes.Secondly,we propose a physical machine memory load prediction algorithm based on the low-order Markov model and the influence of memory state between different moments.Then we combine future memory load state with the benefit of service migration and the dynamic programming algorithm to provide the best edge cloud service migration strategy.This strategy can avoid a large number of user service migration activities which lead to high load state of edge cloud nodes and improve resource utilization of the entire system.Finally,the effectiveness of the strategy is proved by simulation experiments of four service migration strategies.The maximum number of users,cloud service quality,average user service migration data size and average delay are analyzed under different strategies.