A Differential QoS Service Migration Decision Method Under SDN Framework

In recent years,with the development of cloud computing technology and the increasing number of mobile intelligent services,the mobile traffic presents a geometric growth,and the requirements of the service to QoS are also increasing.This has put forward high requirements for the communication network.The current centralized structure communication network has a poor matching ability to the communication service.For example,when the user moves far away from the cloud server initiated from the service.The continuous communication with the original server will cause the waste of the network resources and the increase of the time delay of the user,and ultimately leads to the poor user experience.In order to solve this series of problems,the next generation 5G technology uses the SDN architecture in the core network to enhance the communication network's ability to match the service,so as to meet the QoS requirements of the service.Under the framework of SDN,the concept of Follow-me Cloud is proposed to solve the problem of delay caused by user mobility.According to the change of user location,the network transmigrates the service between different servers in different locations,so that users always communicate with the best server.In order to solve the problem of whether the cloud service is migrated,this paper proposes a differentiated QoS service migration decision method based on Markov Decision Process.The service is divided into real-time and non real time services,and there are different requirements for delay.The state space is constructed based on the distance between users and the service type.The return functions are constructed by two QoS indexes of time delay and obtaining rate.The two QoS indexes have different influence weights on the size of revenue according to the real time and non real time property of the service.The aim is to find an optimal strategy to maximize the overall global revenue of user experience and network overhead within a limited mobile range.Simulation results show that this algorithm can achieve greater global benefits than single service and single QoS index.