Dynamic Configuration Mechanism Of Service Function Chain In NFV Based Mobile Networks

After more than 30 years of explosive growth,mobile communications have developed rapidly.At present,the fifth generation of mobile communication(5G)technology has become an important engine to improve social information applications.The increase of the number of users and the demand for business traffic has been prompting a revolution in the 5G network architecture.Two key network technologies for 5G systems,Network Function Virtualization(NFV)and Software Defined Network(SDN),have attracted wide attention from both research and industrial communities around the world.In the NFV network,the network function will no longer rely on traditional expensive hardware devices,and the virtual network function instances can be deployed on the general server through software,which greatly reduces the network operation cost of the operator.In the NFV network,the Service Function Chain(SFC)can link virtual network functions and complete user requests,making it an important breakthrough point for NFV technology.SFC not only supports fine-grained and flexible service delivery in the network,but also supports modification of service functions and movement of loads.The research emphasis of this thesis is the deployment,migration and scheduling of virtual service function chains in NFV mobile networks.Different deployment proposals have significant change to network performance for the deployment of service function chains in NFV networks.Considering the dependence between processing delay and computing resources,this thesis proposes a deployment scheme of virtual network function chain.In order to investigate the performance of the scheme,we use the end-to-end delay as a precondition to establish an elastic resource allocation model.To maximize the resource utilization of the network is our target.In this thesis,the model is abstracted into a mixed integer quadratically constrained programming problem.A deployment scheme of the service function chain based on predictor corrector primal-dual interior point algorithm is proposed for this programming problem.Through experimental simulation,it is found that the deployment scheme can reduce the resource occupancy rate of the network while ensuring the end-to-end delay of the user.Next,for some mobile scenarios,the traditional deployment solution can no longer meet the user's requirements.In view of this situation,this thesis considers the migration of the service function chain.Moreover,it makes different migration plans as the change of the ingress network element of the SFC.For the migration reconfiguration problem of SFC,according to characteristics of mobile data change,this thesis introduces Markov decision process for modeling analysis,and proposes a migration algorithm based on deep Q learning.Through experimental simulation,it is found that compared with the greedy algorithm and static deployment scheme,the user satisfaction and system benefits are greatly improved in the intelligent migration algorithm.As the number of mobile users increases,the queuing delay of user requests increases dramatically.In view of the scheduling problem of the service function chain,this thesis considers the improvement of network performance as well as the quality of user service.On the basis of reducing the average end-to-end delay of the user,it improves the utilization ratio of network node resources and link bandwidth resources.To solve the issue,the reinforcement learning model is established.The basic algorithm and the shortest SFC priority algorithm are used as comparison algorithms.This thesis proposes a service function chain optimization scheduling scheme that can improve network performance.This thesis studies the optimal deployment and migration scheduling of virtual network function chains in NFV mobile networks.According to different traffic characteristics,an optimization model or a machine learning model is established,and related algorithms of the matching model are proposed to improve the end-to-end delay of the user and the utilization of network resources.