Research On Deployment And Dynamic Configuration Of Service Function Chain In Mobile Cellular Network

With the rapid development and application of artificial intelligence,Internet of things,cloud computing and other technologies,mobile cellular data presents an explosive growth trend.Heterogeneous data with huge differences puts forward higher requirements for the service quality assurance of mobile cellular network,which brings great challenges to the traditional middleware chimney cellular network structure based on specific network physical facilities.At the same time,such as routers,switches and other network hardware facilities are fixed,whose quantity are huge.It is difficult to meet the massive and dynamic network requests,resulting in the network efficiency greatly reduced.Therefore,the Service Function Chain(SFC)based on Network Function Virtualization(NFV)is introduced into cellular mobile network,and it has become a hot research direction in academic and industrial in recent years.The deployment of service function chain is optimized by setting different optimization objectives,such as minimizing network delay,minimizing energy consumption and maximizing resource utilization.This thesis studies the deployment of service function chain based on the joint load balancing of general server and physical link,the deployment strategy based on neural combination optimization to minimize the total power consumption of service function chain,and the dynamic resource allocation of service function chain.1.Research on service function chain deployment based on network load balancing.In this study,the service function chain deployment problem is modeled as a mixed integer programming model(MILP),and a service function chain deployment algorithm based on load balancing is proposed.The algorithm is divided into two parts:service function chain link mapping algorithm and virtual network function instance deployment algorithm.Link mapping algorithm effectively selects an optimal path from the global path as the link of service function chain by calculating the current load of general server and physical link;virtual network function instance deployment algorithm further deploys virtual network function instance according to the load of general server on the optimal path.Simulation results show that the algorithm can consider the load of general server and physical link at the same time,realize the optimization of service function chain deployment decision,avoid the network bottleneck caused by overload of general server or physical link,and effectively improve the reliability of cellular network,which has reduced the load of general server and physical link about 20%comparing traditional greedy algorithm.2.Research on service function chain deployment to minimize system power consumption.Aiming at the problem of high total power consumption and low usage efficiency of network resources,this research considers the minimum total power consumption of the network as the optimization goal under the premise of satisfying the deployment quality of the service function chain,and mathematically modeling the deployment problem of the service function chain.At the same time,for the problem that heuristic algorithms or meta-heuristic algorithms often fall into local optimal values in the optimization process,a service function chain deployment algorithm based on neural combination optimization is studied.The algorithm uses reinforcement learning combined with deep neural networks to make a service function chain deployment strategy to reduce the power consumption of general servers and physical links in the network structure to achieve the deployment goal of minimizing total network power consumption.Through experimental simulation analysis,it is proved that under the premise of ensuring the quality of deployment,this algorithm effectively reduces the total power consumption of the system by nearly 19%compared with the traditional greedy algorithm.3.Research on service function chain deployment based on dynamic resource demand forecasting.This research takes the dynamic resource allocation of virtual network functions in the service function chain as the goal,and proposes a dynamic resource allocation model of the service function chain based on the demand resource prediction module of virtual network function instances.Research shows that the resource demand of virtual network function instances and network requests are a non-linear relationship.The model uses machine learning to fit the relationship between the two,and can predict the resources that the virtual network function instance needs to allocate based on the user's network request.Then allocate service function chain resources according to demand.The experimental results show that the model effectively reduces the total power consumption of the system by nearly 34%compared with the traditional static resource allocation method,and effectively solves the problems of insufficient resource allocation and excessive resource allocation.