Research On Global Network Optimization Based On Software Defined Network

With the popularity of commercial use of the fifth-generation mobile communication networks,the networks face a dramatic increase in mobile data traffic and a considerable number of device access requests.The traditional network is centered on the terrestrial communication network with a fully distributed architecture,which makes it difficult to achieve global optimization in cross-domain and massive devices.First,the ground network has limited capacity and coverage,making providing services for the exponentially growing terminals and communication devices difficult.For this reason,the space-air-ground integrated network,which is based on the ground network and extends by the space and air networks to provide information assurance for various network applications in wide-area space,has received wide attention in recent years.Second,the fully distributed network architecture has complex protocols and slow network upgrades,making it challenging to meet the requirements of new services and application scenarios for latency,bandwidth,and reliability.For this reason,software defined networks decouple the control plane and data plane,improving network control management capabilities and data processing performance and reducing network application upgrade cycles.In this thesis,we focus on network control optimization,apply software defined network architecture to global optimization,improve the global resource utilization of space-air-ground integrated networks,and realize automatic neighborhood relation optimization of large-scale cellular cells.The main work of the thesis is as follows.First,this thesis introduces the technical background of the space-airground integrated network and cellular cell neighbor relationship optimization.This thesis summarizes the related research on space-airground integrated network architecture and routing strategy,analyzes the difficulties in network routing,and introduces the self-optimizing networks and automatic neighborhood optimization in network management.Second,based on the software defined network architecture,this thesis proposes a resource-aware load-balanced routing strategy,which uses the network resource state to weigh the forwarding cost,and forwards data flows consuming different resource classes to the optimal path.Under limited network resources,the proposed strategy takes the overall network resource utilization as the optimization target and extends the average network service time.Furthermore,this thesis uses node centrality to downscale the underlying network and deploys the resource-aware load balancing policy at several critical nodes to reduce control overhead.Simulation results show that the strategy proposed in this thesis can increase the number of forwarded flows and maintain a lower control overhead compared with the traditional open shortest path first algorithm.Finally,this thesis designs a hybrid distributed and centralized neighbor optimization policy based on the software defined network architecture.The distributed strategy uses local data of base stations to train deep neural networks for immediate neighbor relation optimization,and the centralized strategy introduces global data to improve the accuracy of local models.The neighbor relation optimization strategy proposed in this thesis has high scalability for the input data dimensions while solves the problem of high cost and accuracy guarantee of manual configuration of massive cellular cells.The experimental results show that the hybrid optimization strategy proposed in this thesis can improve the automatic neighbor relation accuracy while reducing the amount of data migration.