Research On Business Transmission Optimization Of Software-defined Networking

Software-defined Networking(SDN)has achieved the decoupling of the control plane and the data plane,and innovatively solved the problems such as rigid structure of the traditional network system,difficult management,and difficulty in updating.In recent years,it has become a hot spot in academic research and industrial application.However,with the continuous development of all kinds of network infrastructure based on SDN,especially data center network,a large number of network applications and network traffic show explosive growth,making it difficult to guarantee service quality and poor end-user experience in the service transmission process of SDN.To be specific,firstly,with the diversified development of network services,network service state information presents high-dimensional characteristics.The existing SDN is limited by its node perception ability and rigid service distribution mechanism,which makes it difficult to efficiently complete global network service perception and provide accurate perception basis for business transmission.Secondly,in the distributed multi-controller deployed SDN network,if the sub-domain division is unreasonable and the cooperation between controllers under multi-domain management is poor,the SDN service transmission efficiency will be seriously affected.Finally,the network traffic has the characteristic of uneven distribution in time and space.In the case of sudden traffic,the controller is prone to overload or light load,leading to unstable service transmission.First,with the diversified development of network services,the status information presents a high-dimensional feature.Existing SDN is limited by its node awareness and rigid task allocation mechanism,so it is difficult to efficiently complete global network status awareness tasks and cannot provide accurate perception basis for optimizing SDN transmission.Second,in an SDN network deployed by distributed multi-controllers,if the division of sub-domains is unreasonable and the coordination among controllers under multi-domain management is poor,the control efficiency of SDN transmission will be seriously affected.Finally,there is a non-uniform distribution of time and space in network traffic.Therefore,in the case of traffic bursts,the controller is prone to overload or light load,resulting in loading imbalance.This paper is supported by the National High Technology Research and Development Program(863 Program)“Research on the Architecture and Key Technologies of Software-defined Networking”.It focuses on the transmission optimization of SDN,and conducts optimization from four aspects: SDN intelligent transmission framework,high-dimensional state awareness algorithm,multi-domain collaborative control mechanism and network traffic load balancing strategy.The main research work and results are as follows:(1)For software defined network transmission optimization of business requirements,the cluster movement in biological individual voluntary movement,the overall characteristics of intelligent coordination,will implement the distributed network state regional autonomous perception co-ordinated global load balancing for network optimization goal,designs a framework based on cluster motion SDN network business transmission in SDN cluster motion,introduced in the network,gives the framework of function model and the corresponding data layerThe key mechanism of realizing SDN service transmission framework based on cluster movement is described in detail to provide architectural support for realizing intelligent decision and efficient control of SDN service transmission.Finally,the prototype verification system proves the effectiveness of the framework for optimization of service transmission.(2)To solve the problem that node performance is limited and it is difficult to carry out highdimensional state perception of network services,a SDN service perception algorithm based on dimension reduction and allocation is proposed under the framework of SDN service transmission.The algorithm is divided into two stages: business reduction and business allocation.In stage 1,the high-dimensional perceptive services are divided into several low-dimensional perceptive subservices by using the clustering idea of k-center method.In stage 2,according to the performance of network nodes,one or more sub-services are selected or matched by nodes for perception.At the same time,adjustment parameters are introduced to adjust the probability of each node participating in the business perception,and tradeoff factors are set to balance the perception cost and the number of network nodes participating in the perception.Simulation results show that the service perception rate of the algorithm tends to 100% with the increase of the number of nodes,the perception cost is lower and becomes more obvious with the increase of network size,and the node perception equilibrium is proportional to the number of nodes.(3)In view of SDN domain problem of poor control of collaborative business transfer,business in SDN transmission framework,this paper proposes a SDN based on neighbor contextaware multi-domain collaborative control mechanism,first,design the spread based on neighbor clustering network domain algorithm,cognitive state information between network nodes(hop and flow request rate),defined and compute nodes attract and belonging degrees,based on the neighbor spread process of SDN multi-domain;Then,a controller load optimization algorithm based on cooperative mapping is designed.By implementing bidirectional mapping between switch and controller,the connection relationship between switch and controller is optimized to balance the controller load.Simulation results show that the switk-controller delay decreases by an average of 34.5% and the controller load balancer increases by at least 26.7%.(4)To solve the problem of controller load imbalance in the case of network traffic burst,a switch adaptive migration strategy for SDN controller load balance was proposed under the SDN service transmission framework.First,add functional modules to each controller node,dynamically set the controller overload determination threshold,and then realize the optimization and adjustment of controller load through load collection and measurement,overload determination,migration domain selection and adaptive migration.Finally,simulation results show that compared with the existing switch migration algorithm,the migration efficiency of this strategy is improved by 19.7%.