Research On QoS Guarantee Technology Based On SDN

With the sharp increase in network traffic,people's demand for networks is becoming more and more diverse.The traditional network architecture has been unable to meet people's needs for the network.Software-defined networking(SDN-Software Defined Network)attempts to propose a new network architecture to solve these problems.The most significant feature of the SDN network architecture is the separation of the control plane and the data plane.Programmable network traffic and centralized control.However,under the new network architecture,the problems faced by the network have not disappeared.Due to the architecture update,its solution is different from the previous traditional network architecture.Therefore,how to guarantee the quality of service(QoS)under the SDN network? It has become a focus of current research.This paper mainly focuses on two issues.The first is that the existing QoS evaluation model has only a qualitative evaluation model and no quantitative evaluation model.Therefore,this paper constructs a QoS evaluation model based on business matching degree.Under the SDN network architecture,this evaluation model divides QoS attributes into measurable operation attributes and predictable performance indexes.Through the extracted QoS operation indexes The performance index is predicted,and a dynamic QoS awareness model based on Markov is established.The weight coefficient is obtained by Markov state transition probability of QoS operation cost.At the same time,the cost attribute is introduced to evaluate the service quality of a specific service in cloud data center in combination with operation cost.Secondly,the effect of traditional load balancing method on SDN network optimization is not obvious Problem,this paper tries to combine swarm intelligence optimization algorithm to achieve load balancing,and make full use of the centralized characteristics of SDN network control plane.Firstly,the adjacency matrix of network topology is constructed in memory by using lldp and SNMP protocol,then the problem is transformed into the shortest path optimization problem,and the following two improvements are made for the load balancing problem of the network.One is to initialize the initial value of pheromone on the link by using Dijkstra shortest path tree to make the initial path more instructive,the other is to guide the iteration by using service source vector Process,so that the path selected in iterative path finding is always closer to the end point and the node with higher resource availability;Finally,the simulation topology is built by simulation tools,and the two improved methods are verified by experiments.The simulation results show that the improved QoS evaluation model in this paper has better availability and applicability;at the same time,the improved swarm intelligence optimization algorithm is successfully applied to the research of load balancing based on SDN,and has good characteristics.