Research On The Network Delay Measurement Methods Based On SDN

The number of devices connected to the Internet and network applications continues to grow every year.Traditional networks can no longer meet the needs of Internet upgrading and increasing scale.The high flexibility and scalability of Software Defined Network(SDN)re-energize the Internet,it also brings new opportunities for the research of network measurements.Managers can more conveniently monitor various performance state parameters in SDN through network measurement.Among the performance state parameters of the network,time delay is an important indicator for judging whether the network is congested,and it is also an important factor that affects users’ network experience.How to reasonably and accurately measure and predict network delay at a lower cost is still a problem that needs to be solved urgently.Therefore,this paper studies the network link delay measurement and prediction in SDN.The main work of this paper is as follows:Firstly,for the problem of how to arrange measurement nodes reasonably in the process of link delay measurement,a multi-objective link delay measurement node selection method is proposed in this paper.This method comprehensively considers the three factors of the number of measurement nodes,the total length of the node measurement path and the calculation load balancing parameters of the measurement plan,and abstracts the selection of the delay measurement node plan into a multi-objective optimization problem.By adopting an improved multi-objective gray wolf optimization algorithm to solve the measurement node selection problem,the generated node selection plan reduces the cost of link delay measurement while ensuring the accuracy of the actual measurement.Secondly,to avoid the high cost of measuring multiple network performance indicators,this paper proposes a link delay prediction model based on graph neural network,aiming at accurately predicting network link delay when other network state parameters are known.In this model,the routes and links in the network routing scheme are equivalent to graph nodes,and the mapping relationship between the paths and link nodes is established by using the message passing neural network framework.On this basis,the message transfer process between the path and link nodes is optimized by using simple loop unit and attention mechanism,and the delay state characteristics of the path are read by multi-layer perceptron with double hidden layers.Finally,the delay predicted value of each link in the network is calculated.This method can effectively reduce the measurement of performance indexes in the network and accurately predict the delay performance parameters in any topology.Finally,the comparative experimental results show that the delay measurement node selection method proposed in this paper has lower measurement cost and higher measurement accuracy than the existing methods,and the link delay prediction model proposed in this paper has higher prediction accuracy than existing prediction models.