The proposal of Software Defined Network provides a new solution to change the difficult situation of traditional network management.As two important indicators of network measurement technology,network topology and link delay information play an important role in evaluating network performance.The common SDN topology discovery methods and delay measurement methods will produce a large number of redundant data packets,increase the burden of SDN controller and affect the network performance.Based on SDN network,this thesis studied the problem that SDN controller needs to deal with too many redundant data packets in the process of topology discovery and delay measurement.The main work is as follows:(1)According to the problem of packet redundancy caused by the flooding detection method used in the common SDN network topology discovery process,s-nOFDP was proposed.By using the replication function of the switch for packets and using the link event listening mechanism,the number of packets to be processed by the SDN controller in the topology discovery process is reduced.(2)In order to solve the problem of data packets redundancy caused by measuring the dedicated link between the controller and the switch in the common method,this thesis adopted the idea of passive measurement,by deploying measurement points in the network to minimizes redundancy.This thesis presented a measurement points selection algorithm based on Markov process optimized spectral clustering algorithm and a path allocation algorithm based on the principle of minimum depth first,to select suitable locations for deploying measurement points and to assign detection paths to improve the accuracy of time-delay measurement.(3)In this thesis,Mininet and Ryu is used to build the simulation environment of the experiment.By setting up the comparison experiment of single cycle and multi cycle topology discovery,the results show that s-nOFDP can effectively reduce the number of data packets produced by SDN controller and reduce the burden of SDN controller.By setting the comparative experiments of measurement points selection and path allocation,the results show that the measurement points selection algorithm based on Markov process optimized spectral clustering algorithm can better deploy the measurement nodes and configure the detection path,so as to improve the accuracy of delay measurement. |