Research Of Fast Flow Classification Strategy And Adaptive Routing Algorithm In SDN

As a new type of network architecture,SDN realizes the separation of network control plane and data forwarding plane,has advantages of programmability,controllability and high security.It has been widely used in cloud computing and data centers.However,there are shortcomings such as large transmission delay and high packet loss rate in SDN operation.These problems are usually related to network routing,the detection of large data flows and route calculation are two important factors.Aiming at problems of large delay in data flow acquisition,bandwidth consumption,and large computing resources in SDN routing,a fast flow classification routing strategy is proposed.This strategy mainly adopts the method of extending the packet-in message in OpenFlow protocol to implement fast large flow detection,to solve problems in the process of identifying big data flow,and to use the rerouting processing mode for the identified big data flow to make it as much as possible,and get more bandwidth,and increase transmission speed,and reduce data transmission delay.In order to solve the problem of the controller periodically detecting the low efficiency of congestion,the system adds a congestion warning mechanism to improve the efficiency of congestion detection,reduce the probability of network congestion.Experimental results show that this strategy achieves the purpose of reducing the average transmission delay and packet loss rate of the network.Aiming at problems of large transmission delay and high packet loss caused by irrational resource allocation of SDN route calculation,an adaptive SDN routing algorithm based on genetic algorithm is proposed.The algorithm collects the network state information through the OpenFlow technology,which serves as a reference for the calculation of the forwarding path by the genetic algorithm.When designing the algorithm,the conditions of the crossover and mutation operations of the genetic algorithm are constrained to avoid generating invalid paths,thereby reducing the search space of the optimized solution and reducing the computational overhead.Meanwhile,in order to avoid the delay and distribution caused by the allocation of small data streams to the big data streams and the big data streams to the same link,the algorithm also quickly introduces the stream classification policy to achieve the classification of large and small data streams,and to improve the network performance.Experimental results show that the algorithm reduces the transmission delay and packet loss rate of the network to some extent.