Research And Implementation Of Adaptive Flow Control Technology For SDN Network Based On Anomaly Detection

With the continuous development of network technologies,the introduction and application of technical concepts such as big data and cloud computing,the network load of data centers is large,and the network structure is becoming more and more complex.The emergence of a software-defined network architecture provides a viable solution for people to solve this problem.It provides a new way of thinking for traditional networks.The core idea is to separate the control layer from the forwarding layer.This novel architecture brings great convenience and flexibility to network management,while bringing new opportunity and challenge.The idea of simplifying network operation by SDN is commonly recognized by today's complex network platforms,greatly reducing the complexity of the network system and the burden of exchange and forwarding,so that the compatibility of each interface is greatly facilitated by users and managers.This paper studies the characteristics of SDN architecture and related technical framework,and builds a distributed experimental environment integrated by OpenDayLight and OpenStack.By detecting the abnormal conditions in the network and then feeding back to the controller,the SDN architecture can flexibly deliver and modify the characteristics of the policy,and implement the related technologies for dynamically controlling network traffic.Based on the research of the commonly used static load balancing algorithm and some dynamic load balancing schemes proposed by the academic community according to the characteristics of SDN architecture,this paper proposes a load balancing scheme based on server anomaly feedback,which effectively guarantees the network service quality.At the same time,the network load of the controller is greatly reduced compared to the dynamic load balancing proposed before.Finally,this paper designs and implements a security protection method in which the intrusion detection program with machine learning function is placed in the network as a security device in the SDN network architecture.An improved k-means algorithm was implemented using Mllib on the Spark platform.The deployment method of separating the training module from the detection module is used,the system resources occupied by the detection program are reduced,and finally the flow table is modified by the abnormal traffic information reported according to the abnormality detection program,and the abnormal traffic is cut off,thereby utilizing the SDN network.The advantages enable dynamic security policy modifications.