Research On Machine Learning-based SDN Network Traffic Situation Awareness Technology

In the past 30 years,the world is rapidly transforming into an information society.The rise and evolution of Internet technology has played a very important role in this process.As the network scale continues to increase,network management is facing many challenges such as a surge in network traffic and increased network load.How to efficiently and stably operate the entire network system has become the focus of social network managers today.In response to this,the industry has born a network traffic situation awareness technology that uses Software Defined Network(SDN)as a new network management method.This technology provides a lot of help to improve the management capabilities of network systems.The key to network traffic situational awareness is to extract,merge,analyze,and predict future trends based on traffic-related situational information in the network,and the results of the perception can assist network managers to make perception decisions to reduce potential risks or avoid losses.Combining the strong management and control capabilities of SDN network controllers on network traffic data and the outstanding advantages of machine learning methods in data analysis and data mining,this paper did a research on SDN network traffic situation awareness technology based on machine learning.The main work is as follows:(1)Research the traffic situation prediction technology,in order to solve the traditional support vector regression method training is sensitive to parameter adjustment,easy to fall into local shortcomings and other shortcomings,the gray wolf optimization algorithm is introduced into SVR training,and a gray wolf algorithm based optimizes the support vector regression situation prediction method is proposed,which is the GWOSVR(Gray Wolf Optimization,GWO)algorithm.It designs the parameters and fitness functions of the model in detail,gives a detailed prediction process,and then conducts experimental tests.Finally,comparison with other models verifies its superiority.(2)The multi-controller load balancing scheme based on traffic situation prediction to realize the knowledge application of the situation prediction results in the specific field of multi-controller load balancing and complete the simple situation decision prototype design.This solution combines the traffic situation prediction algorithm and the predicted load value to determine whether the load balancing degree of the control plane is lower than the threshold.If it is judged that load unevenness may occur,pre-processing is performed in the form of replacement migration to avoid overload controller and affect network performance.Related research mostly adopts overload post-processing method,and simulation results show that the overload pre-processing scheme in this paper can effectively improve the load balancing degree of SDN control plane.(3)Based on the research on situation prediction technology and multi-controller load balancing scheme,the system needs analysis,the design and implementation of network traffic situation data collection system,build SDN network experiment platform,design and implement each sub-module,and then complete the entire SDN Network traffic situation awareness system.