Research On QoS Guarantee Mechanism For Network Traffic Based On SDN

With the rapid development of Internet technology,many new network applications and services have emerged,and the new network applications have put forward differentiated demands on the network transmission process in terms of latency,bandwidth,packet loss rate,etc.How to ensure the quality of service(QoS)for different types of network applications has become an urgent problem.Traditional network architectures have limited global views and insufficient control over service flows,and their best-effort network transmission mechanisms can no longer meet the QoS requirements of different service types of network applications.The emergence of Software Defined Network(SDN)brings a good solution to this problem,with a global view of the network available in real-time at the control level and the ability to control globally,enabling flexible QoS guarantee mechanisms.To meet the QoS requirements of different traffic service types in SDN networks,this paper proposes an SDN-based QoS guarantee mechanism for network traffic,with the following three main research elements:1)The SDN-based network traffic classification method is designed.Convolutional neural network technology is introduced and combined with the SDN-based network traffic classification architecture to classify traffic into different service types according to the different QoS requirements of applications;a differentiated service model is introduced to label different network traffic services.introducing nearest neighborhood component analysis and stochastic gradient descent,a feature dimensionality reduction algorithm is designed,and effective features are used for machine learning algorithm performance testing and model training.The experimental results show that the convolutional neural network algorithm can perform better classification and thus improve the network performance.2)The SDN-based network traffic queue scheduling mechanism was established.Combining SDN technology and network traffic classification methods,the SDN-based queueing scheduling model is designed;M/M/1 and M/D/1 queuing theories are introduced under the new model,and the SDN-based queueing theory end-to-end delay model is established;by analyzing common queueing scheduling algorithms,the multi-level feedback queuing algorithm has better scheduling performance.Experimental results show that the mechanism can meet the QoS requirements of network traffic of different service types to prevent traffic congestion in controller queues.3)The SDN-based network traffic routing planning strategy is proposed.Combining network traffic classification methods and queue scheduling mechanisms,the SDN-based realtime routing planning framework is designed to guarantee the overall performance of multitype network traffic routing and forwarding.QoS routing algorithms based on adaptive genetic ant colony and deep reinforcement learning are designed.Experimental results show that both routing algorithms can achieve better QoS performance optimization to meet the QoS requirements of multiple services.