Model Design And Research On Traffic Controlbased On SDN

With the development of diversification of network applications, the increase of network traffic in number and the sustained growth of bandwidth requirement, as an important mechanism to optimize network performance,traffic control engineering has been paid more and more attention by network operators. The current flow control methods of IP networks cannot meet the QoS needs of different services due to the limitations of routing, which leads to bottlenecks in the research of flow control method of the existing networks. SDN is proposed to build a new kind of network, by separatingdata plane and control plane. According to the entire network topology and status information, the controller can find the best next hop to forward data to ensure the QoS needs for different network applications. This paper focused onresearching traffic control method based on SDN.First of all, the SDN and its main communications protocol OpenFlow were introduced in this paper, the forwarding of these data flow was described in details, and the existing traffic control method in SDN was analyzed.Then the management planning and traffic control model on both data layer and control layer were established separatelybased on the uneven distribution of resources and the congestion problems on the data plane in existing SDN traffic.Secondly, the traffic control methodon SDN control plane was researched.By analyzing the forwarding characteristics of the control layer,it can be found that theseissues are similarwith the maximum concurrent flow in the conventional multi-commodity flow model. In this paper, the multi-commodity flow algorithm based on multi-path mapping is applied to the traffic control engineeringof SDN control plane.The virtual network mapping mathematical model was established based on the OpenFlow, and using the multi-commodity flow algorithm, this model was solved.The two multi-commodity flow algorithm based on single-path mapping and multi-path mapping were compared by simulation experiments to prove that multi-path mapping algorithm has better performance, higher resource utilization, and more acceptation of service requests.Then, queue managementalgorithm was used in the data plane. RED is the most popular algorithm onthe prevention of network congestion, and further measures can be taken promptly when the unexpected traffic congestion happened in the network. After analyzing the deficiencies of the RED and its improved algorithms in packet prioritization, fairness and other aspects, a new algorithm based on packet priority was proposedfor SDN dataplane in this paper. This model classifies different data flow into different groups which marked by different colors.The different groups were maneged with the same set ofparameters, and the packet arrival and departure impact on the average queue length was considered at the same time.Simulation results show that the loss rate, delay etc. calculated based on this model canmeet the various needs of the network.Finally, this paper summarized the SDN traffic control technology,and lookedforward to several aspectsof future work. Itpointed out thatalthough many method have been proposed, the research in SDN traffic control are still in the initial stage, and its reliability and scalability still need further improvement.