Research And Implementation Of SDN Switch Configuration And Control Technologies

Software Defined Networking(SDN) technology separates the control plane and the data plane, which satisfies application requirements with good controllability and manageability. The centralized control mechanism of SDN provides a global view of network, which has brought a new approach for the research of the data center network load balancing. With the introduction of SDN technology, some new load balancing algorithms have emerged. These algorithms have concerned the global load balancing performance of SDN networks, but they have ignored that load balancing scheduling may affect Quality of Service(QoS).This thesis proposes analgorithmof controlling SDN switchbased on QoS and dynamic load balancing, SAS(Scheduling According to Stickiness), in which the concept of link stickiness value is proposed andusedin order to estimate the level of dispatch influenceon flow’s performance. By prior scheduling lowstickiness value flow, the effect on flow performance is reduced, and the purpose of optimizing scheduling is achieved. The approximation algorithms to multi-constrained routing are used, to improve bandwidth utilization while the QoS requirementof flows is considered, and the influenceonthe flow performance is effectively reduced.This thesis designs and implements a prototype system based on Floodlight, an open control platform.Aiming at two test targets, including the load_balancing effect and its generated flows’ QoS, the test approach is designed and implemented. Experimental test results show that in case of different QoS requirement flows competing for limited network resources and resulting in network load deviations from the balance state, the technology approach proposed in this thesis can improve the bandwidth utilization while fufiling the requirements of QoS on flow, and reduce the influence on flow’s performance.