Research On Routing Algorithm For Load Balancing Based On SDN

With the rapid development of computer network,the data transmitted in the network becomes more and more complex,and the data center network will face the challenges of topology discovery,routing scheduling,load balancing and so on.As a new generation of network architecture,Software-Defined Networking(SDN)provides a good way to solve the problem of path congestion caused by the growth of network traffic because of its centralized network management and programmability.In order to achieve the goal of reasonable distribution and transmission of network traffic,this thesis studies the routing algorithm and strategy of network load balancing based on SDN technology.Aim at the different working stages of SDN,we reduce the load on controller by using improved topology discovery mechanism at the stage of topology discovery,we solve the problem of network congestion by using efficient dynamic link load-balancing routing algorithm and find the optimal path allocation scheme by calculating the traffic change quantity of the rerouting at the stage of routing planning,we avoid the transmission interruption of network traffic in the rerouting process by using priority-constrained flow table updating strategy at the stage of flow table updating.Firstly,this thesis introduces the research background and significance of the subject,the current situation of studies both at home and abroad and the contents of the thesis.Then,the thesis gives a brief overview of the architecture and related technologies involved in this thesis.Then,an improved topology discovery mechanism based on link layer discovery protocol is proposed,by evaluating the cost and overhead of topology discovery implemented by current mainstream SDN controller.It reduces the load on controller by reducing the number of messages which are sent and processed by the controller.And the experimental results show that the improved topology discovery mechanism greatly improves the performance of topology discovery.Next,a dynamic link load balancing routing algorithm is proposed.At the initialization stage of routing,the algorithm computes all possible paths between switches and select the initial path according to the link remaining bandwidth.During the stage of dynamic path optimization,the algorithm checks the load balancing state of the network periodically.Once load imbalance is detected,a multi-link algorithm is invoked to reroute the path of data flow whose path covers the link set of minimum remaining bandwidth.If still cannot balance load,a single-link algorithm is invoked to reroute the sub-path of data flow whose path covers the link of minimum remaining bandwidth.The experimental results show that the proposed routing algorithm can achieve higher link average bandwidth utilization and data flow average throughput compared with the current LABERIO algorithm.And a priority-constrained flow table updating strategy is proposed to ensure that the flow table entries are updated in the correct order in the rerouting process.It effectively avoids the interruption of data flow transmission.Finally,the possibility of getting multiple optimal solutions is realized in the simplex algorithm of linear programming.And a method of calculating the traffic change of rerouting is proposed.In order to find the best solution with the smallest change of traffic among the multiple optimal solutions,we get the total traffic change by accumulating the traffic change of each data flow in the network.This reduces the out of sequence during data packets transmission in the rerouting process.