Research On Multilink Load Balancing Strategy Based On SDN

With the rise of new business such as cloud computing and big data,the network scale and the number of network users show an explosive growth.As an important infrastructure for IT services,data centers are increasingly burdened with the network they carry.How to achieve balanced network load distribution and reduce network congestion has become an urgent problem for data centers.Simply improving the traditional TCP/IP network architecture can no longer meet current data center network requirements.In this big background,the new network architecture SDN emerges as the times require.In essence,the SDN system is a network layered architecture,its core idea is to realize the decoupling of the control layer and the data layer,and to complete the centralized control of all functions of the network by the SDN controller.Compared to the distributed control of traditional network architecture,the centralized control function of SDN network and the software programmable method greatly increase the flexibility of the network.The introduction of SDN into data center to alleviate the traffic pressure of data center has attracted more and more attention.The fat tree topology widely used in current data centers can provide multiple equal-cost paths for network data flow transmission.The multi-path characteristics of this topology can improve network throughput and bandwidth.However,the traditional routing algorithm ECMP used in the data center only distributes the data stream approximately equally across the number of equal-cost paths.This algorithm is not suitable for the complex and changeable data center traffic.The centralized architecture advantage of the software-defined network enables the network to dynamically adjust the traffic distribution according to the real-time conditions so that the multi-path characteristics can be fully utilized.Therefore,this paper will study the multilink load balancing strategy in SDN network environment.In this paper,the controller Ryu is chosen as the platform for the algorithm implementation.Through the controller design and implementation of three custom modules: information collection module,routing module,flow table forwarding module to complete the program deployment.First,the complete topology of the network is obtained through the information collection module,and the real-time status of each link in the topology structure is monitored centrally.Then the routing module implements the selection of k-reliable path between the source host and the destination host through the optimal k-short path algorithm,and formulates an integrated evaluation model to integrate the real-time network status information collected by each functional module,and the obtained result is short for k.The paths are prioritized to get the most favorable transmission path for current network load balancing.Finally,the traffic forwarding module delivers corresponding data flow processing decisions to all the switching devices on the selected link.The strategy can dynamically adjust network load according to the real-time state of network,alleviate network congestion effectively,optimize network performance and achieve efficient data forwarding.Simulate the Fat-Tree topology on the simulation platform Mininet,and connect it to the Ryu controller with multilink load balancing strategy.The real data center traffic model is simulated under the complete SDN network architecture,and the scheme is tested.By comparing the bandwidth utilization,delay and throughput of different algorithms,it is proved that the load balancing effect is better when the data center traffic is large,the transmission delay is reduced and the bandwidth utilization rate of the network is increased.