Load Balancing Strategy for SDN Multi-controller ClustersBased on Load Prediction

Software-defined networking (SDN) separates the control layer from the data layer, and decisions tomanage the network are issued through a controller. The distributed SDN architecture is an effectivesolution addressing modern WAN SDN architectures and allows multiple controllers to managedifferent parts of the network to ensure efficient and stable operation. To solve the problems of highswitch migration cost, load imbalance, and inefficient load balancing in SDN multi-controller environments,we propose a deep learning-based controller load prediction switch migration (LPSM) strategythat uses a migration switch selection algorithm, target controller selection algorithm, and switchmigration decision algorithm. Then, we propose a load balancing algorithm based on this decisionalgorithm. The final experimental results show that the LPSM reduces the migration cost by 16%and 8%, respectively, compared with time-sharing switch migration (TSSM) and distributed decisionmigration (DDM) strategies, reduces load variance from 0.02 to 0.004 compared with the DDMstrategy, and improves load balancing efficiency by 27.6% compared with the TSSM strategy.