Software Defined Networks (SDN) divide network intelligence and packet forwarding functionalities between control plane and data plane devices respectively. Multiple controllers need to be deployed in the control plane in large SDN networks to improve performance and scalability. In
a multi-controller scenario, finding the adequate number of controllers and their load distribution are open research challenges. In a large-scale network, the control plane load balancing is termed a controller placement problem (CPP). It is observed that of the existing solutions for the
CPP, clustering-based approaches provide computationally less intensive solutions. The proposed augmented affinity propagation (augmented-AP) clustering identifies the required number of network partitions and places the controllers such that the distribution of switches to the controller
is much better than with existing algorithms. The simulation results show that the computed controller imbalance factor of augmented-AP algorithm outperforms the existing k-means algorithm.
Evaluating and Comparing the Performance of Using Multiple Controllers in Software Defined Networks