Dynamic Controller Placement in Software Defined Networks using Spectral Clustering

Abstract

This paper addresses the controller placement problem in Software-Defined Networks (SDNs), which involves determining the optimal locations for multiple controllers in a large network. SDN is a modern network technology that enhances network management efficiency by separating the control plane from the data plane. In SDN, devices in the data plane become forwarding devices, while the decision-making processes are centralized in a controller within the control plane. The placement of controllers is critical in SDN design and directly impacts various network performance metrics such as latency, flow setup time, network availability, controller load balancing, and energy consumption. To address the controller placement problem, this work proposes an algorithm that dynamically determines the arrangement of K controllers with limited resources in a cluster of hosts and switches. The approach involves efficiently clustering switches and assigning a controller to each group. Spectral clustering is employed as the primary technique in this work. The proposed work is implemented using the Mininet emulator with the POX SDN controller. To evaluate its performance, the results are compared with other clustering algorithms, including K-means and K-medoids. The results demonstrate that the spectral clustering-based mechanism performs better than K-means and K-medoids in terms of throughput, delay, and jitter, indicating its superiority in achieving efficient controller placement for SDNs.