GOKA: A Network Patition and Cluster Fusion Algorithm for Controller Placement Problem in SDN

Software Defined Networking (SDN) is an emerging network architecture. It decouples the data plane from the control plane, centralizes the network structure logically and separates it physically, making the network flexible to manage. However, with the increasing network scale of SDN in recent years, it is difficult for a single controller to manage a huge network, easily producing a single point of failure that leads to the network performance reduction even paralysis. As a result, a multi-controller structure emerges and the Controller Placement Problem (CPP) is proposed. With the goal of shortening the latency between the controller and switch as much as possible, we propose Greedy Optimized K-means Algorithm (GOKA) to solve this deployment problem. By comparing with the optimized K-means algorithm and the Brute Force (BF) algorithm, the effectiveness of GOKA is verified. Through the experimental results, GOKA is more stable than optimized K-means algorithm, and promises an optimal or near-optimal solution with a less than 2% margin.