Optimized K-Medoid Algorithm Based Controller Placement In A Software Defined Network: In Case Of Ethio-Telecom

Abstract

A Software Defined Network (SDN) is a new paradigm of computer networking that separates control plane and data plane to promote logical centralisation of network control and central network programming ability. The performance of this type of network is affected by a way of controller placement. The existing algorithm consumes a lot of time in initializing the medoids. Here in this study, we have optimized the existing K-medoid algorithm to improve the performance of the network, to determine a minimum number of controllers, to identify optimum positions of controllers by balancing a load between controllers. To make the optimized K-medoid algorithm efficient optimal, we have used Knuth algorithm instead of random permutation to initialize medoids. The proposed method improves the average latency between nodes and controllers by 6.15% for Pareto Simulated Annealing (PSA). Maximum N2C latency is improved by 4.2 % for PSA, and by 13.31 % for standard K-medoid algorithms. Average C2C latency is improved by 3.11 % and 1.82 for PSA and standard K-medoid algorithms respectively. Lastly, the maximum C2C latency for PSA and standard K-medoid algorithm is improved by 14.23 % and 13.78 %, respectively. We have concluded that the optimized K-medoid algorithm has the best improvement on the evaluated performance metrics, and it determines the minimum number of controllers for the topology in comparisons with the two methods