Optimization of Multi-Controller Locations in SDWAN using Various Method

Abstract

Software Define Wide Area Network (SDWAN) is a solution for utilizing technology in sending information. SDWAN is designed by separating the control plane and data plane by applying the Software Define Network (SDN) concept; therefore, it can use physical devices more effectively and efficiently. With many nodes and spread, node grouping is needed to facilitate control and supervision that requires a controller in each cluster. Installing many nodes on SDWAN will require several controllers to make it more effective and efficient placement. Optimal controller placement will improve the performance of the network. In this study, determining the optimal controller location requires various methods that are interconnected with each other. The algorithms used are the Haversine method, Johnson's Algorithm, Partition Around Medoids (PAM), and Silhouette Analysis. The number of nodes and locations obtained from Zootopology in this research using the Indonesia Biznet network, therefore the recommendation for the optimal number of controllers is obtained using the Silhouette, Gap, Calinski-Harabasz, and Davies-Bouldien evaluation methods. The algorithms aim to get the optimal point by determining the number of controllers and recommendations for the optimal number of controllers as an initial recommendation for a company that uses this method. In this study, the most optimal number of controllers on the Biznet network with 29 nodes were two controllers and an average value of Silhouette analysis calculation of 0.51846.