Controller selection and load balancing in software defined network using multi criteria decision making methods

Abstract

A Software Defined Network (SDN) is a new approach to simplify newlinenetwork management and enable network programmability by decoupling the newlinecontrol plane from the forwarding plane. Logically centralized physically newlinedistributed controller architecture emerges as a result of the issues of scalability newlineand reliability suffered by a single, centralized controller. In order to implement newlinean SDN application, controllers are the fundamental components, however, newlinemany controllers exists such as POX, RYU, Beacon and Floodlight. The main newlinegoal of the thesis is to study the selection of best controller and balance the load newlineamong controllers that satisfy multiple objectives for obtaining load-balanced newlineefficient network. Initially, the research work focuses on the problem of newlinechoosing the appropriate controller, for a load balancing application. Here, a newlineMulti Criteria Decision Making method (MCDM) named Entropy based newlineTechnique for Order Preference Similarity to Ideal Solution (TOPSIS) method is newlineused to select the best controller for the load balancing mechanism. The network newlineemulator called Mininet is used to evaluate among four controllers (POX, RYU, newlineBeacon and Floodlight) over tree topology with different parameters such as newlinethroughput, delay, response time and communication cost. The Entropy based newlineTOPSIS method is used to rank the controllers where Floodlight is chosen as the newlinemost excellent one based on our criteria.