Modeling Control Traffic in Distributed Software Defined Networks

Abstract: Software defined networks (SDNs) have evolved as novel network architecture in modern networking. However, to allow SDN scale for several thousand switches, the control architecture needs to be distributed, with control traffic between switches and controllers, as well as among the controllers. The aim of this master thesis is to develop a mathematical model of the control traffic in this emerging distributed SDN. The work includes the evaluation of the increase in control traffic with the size of the network and the effect of the network topology on the scaling of the control traffic, and the investigation of the benefits of optimizing the number of controllers. Mathematical model of inter-controller traffic and switch to controller traffic has been developed based on previous measurement works on SDN control protocols. This model has been applied to two basic network topologies and the results have been compared. Our model depicts that a random network which exhibit small world property has much lower traffic than regular topologies, and that by optimizing the cluster size the control traffic can be reduced from several Exabits/s to a few Pbits/s.