Multi-Objective Optimization of Controller Placement in Distributed ONOS Networks.

In distributed Software Defined Networking (SDN), multiple controllers need to maintain a consistent view of the network state among the controllers using consensus algorithms, which introduces additional communication overhead and network latency, especially in large-scale networks. Therefore, it is difficult to optimize the placement of controllers considering not only the delay between switches and controllers, but also the delay of consensus algorithms. This research utilizes Open Network Operating System (ONOS) as a platform for building distributed SDN, which employs an Atomix cluster to store and share the network state views across multiple controllers using a consensus algorithm. Therefore, the optimal placement problem in ONOS involves both the placement of ONOS controllers and the placement of Atomix nodes, which makes the problem more complex. In this research, we develop a practical model to estimate the Flow Setup Time (FST) as a measure of ONOS controller response time, and propose a method to determine the optimal placements of controllers and Atomix nodes by a multi-objective optimization. Moreover, we apply our method to several real-world networks on the Internet and visualize the experimental results. Based on the observed variations in FST across different network topologies, we provide a suggestion to improve ONOS performance.