Design and Implementation of a Scalable SDN-OF Controller Cluster

Min Luo, Quancai Li, Man Bo, Ke Lin, Xiaorong Wu, Chenji Li, Sheng Lu, Wu Chou
Software-defined networking (SDN) is a new paradigm to increase network resource utilization, minimize management complexity, and reduce operational cost. While the logically centralized control in SDN offers many unique advantages that can enable globally optimized routing and resource utilization, the controller in SDN needs to be carefully designed to avoid being a performance bottleneck or a potential single point of failure of the network. This paper presents a scalable clustering approach
more » ... for the design and implementation of SDN-OF controller. In the proposed approach, controller instances (threads, and processes on a single or multiple physical servers) will mostly run in equal mode for packet-in processing, and only few such instances will serve as the "master" to process special packets that may result in updating network states. A self-learning adaptive mechanism is incorporated to further optimize the number of the controller instances under the varying operating conditions to adapt to changes in network states, controller workloads, and controller data traffic flows dynamically. As a consequence, it automatically enables load balancing and fail-over protection. The proposed approach has been successfully implemented and tested under some stringent conditions, and the performance advantages are observed.