Redundancy Method for Highly Available OpenFlow Controller
Keisuke Kuroki, Masaki Fukushima, Michiaki Hayashi, Nobutaka Matsumoto · 2014
Abstract—OpenFlow is an important element for achieving Software Defined Networking (SDN) and is expected to be an enabler that solves the problems of today’s network. Thanks to the centralized management with OpenFlow, agile network operation can be achieved with flexible programmability; how-ever, the centralized management implies a significant impact of any outages of the OpenFlow controller. Hence, a high availability technology is indispensable for building the Open-Flow controller. To achieve the highly available system, we have to consider extraordinary events (e.g., power outage) af-fecting the entire data center as well as anticipated server fail-ures within a local system. In this paper, we review the issue in using the conventional redundancy method for OpenFlow con-trollers. Based on this observation, we propose a redundancy method considering both local and global (i.e., inter data-center) recoveries using the multiple-controllers capability that is defined in OpenFlow switch specification version 1.2 and later. The proposed redundancy scheme eliminates virtual IP address-based redundancy and frontend server causing limita-tion of performance scalability, while it achieves competitive role change and failover times. Keywords-OpenFlow; controller; redundancy. I.