Improving distributed controller failure using monitoring and clustering techniques

Abstract

Software Defined Networking (SDN) has become a leading approach for centralized control in advanced networking use cases such as cloud computing. In very large scale network system, it can reach thousands of servers and network devices like switches are even much more. Using single controller now can easily get troubles, and it seems to be exhausted in management. In order to increase fault tolerance and to eliminate bottleneck possibility of single controller, multiple controllers are necessary for SDN and distributed architecture is the one that highly evaluated. But in fact, distributed SDN still got a lot of challenges to be complete, one of them is the consistency. In this thesis, a robustness distributed SDN architecture that implements Atomix framework [1], with some additional features, is proposed.