Efficient switch clustering for distributed controllers of OpenFlow network with bi-connectivity

Abstract

Control planes of Software Defined Networking have faced a scalability issue because they have been widely applied for large scale networks. To address this issue, control planes with multi-controllers such as Onix and HyperFlow have been proposed in which controllers share global network view. Since the global network view includes topology, states and events of a network, shared information could become larger when governed network is large and complicated. Thus, Onix have proposed reduction method of the information by abstracting switches as a single virtual switch. However, there is a lack of discussion about reliability of virtual switches that depends on topology of abstracted networks. To improve reliability of the virtual switches, there are two viewpoints: (1) administrative areas of controllers, and (2) limitation of abstraction areas. Hence, this paper proposes construction method of more dependable virtual switches focusing on bi-connectivity and reduction of shared information using the virtual switches. Our experimental results show that our reduction method can improve reliability of virtual switches and curb the number of edges in a federation graph that is regarded as shared information.