A min-cover based controller placement approach to build reliable control network in SDN

Software defined network (SDN) develops a centralized control plane to manage the whole network. If the scale of the network is large, it is necessary to deploy multiple distributed controllers. In SDN, a switch can only work by relying on flow tables received from its controller. Therefore, controller placement is an important problem to keep the switches working efficiently and improve the reliability of the control network, which consists of controllers, switches and the communication paths between them. However, the existing controller placement approaches are not effective or do not consider the network reliability and the required delay between switches and controllers at the same time. In order to ensure the reliability of the control network and meet the required propagation delay, a min-cover based controller placement approach is proposed. Two metrics are proposed to measure the reliability of a control network, and the definitions of neighborhood and min-cover are provided, based on which the approach try to use less controllers to achieve the reliability and low delay of the control network while guaranteeing the manageability of the network. Simulations show that min-cover based approach can use as less controllers as possible to ensure the reliability of control network and satisfy the required delay at the same time. Moreover, the approach has steadily good performance in networks of different scales and connectivity.

I. Introduction

Software defined network (SDN) is one of the most promising architectures of future network. In SDN, the complex functions are moved to a centralized control plane, which provides a global view of the whole network and a programmatic interface to control the network. The controller can compute flow tables for the switches about how to forward the data packets, while the switches are just responsible for forwarding the packets. By this way, SDN can improve network resource utilization, simplify network management, and reduce operating cost [1]–[2].