Abstract: By dividing the WAN (wide area network) into several SDN (software defined networking) control do-mains with smaller scale and balanced number of nodes, the service capability of the controller can be improved. However, there are still many challenges in determining of SDN control domains optimal number and partitioning of SDN control domains according to the SDN network topology. In this paper, a mechanism of determination and partition of SDN control domains is proposed. It uses the spectral clustering algorithm to solve the problem of determination and partition of SDN control domains with the target of balancing the controller load. Firstly, SDN control domains determination mechanism is designed based on eigengap spectral clustering algorithm to determine the optimal number of control domains. Then, SDN control domains partition mechanism based on normalized Laplacian matrix eigenvector is designed to make the number of switch nodes in each control domain relatively balanced. Finally, the simulation results show that the mechanism of determination and partition of SDN control domains proposed in this paper can determine the optimal number of control domains, and the relative balance of the number of switch nodes in each control domain can be achieved by partitioning SDN control domains.

Key words: software defined networking (SDN), control domains, load balancing, spectral clustering algorithm

摘要： 通过将广域网划分成若干规模较小、节点数量均衡的软件定义网络（SDN）控制域，可以提高控制器的服务能力。然而，根据SDN网络拓扑进行控制域的最优数量确定与划分仍然面临许多挑战。提出了一种SDN控制域确定与划分机制，以控制器负载均衡为目标，基于谱聚类算法解决控制域确定与划分问题。首先，基于本征间隙谱聚类算法设计了SDN控制域确定机制，用于确定控制域的最优数量；然后，设计了基于归一化Laplacian矩阵特征向量的SDN控制域划分机制，使得各控制域所拥有的交换机节点数量相对均衡；最后，仿真结果表明，提出的SDN控制域确定与划分机制能够确定最优的控制域数量，并且通过SDN控制域划分可实现各控制域交换机节点数量的相对均衡。

关键词: 软件定义网络（SDN）, 控制域, 负载均衡, 谱聚类算法