Distributed Software-Defined Networking Management

Abstract

Distributed software-defined networking (SDN) architecture satisfies the minimum requirements for WANs. The distributed controllers are connected in various topologies, including hierarchical and flat, which include logically centralized, physically distributed, and fully distributed controllers. The distributed SDN architectures are qualitatively explored as a more suitable solution for managing fluctuating networks in large-scale deployments, with the goal of optimizing overall network performance, particularly for applications that can tolerate some level of inconsistency, such as load balancing or routing. The logically centralized, physically distributed SDN controller architecture allows SDN controllers, in conjunction with the deployed SDN applications, to centrally coordinate the network due to the conciliated global network view. That is created through the synchronization process between controllers. However, inter-controller synchronization creates an overhead that affects the system’s performance. Additionally, the amount of inter-controller synchronization is vulnerable to the chosen consistency approach the application can tolerate. Although static eventual consistency is frequently employed in modern SDN systems to provide effective scalability, it is argued that it doesn’t place limits on the state inconsistencies that SDN applications will tolerate. Hence, the adaptive consistency models need to be investigated. The study showed that a flat, logically centralized, physically distributed architecture with an adaptive consistency approach would be more suitable for solving large-scale fluctuating network management considering scalability, reliability, and maximizing performance.

Keywords

• Adaptive consistency
• Distributed SDN architecture
• Large-scale networks
• Logically centralized physically distributed controllers
• Network performance optimization