Reducing Latency in SDN: An Innovative Closed-Loop Migration Collaboration Architecture

Abstract

Due to the asynchronous nature of software-defined networking (SDN), existing flow installation models face a common issue: data packets often reach intermediate switches before the corresponding flow rules are installed. In such cases, the switch either discards the packets or forwards them to the controller, leading to bandwidth waste and increased flow latency; this is unacceptable in networks with stringent latency requirements. This article proposes a new architecture that enables low-latency flow transmission and is compatible with all flow installation models. It achieves a closed-loop of flow migration between the control and the data plane. This greatly enhances the flexibility and resilience of SDN. We present two solutions: first, we scientifically tested the approach's feasibility through experiments, identifying the balance point of controller flow load. Second, we use a simple and existing optimization strategy to reduce SDN resource consumption, improving performance to meet practical application standards. It reduces latency by 66.03% compared with the traditional method and 73.47% compared with the EFastLane method. To our knowledge, we are the first to explore how to implement flow migration to the controller. This architecture is compatible with all flow installation models and can be used in various scenarios to improve the efficiency and reliability of SDN. Our findings should facilitate further progress in deploying SDN in real-world networks.