Avoiding SDN Application Conflicts With Digital Twins: Design, Models and Proof of Concept

Software-Defined Networking (SDN) enables flexible and programmable control over network behavior through the deployment of multiple control applications. However, when these applications operate simultaneously, each pursuing different and potentially conflicting objectives, unexpected interactions may arise, leading to policy violations, performance degradation, or inefficient resource usage. This paper presents a Digital Twin (DT)-based framework for the early detection of such application-level conflicts. The proposed framework is lightweight, modular, and designed to be seamlessly integrated into real SDN controllers. It includes multiple DT models capturing different network aspects, including end-to-end delay, link congestion, reliability, and carbon emissions. A case study in a smart factory scenario demonstrates the framework's ability to identify conflicts arising from coexisting applications with heterogeneous goals. The solution is validated through both simulation and proof-of-concept implementation tested in an emulated environment using Mininet. The performance evaluation shows that three out of four DT models achieve a precision above 90%, while the minimum recall across all models exceeds 84%. Moreover, the proof of concept confirms that what-if analyses can be executed in a few milliseconds, enabling timely and proactive conflict detection. These results demonstrate that the framework can accurately detect conflicts and deliver feedback fast enough to support timely network adaptation.