AICL: Autonomous Intent-Driven Control Loop to Mitigate Intent Drifts in Software-Defined Networking

Intent-driven networks have emerged as a promising paradigm with wide applications to autonomous networks, which can refine high-level intents into executable network configurations. Current research primarily focuses on tracking performance indicators to address performance drift, which prevents networks from achieving expected performance metrics. However, semantic drift, which refers to the inconsistency between original intents and actual network configurations, is always overlooked. As a result, underlying configurations may fail to precisely reflect the original intent. There is a lack of a generic intent full lifecycle management framework to mitigate both performance and semantic drifts. In this article, we first introduce the Autonomous Intent-driven Control Loop (AICL) framework to illustrate the control loop design and key functions of intent full lifecycle management. Then, we present an intent representation and drift formulation to measure the semantic drift clearly. Our proposed framework ensures performance and semantic consistency through real-time monitoring and closed-loop validation. Extracting the semantics of network configurations can seamlessly bridge the gap between declarative intents and underlying policies. Finally, we demonstrate the feasibility and effectiveness of the AICL framework by conducting experiments in the software-defined networking environment.