Adaptive Network Configuration using Intents in Dynamic Networks

Abstract

In modern healthcare environments, reliable and adaptive networking is critical to ensuring the continuity of patient care, telemedicine, and hospital operations. Traditional static network configurations are inadequate for handling the dynamic and diverse traffic requirements of medical applications. This thesis presents an intelligent, intent-based network configuration system that leverages Software-Defined Networking (SDN), Network Function Virtualization (NFV), and Generative AI to automate network slicing and service delivery in hospital infrastructures. The proposed system allows users to express high-level network intents in natural language, which are interpreted using a large language model API and translated into precise configurations for SDN flow control and VNF deployment. Using ONOS as the SDN controller and Mininet for emulation, the system dynamically creates and isolates slices for critical use cases such as patient monitoring, video consultations, and administrative services. Each slice is provisioned with bandwidth policies tailored to its operational needs. Additionally, Monitoring VNFs are deployed per slice to validate real-time performance and ensure compliance with user-defined intents. The system achieves automated intent assurance, demonstrating its capability to maintain quality of service (QoS) under dynamic conditions. Experimental evaluation shows that the architecture reliably enforces traffic isolation, prioritization, and feedback-based monitoring within an emulated hospital network.