Leveraging generative AI for intent-based networking operations in network slices

Large Language Models (LLMs) are among the most popular Generative AI models. They bring benefits like a natural language interface and the automation of complex tasks. Despite their potential, few studies have implemented LLMs for network management. This paper addresses that gap, showcasing in a p...

Full description

Bibliographic Details
Authors: Adanza, D, Gifre, L, Alemany, P, Natalino, C, Monti, P, Muñoz, R, Vilalta, R
Format: article
Status:Published version
Publication Date:2025
Country:España
Institution:Centre Tecnològic de Telecomunicacions de Catalunya (CTTC)
Repository:r-CTTC. Repositorio Institucional Producción Científica del Centre Tecnològic de Telecomunicacions de Catalunya (CTTC)
OAI Identifier:oai:cttc.fundanetsuite.com:p8859
Online Access:https://cttc.fundanetsuite.com/Publicaciones/ProdCientif/PublicacionFrw.aspx?id=8859
Access Level:Embargoed access
Keyword:Large language models
Software-defined networking
Intent-based networking and network slices
Description
Summary:Large Language Models (LLMs) are among the most popular Generative AI models. They bring benefits like a natural language interface and the automation of complex tasks. Despite their potential, few studies have implemented LLMs for network management. This paper addresses that gap, showcasing in a practical scenario how network management can be efficiently enhanced by automating tasks such as network configuration and traffic analysis, thereby reducing downtime and improving efficiency. This study presents a LLM agent integrated with a cloud-native SDN controller (ETSI TeraFlowSDN) designed with Retrieval-Augmented Generation (RAG) capabilities to operate with intent-based network operations. The LLM agent understands the context and triggers different operations, such as intent creation, query, and explanation. The results demonstrate a system capable of automating network operations with a factual accuracy of 93% with reasonable computation times, demonstrating how the developed LLM agent can enhance network management.