Intent-Based Management of Next-Generation Networks: an LLM-Centric Approach
Abdelkader Mekrache, Adlen Ksentini, Christos Verikoukis
Abstract
Intent-Based Networking (IBN) management has emerged as an alternative approach to simplify network configuration and management by abstracting the complexities of low-level configurations. Existing IBN solutions typically rely on human-readable structures like JSON or YAML to define Intents, which still require expertise in understanding these structures. A natural evolution of IBN is to use natural language instead of defined structures. However, this approach introduces complexities related to natural language understanding. Fortunately, Large Language Models (LLMs) offer a promising solution. In this paper: (i) We propose a novel LLM-centric Intent Life-Cycle (LC) management architecture designed to configure and manage network services using natural language. The architecture spans the complete Intent LC, encompassing decomposition, translation, negotiation, activation, and assurance; (ii) We identify key open issues and challenges related to IBN within our proposed architecture; (iii) We demonstrate the effectiveness of the architecture by developing a component within the EURECOM 5G facility <xref ref-type="bibr" rid="ref1" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[1]</xref>, leveraging LLMs to implement the essential Intent LC procedures; (iv) We validate the proposed system through realworld deployment, showcasing its capability to define, decompose, translate, and activate Intents using natural language.