Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Towards End-to-End Network Intent Management with Large Language Models

Lam Dinh, Sihem Cherrared, Xiaofeng Huang, Fabrice Guillemin

TL;DR

This paper investigates end-to-end network intent management using large language models to translate user intents into RAN and Core network configurations for 5G/6G. It introduces FEACI, a five-daceted metric for Format, Explainability, Accuracy, Cost, and Inference time to evaluate LLM-generated deployment plans, and compares closed-source models (Gemini 1.5 Pro, GPT-4) with open-source ones (Llama, Mistral) under ZERO/ONE/FEW prompting. The authors propose an ibn framework with Business Intent Resolver and Service Intent Resolver aligned with TM Forum ODA, translating natural-language intents into TMF-formatted cfs and rfs outputs that feed the Network Resource Orchestrator. Results show open-source models achieving comparable or superior translation performance while offering lower hardware and cost barriers, particularly under few-shot prompting.

Abstract

Large Language Models (LLMs) are likely to play a key role in Intent-Based Networking (IBN) as they show remarkable performance in interpreting human language as well as code generation, enabling the translation of high-level intents expressed by humans into low-level network configurations. In this paper, we leverage closed-source language models (i.e., Google Gemini 1.5 pro, ChatGPT-4) and open-source models (i.e., LLama, Mistral) to investigate their capacity to generate E2E network configurations for radio access networks (RANs) and core networks in 5G/6G mobile networks. We introduce a novel performance metrics, known as FEACI, to quantitatively assess the format (F), explainability (E), accuracy (A), cost (C), and inference time (I) of the generated answer; existing general metrics are unable to capture these features. The results of our study demonstrate that open-source models can achieve comparable or even superior translation performance compared with the closed-source models requiring costly hardware setup and not accessible to all users.

Towards End-to-End Network Intent Management with Large Language Models

TL;DR

This paper investigates end-to-end network intent management using large language models to translate user intents into RAN and Core network configurations for 5G/6G. It introduces FEACI, a five-daceted metric for Format, Explainability, Accuracy, Cost, and Inference time to evaluate LLM-generated deployment plans, and compares closed-source models (Gemini 1.5 Pro, GPT-4) with open-source ones (Llama, Mistral) under ZERO/ONE/FEW prompting. The authors propose an ibn framework with Business Intent Resolver and Service Intent Resolver aligned with TM Forum ODA, translating natural-language intents into TMF-formatted cfs and rfs outputs that feed the Network Resource Orchestrator. Results show open-source models achieving comparable or superior translation performance while offering lower hardware and cost barriers, particularly under few-shot prompting.

Abstract

Large Language Models (LLMs) are likely to play a key role in Intent-Based Networking (IBN) as they show remarkable performance in interpreting human language as well as code generation, enabling the translation of high-level intents expressed by humans into low-level network configurations. In this paper, we leverage closed-source language models (i.e., Google Gemini 1.5 pro, ChatGPT-4) and open-source models (i.e., LLama, Mistral) to investigate their capacity to generate E2E network configurations for radio access networks (RANs) and core networks in 5G/6G mobile networks. We introduce a novel performance metrics, known as FEACI, to quantitatively assess the format (F), explainability (E), accuracy (A), cost (C), and inference time (I) of the generated answer; existing general metrics are unable to capture these features. The results of our study demonstrate that open-source models can achieve comparable or even superior translation performance compared with the closed-source models requiring costly hardware setup and not accessible to all users.

Paper Structure

This paper contains 16 sections, 1 equation, 9 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related works
  3. Intent extraction
  4. Intent activation/assurance
  5. Our approach
  6. End-to-End Network Intent Management
  7. Network Intent Architecture
  8. Large Language Models
  9. From natural language to network deployment with llms
  10. Prompting
  11. LLM evaluation for intent translation/resolution - FEACI
  12. Results and discussion
  13. Scenario description
  14. Simulation setup
  15. Numerical results
  16. ...and 1 more sections

Figures (9)

  • Figure 1: Network architecture with Intent layer
  • Figure 2: Translation and resolution block with business and service resolver
  • Figure 3: Illustration of an llm workflow
  • Figure 4: Intent translation and resolution with llms
  • Figure 5: Zero-shot prompting (ZERO)
  • ...and 4 more figures