Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Large Language Models for Zero Touch Network Configuration Management

Oscar G. Lira, Oscar M. Caicedo, Nelson L. S. da Fonseca

TL;DR

The paper tackles the challenge of achieving zero-touch network management with privacy-preserving automation. It introduces LLM-NetCFG, a local, modular LLM-based system that translates operator intents into device configurations, verifies them, and deploys them with minimal human intervention, leveraging a Zephyr-7B beta model for on-site processing. A verification pathway using Batfish and a carefully designed orchestration layer ensure configuration correctness and feasibility before deployment. Empirical results from a prototype with 90 intents demonstrate strong classification accuracy (92.2%) but reveal challenges such as misclassification into an 'Other' class and longer verification cycles for complex intents, highlighting the need for further fine-tuning and robust prompt design. The work maps LLM capabilities to ZSM components, discusses future research directions, and underscores the significance of privacy, security, and end-to-end automation in autonomous networks.

Abstract

The Zero-touch Network & Service Management (ZSM) paradigm, a direct response to the increasing complexity of communication networks, is a problem-solving approach. In this paper, taking advantage of recent advances in generative Artificial Intelligence, we introduce the Network ConFiguration Generator (LLM-NetCFG) that employs Large Language Model and architects ZSM configuration agents by Large Language Models. LLM-NetCFG can automatically generate configurations, verify them, and configure network devices based on intents expressed in natural language. We also show the automation and verification of network configurations with minimum human intervention. Moreover, we explore the opportunities and challenges of integrating LLM in functional areas of network management to fully achieve ZSM.

Large Language Models for Zero Touch Network Configuration Management

TL;DR

The paper tackles the challenge of achieving zero-touch network management with privacy-preserving automation. It introduces LLM-NetCFG, a local, modular LLM-based system that translates operator intents into device configurations, verifies them, and deploys them with minimal human intervention, leveraging a Zephyr-7B beta model for on-site processing. A verification pathway using Batfish and a carefully designed orchestration layer ensure configuration correctness and feasibility before deployment. Empirical results from a prototype with 90 intents demonstrate strong classification accuracy (92.2%) but reveal challenges such as misclassification into an 'Other' class and longer verification cycles for complex intents, highlighting the need for further fine-tuning and robust prompt design. The work maps LLM capabilities to ZSM components, discusses future research directions, and underscores the significance of privacy, security, and end-to-end automation in autonomous networks.

Abstract

The Zero-touch Network & Service Management (ZSM) paradigm, a direct response to the increasing complexity of communication networks, is a problem-solving approach. In this paper, taking advantage of recent advances in generative Artificial Intelligence, we introduce the Network ConFiguration Generator (LLM-NetCFG) that employs Large Language Model and architects ZSM configuration agents by Large Language Models. LLM-NetCFG can automatically generate configurations, verify them, and configure network devices based on intents expressed in natural language. We also show the automation and verification of network configurations with minimum human intervention. Moreover, we explore the opportunities and challenges of integrating LLM in functional areas of network management to fully achieve ZSM.
Paper Structure (20 sections, 5 figures)

This paper contains 20 sections, 5 figures.

Table of Contents

  1. Introduction
  2. Large Language Models
  3. Architectures
  4. Tokenization and Training
  5. Prompts and Prompt engineering
  6. Fine-tuning
  7. LLMs for Network Management Functionalities: What Has Been Done?
  8. LLM-aided ZSM
  9. LLM for ZSM Components
  10. Choosing LLMs For ZSM Components
  11. LLM-NetCFG: Operation and Deployment
  12. Overall operation
  13. Prompts for Translation and Generation
  14. Use Case: Interface, ACL, Router, and Tunnel Configuration
  15. Prototype and Test Environment
  16. ...and 5 more sections

Figures (5)

  • Figure 1: LLMs Into ZSM Components.
  • Figure 2: LLM-NetCFG Structure.
  • Figure 3: Classification, Translation and Generation Prompts - Example.
  • Figure 4: Accuracy in classifying requirements' type
  • Figure 5: Total Processing Time By Requirement Complexity For Each Class.