Towards intent-based management for Open Radio Access Networks: an agile framework for detecting service-level agreement conflicts

Radio Access Networks (RAN) management and orchestration are challenging due to the network’s complexity and dynamics. Management and orchestration rely on enforcing complex policies derived from mapping high-level intents, expressed as Service-Level Agreements (SLAs), into low-level actions to be d...

Full description

Saved in:
Bibliographic Details
Published in:Annales des télécommunications 2024, Vol.79 (9-10), p.693-706
Main Authors: de Oliveira, Nicollas R., Medeiros, Dianne S. V., Moraes, Igor M., Andreonni, Martin, Mattos, Diogo M. F.
Format: Article
Language:English
Subjects:
Artificial neural networks
Circuits
Communications Engineering
Complexity
Computer Communication Networks
Engineering
Information and Communication
Information Systems and Communication Service
Mapping
Modular systems
Natural language processing
Networks
Neural networks
Policies
R & D/Technology Policy
Signal,Image and Speech Processing
Citations: Items that this one cites
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Radio Access Networks (RAN) management and orchestration are challenging due to the network’s complexity and dynamics. Management and orchestration rely on enforcing complex policies derived from mapping high-level intents, expressed as Service-Level Agreements (SLAs), into low-level actions to be deployed on the network. Such mapping is human-made and frequently leads to errors. This paper proposes the AGility in Intent-based management of service-level agreement Refinements (AGIR) system for implementing automated intent-based management in Open Radio Access Networks (Open RAN). The proposed system is modular and relies on Natural Language Processing (NLP) to allow operators to specify Service-Level Objectives (SLOs) for the RAN to fulfill without explicitly defining how to achieve these SLOs. It is possible because the AGIR system translates imprecise intents into configurable network instructions, detecting conflicts among the received intents. To develop the conflict detection module, we propose to use two deep neural network models, Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU). The deep neural network model determines whether intents and policies are conflicting. Our results reveal that the proposed system reaches more than 80% recall in detecting conflicting intents when deploying an LSTM model with 256 neurons.
ISSN:0003-4347
1958-9395
DOI:10.1007/s12243-024-01035-3