Optimised 5G Network Sub Slicing Orchestration (ONSSO) in a fully virtualised smart company using machine learning

Abstract

The traditional method of one size fits all for network resource allocation no longer works for various vertical with the limited available network resource to meet the increasingly rise in demand for diverse requirements to satisfy network quality of service (QoS), quality of experience (QoE) and service level agreements (SLA). The complexity of allocating various network resources to each vertical as required and on-demand has been a major challenge with the flexible 5G network. This research examined existing studies on network slicing, which serves as an enabler for 5G and aim to identify and solve the challenge with efficiently allocating the limited network resource to orchestrate the 5G network slices on-demand. It also investigated the existing challenges of managing current cross domain network resources in an enterprise network with dynamic and unstructured network slice requirements and then proposes a framework to efficiently orchestrate the network slice to meet the varying latency, data rate, mobility, and reliability needs thereby reducing capital and operational expenditure. In this study, the proposed advanced machine learning (ML) algorithm pipeline model designed for optimised sub-slicing of 5G network slice dynamically on demand, re-learning, and service function chaining to orchestrate bespoke network slice is called the Enhanced Subslice Model (eSS). This study further proposes a smart and novel end-to-end adaptive resource management framework which is defined as ONSSO. This proposed framework comprises of functions ranging from the analysis of the diversified network service on a network, to the implementation of the proposed ML-based eSS model which incorporates a series of supervised machine learning algorithms as well as the LazyPredict module to find and suggest the best-fit model, to the making of intelligent decisions for the specific network data environment being monitored in real time or analysed with historical data. It also includes functions ranging from implementing caching to improve the user experience and performance of orchestration right through to employing reinforcement learning techniques for admission control, as well as predicting future network traffic data pattern thereby ultimately adapting slice configurations in real-time and efficiently orchestrating alternative best fit bespoke network slices. This ONSSO is designed to enable and optimise processes within a newly proposed vertical called Enterprise Company Network as a Service (CNaaS) which is based on an industry 4.0 smart company. The integration of the ONSSO into CNaaS provides a scalable and optimised solution to network sub-slicing orchestration in an enterprise network. This CNaaS was ultimately used as a case study in validating the model through simulations conducted using python with MATLAB and Simulink, where different network parameters were analysed under varying conditions. These simulations enabled the identification of optimised resource allocation strategies across multiple slices, considering factors such as latency, bandwidth, and energy efficiency.Results from the MATLAB simulation demonstrate that the proposed model significantly enhances the flexibility and scalability of network subslice orchestration for dynamic and adaptive network management in future mobile networks. When compared with traditional static orchestration methods, the sub-slicing of the network slice with the proposed machine learning-based pipeline exhibited improvements in resource utilisation, enhancing QoS and QoE for end-users and reduced the risk of denial of service. This research offers a novel approach to network slice orchestration, with potential applications in various sectors such as smart cities and autonomous vehicles. The proposed model not only contributes to the ongoing development of 5G and 6G networks but also provides a foundation for future research on adaptive network management. This research aims to initiate discussion on the new suggested concept of enhanced subslice as well as robotic process automation of network slice orchestration, provide the foundation for the application model which application programming interface can be built upon and spur development on true fully automated self-organising network slice orchestration.

Date of Award	May 2025
Original language	English
Awarding Institution	University of Bedfordshire
Supervisor	Enjie Liu (Supervisor) & Renxi Qiu (Second supervisor)