Converged RAN/MEC slicing in beyond 5G (B5G) networks
(English) The main objective of this thesis is to propose solutions for implementing dynamic RAN slicing and Functional Split (FS) along with MEC placements in 5G/B5G. In particular, this thesis is divided into three parts. In the first part (Chapter 3), we model a joint slicing and FS optimization...
| Autor: | |
|---|---|
| Formato: | tesis doctoral |
| Estado: | Versión publicada |
| Fecha de publicación: | 2023 |
| País: | España |
| Recursos: | CBUC, CESCA |
| Repositorio: | TDR. Tesis Doctorales en Red |
| OAI Identifier: | oai:www.tdx.cat:10803/690188 |
| Acesso em linha: | http://hdl.handle.net/10803/690188 https://dx.doi.org/10.5821/dissertation-2117-403562 |
| Access Level: | acceso abierto |
| Palavra-chave: | Àrees temàtiques de la UPC::Enginyeria de la telecomunicació 621.3 |
| Resumo: | (English) The main objective of this thesis is to propose solutions for implementing dynamic RAN slicing and Functional Split (FS) along with MEC placements in 5G/B5G. In particular, this thesis is divided into three parts. In the first part (Chapter 3), we model a joint slicing and FS optimization in the 5G RAN with the objectives of optimizing the centralization degree and throughput. In this work, the RAN slicing allowed a customized FS deployment per slice, thus optimizing the available resources, e.g., transport network capacity and Remote Radio Head (RRH) or Central Unit (CU) computational capacity. Next, we present the second part in Chapter 4 by extending the first work by proposing SlicedRAN: service-aware network slicing framework for 5G RAN to create isolated RAN slices based on the service requirements with customized functional splits per slice. The proposed framework investigates the bottlenecks in the capacity of RRHs Fronthaul/Backhaul (FH/BH) network capacity along with a minimum level of Service Level Agreement (SLA) for each slice imposed by the different service types. Finally, in the last part presented in Chapter 5, we investigate dynamic RAN/MEC slicing framework in Open-RAN (O-RAN) architecture to dynamically place the RAN protocol stack of Virtual Network Functions (VNFs) and MEC server per slice. This framework contains the bottlenecks in the capacity of Open-RAN Radio Units (O-RUs), MEC server computation capacity, together with a customized FS per slice, to jointly solve the challenge of operating cost-efficient edge networks and maintaining the served traffic with various QoS criteria. We use a robust Benders decomposition algorithm, which reduces the computation complexity while ensuring an exact and optimal global solution. The proposed algorithm successfully optimizes the joint throughput and system cost in various traffic scenarios while satisfying QoS criteria, as shown by trace-driven simulation results. Hence, in order to determine the right MEC settings for on-demand traffic and alter the MEC type to satisfy the QoS requirements of various User Equipment (UEs) belonging to different slice types, we explore the compute and storage capacity for MEC services such as Enhanced Mobile Broadband (eMBB) and ultra-Reliable and Low-Latency Communications (uRLLC). The overall conclusion of the present findings demonstrates a trade-off between the throughput attained and the cost incurred to the network. As a result, we investigate multi-objective optimization to construct slices while optimizing throughput and decreasing computational cost objectives, and we compare its performance to that of a single objective (maximizing throughput). The findings demonstrate that a throughput increase of up to 160% can be made possible by increasing 78% in the computation cost for a single objective when compared with multi-objective without prioritization. In addition, comparing a single objective with a multi-objective with priority in throughput, it increases throughput by up to 82% and adds 17% to computation costs. Consequently, a single objective of maximizing throughput can result in high throughput at the expense of high cost. It is possible to achieve almost half the amount of throughput using multi-objective with prioritization in throughput, whereas costs can be reduced five-fold. |
|---|