Ajayi, Jesutofunmi Ademiposi (2025). Multi-Dimensional Network Slicing: Algorithmic and Learning-based Methods. (Thesis). Universität Bern, Bern
|
Text
25ajayi_ja.pdf - Thesis Available under License Creative Commons: Attribution-Noncommercial-No Derivative Works (CC-BY-NC-ND 4.0). Download (2MB) | Preview |
Abstract
Next‐generation mobile networks are expected to support a wide range of demanding applications and services that have strict and varying performance requirements. To meet these requirements, Network Slicing (NS) has emerged as a powerful technique to enable cost‐effective, multi‐tenant communications and services over a shared physical mobile network infrastructure. However, the effective realization of the NS paradigm hinges on the ability to manage the end‐to‐end lifecycle of network slices in a dynamic, efficient, and automated manner. This challenge is exacerbated by the multi‐dimensional slice requirements such as bandwidth, latency, and CPU and the unpredictable, online arrival of slice requests. To address the challenges of managing the end‐to‐end lifecycle of network slices, this thesis proposes online, data‐driven solutions that are capable of adapting to dynamic conditions in mobile networks by optimizing multi‐dimensional resource allocations, enabling scalable, real‐time slice orchestration across heterogeneous infrastructures and proactively optimizing network slice performance to ensure service‐level compliance. First, we investigate the slice admission control problem, focusing on the setting where slice requests arrive sequentially and must be admitted or rejected in real‐time without prior knowledge of future resource demands. We propose an online algorithm that dynamically incorporates system resource utilization to guide admission decisions, and therefore resource allocations, with the aim of maximizing the long‐term revenue of infrastructure providers. Second, building on this foundation, we address the problem of online policy selection under non‐stationary network conditions. By modeling the policy selection task as a multi‐armed bandit problem, we propose a data‐driven solution that learns to select the most effective admission policy across time‐varying network conditions. This approach balances exploration and exploitation while detecting and reacting to changes in environmental dynamics. Third, we address the problem of scalable slice provisioning in large‐scale, distributed networks, and propose a hierarchical solution for the online network slice provisioning problem in which a service function chain must be effectively mapped onto the network infrastructure, while optimizing for multiple objectives. Finally, we propose a proactive optimization framework for the problem of allocating resources to heterogeneous network slices. The proposed framework aims to learn an effective resource allocation strategy in virtual radio access network environments by anticipating traffic fluctuations and proactively adjusting network slice resources for optimal performance.
| Item Type: | Thesis |
|---|---|
| Granting Institution: | Faculty of Science, University of Bern |
| Dissertation Type: | Single |
| Date of Defense: | 26 September 2025 |
| Subjects: | 000 Computer science, knowledge & systems 500 Science > 510 Mathematics |
| Institute / Center: | 08 Faculty of Science > Institute of Computer Science (INF) |
| Depositing User: | Jesutofunmi Ademiposi Ajayi |
| Date Deposited: | 16 Oct 2025 11:53 |
| Last Modified: | 16 Oct 2025 11:53 |
| URI: | https://boristheses.unibe.ch/id/eprint/6757 |
Actions (login required)
 |
View Item |