Operator-Centric Strategies for Enhancing Performance and Reliability of Cloud-Native 5G Core and Wi-Fi Access Networks

<p dir="ltr">The fundamental shift towards complex, disaggregated network architectures necessitates new operator-centric strategies to ensure high performance and reliability. This thesis introduces a suite of such methods tailored for implementation within operator-managed infrastructure, specifically cloud-native 5G core and centralized Wi-Fi access networks. </p><p dir="ltr">Within the cloud-native 5G core, the thesis introduces strategies to enhance both performance and reliability. First, it adopts a stateless network function paradigm to improve reliability and optimize the state cost through shared state access, non-blocking interactions and data embedding techniques to reduce call setup time. Reliability is further enhanced by a system named ZTX-SEM, a zero-trust security architecture that maintains high performance and low resource utilization, through mechanisms such as protocol-agnostic packet interception, proactive authentication and optimized key lookups. The thesis further improves the operational reliability of cloud-native 5G core with a redesigned transformer architecture for anomaly detection through a system we name Janus. Its single-pass dual-mask attention mechanism and curriculum learning framework enable Janus to identify complex anomalies by capturing both the procedural flow and field-level semantics of 5G call-flows.</p><p dir="ltr">In the domain of Wi-Fi access networks, the thesis focuses on strategies for enhancing network performance, specifically user Quality of Experience (QoE). Leveraging insights from a measurement study on how centralized Wi-Fi configurations affect user QoE, the thesis introduces Maestro, a dynamic resource allocation framework designed for implementation at the central wireless local area network controller. Maestro combines user experience prediction with reinforcement learning, and adaptively allocates access categories and priority queues for application flows to enhance overall QoE and fairness.</p><p dir="ltr">The thesis validates the proposed operator-centric strategies through comprehensive experimental evaluation on 5G and Wi-Fi testbeds. The results demonstrate significant improvements in latency, security, resource efficiency, and overall end-user experience.</p>