mso-ansi-language:EN-US">The rapid evolution of time-sensitive applications like holographic communication, the tactile Internet, autonomous systems, and the metaverse demands ultra-low latency, high reliability, and predictable network performance. These requirements call for deterministic network behaviors, yet traditional architectures struggle to meet such demands in dynamic environments with varying traffic patterns.
mso-ansi-language:EN-US">This thesis focuses on the transition from 5G to 6G, emphasizing deterministic networking enabled by Software-Defined Networking (SDN). While SDN offers programmability and centralized control, its capabilities for optimizing the data plane (DP) to support time-sensitive applications remain underexplored. The work proposes SDN-based frameworks and solutions to ensure deterministic end-to-end (E2E) latency and quality of service (QoS) in next-generation networks.
mso-ansi-language:EN-US">Key contributions include:
mso-ansi-language:EN-US">SDN-Based Path Selection Using Dynamic Queuing: The SDLL framework leverages dynamic queuing mechanisms and traffic engineering (TE) to prioritize delay-sensitive flows, integrate 5G QoS Flow Identifier (QFI), and ensure scalable and flexible traffic management across transport networks (TN).
mso-ansi-language:EN-US">GNN-Assisted Admission Control (GNN-AC): This mechanism predicts network latency using a two-layer architecture with RouteNet-F for real-time latency prediction and an admission control layer to regulate traffic dynamically, ensuring QoS while avoiding congestion.
mso-ansi-language:EN-US">L4S-Based Congestion Control: Integrating Low Latency Low Loss Scalable Throughput (L4S) protocols, this framework uses Explicit Congestion Notification (ECN), dual-queue isolation, and adaptive transmission rates to maintain ultra-low latency and high throughput for critical applications.
mso-ansi-language:EN-US">eBPF-Enhanced Edge Processing in CEC Interconnection: The HELIOS framework utilizes eBPF for efficient packet processing at the kernel level, managed by an SDN controller. This reduces processing overhead and ensures efficient cloud, edge, and far-edge interconnections.
mso-ansi-language:EN-US">Extensive evaluations in virtual and physical setups, focusing on 5G TN backhaul and midhaul, demonstrate significant reductions in latency, improved resource utilization, and scalability under dynamic conditions. Validated through IEEE publications, this research bridges application demands and network capabilities, providing a foundation for SDN-enabled deterministic communication in 6G networks. It addresses dynamic traffic management, admission control, congestion control, and edge processing, enabling next-generation services with stringent performance requirements.
