Pu Yang

Topic

Traffic-Aware Routing Optimization with Stable Forwarding Adaptation

Department of Electrical and Computer Engineering

Date & location

• Monday, August 17, 2026
• 9:30 A.M.
• Engineering & Computer Science Building, Room 467

Examining Committee

Supervisory Committee

• Dr. Lin Cai, Department of Electrical and Computer Engineering, University of Victoria (Supervisor)
• Dr. Amirali Baniasadi, Department of Electrical and Computer Engineering, UVic (Member)
• Dr. Kui Wu, Department of Computer Science, UVic (Outside Member)

External Examiner

• Dr. Shiwen Mao, Department of Electrical and Computer Engineering, Arburn University

Chair of Oral Examination

• Dr. Jody Klymak, School of Earth and Ocean Sciences, UVic

Abstract

Modern networks must increasingly optimize service quality in addition to providing reachability. Applications such as real-time communication, cloud services, and AI data-center workloads require routing systems to exploit traffic information when making forwarding decisions in order to meet latency, throughput, and reliability objectives. In these environments, a path may remain reachable while failing to deliver the desired service quality.

However, traffic information differs fundamentally from routing information. Routing state is derived from topology, reachability, and policy constraints, and is valued for its stability and consistency. Traffic information, by contrast, is local, transient, and rapidly changing. While such information can improve forwarding decisions, treating transient traffic measurements as routing state may introduce excessive control-plane dynamics and routing instability. This creates a fundamental challenge: how can networks exploit traffic information without sacrificing routing stability?

This dissertation develops a stable forwarding adaptation approach in which stable routing state constrains forwarding choices while dynamic traffic information guides forwarding behavior rather than redefining routing state.

Guided by this principle, the dissertation develops four traffic-aware routing systems. Delay-Guaranteed Routing (DGR) demonstrates how controlled path diversity guided by local traffic information can improve delay-sensitive traffic delivery. Deadline-Driven Routing (DDR) addresses the staleness of distributed traffic observations through queue-state estimation, enabling more accurate deadline-aware forwarding decisions. Information-Rich Routing (IR) generalizes these insights into a unified framework that separates routing state from traffic-driven forwarding adaptation. DF-LRA extends the same principle to lossless AI data-center fabrics, where controlled detour admission mitigates congestion propagation and flow-control risks.

Together, these systems demonstrate that traffic awareness and routing stability need not be conflicting objectives. By separating stable routing state from dynamic traffic information, networks can become increasingly service-aware without sacrificing the stability, auditability, and deployability expected of production routing systems. More broadly, it provides a unified perspective on traffic-aware routing across both large-scale networks and emerging AI infrastructure.