Optimising Quality of Service (QoS) in Edge and Fog Computing: A Comparative Analysis of DiffServ and IntServ Mechanisms in Heterogeneous Network Environments.

Abstract

This thesis examines how Quality of Service (QoS) can be preserved in distributed edge–fog–cloud systems when heterogeneous traffic competes for limited uplink capacity and latency-sensitive applications require low delay, low jitter, and reliable packet delivery. It compares the suitability of DiffServ and IntServ in hierarchical edge–fog environments and investigates whether a hybrid approach can provide stronger protection for critical traffic under congestion. The study combines controlled ns-3 simulation with physical validation on a Raspberry Pi 4 testbed using Open vSwitch and a Ryu SDN controller, evaluating delay, jitter, loss, throughput, and SLA compliance across multiple traffic patterns and bottleneck conditions. The principal contribution is an adaptive hybrid QoS architecture that integrates ML-based bandwidth prediction, water-filling allocation, PI feedback control, queue-depth feedforward, and confidence-gated rate adaptation within a delay-driven per-class bandwidth management framework. Results show that the Hybrid architecture delivered the strongest and most consistent EF protection in simulation, reducing mean EF delay from 226.91 ms to 9.04 ms and achieving 100% EF SLA compliance, while the physical testbed directionally confirmed its advantage over Baseline and DiffServ under congested conditions. Overall, the thesis demonstrates that adaptive hybrid QoS is the most effective of the evaluated approaches for multi-service edge–fog deployments where robust protection of delay-sensitive traffic is the primary objective.