QoS-Aware Self-Adaptive Random Early Detection for Congestion Control in Networks With Multiple Classes of Traffic
TL;DR
QASARED unifies class-based QoS differentiation and load-aware adaptability within a single self-regulating framework, enabling it to respond smoothly to dynamic congestion conditions without relying on external control feedback or manual parameter tuning.
Abstract
An efficient congestion control mechanism remains a fundamental requirement for achieving Quality of Service (QoS) assurance in modern networks carrying heterogeneous traffic classes. Classical Active Queue Management (AQM) schemes such as Random Early Detection (RED) and its widely deployed weighted variant (WRED) rely on static configuration parameters, which limit their adaptability and responsiveness under dynamic traffic loads. This paper presents QoS-Aware Self-Adaptive Random Early Detection (QASARED), a load-sensitive AQM mechanism that dynamically adjusts its packet dropping behavior based on both network load intensity and traffic class priority. Unlike WRED, QASARED unifies class-based QoS differentiation and load-aware adaptability within a single self-regulating framework, enabling it to respond smoothly to dynamic congestion conditions without relying on external control feedback or manual parameter tuning. Simulation experiments were conducted to evaluate the performance of QASARED against WRED across various traffic load conditions. The results show that QASARED achieves improved buffer utilization, enhanced stability, fairness in handling different traffic classes, and real-time QoS differentiation in different network load conditions without sacrificing simplicity or compatibility with existing RED/WRED architectures. These characteristics demonstrate the potential of QASARED as an efficient congestion control solution for modern multi-class communication networks and emerging edge environments.