QoE-Oriented Routing Mixing Application KPIs and Link Metrics Through Machine Learning

Abstract

As the complexity of network end devices and applications grows, network managers face increasing difficulty in meeting specific end user requirements, leading to reduced user experience and inefficient resource management. This paper introduces a Quality of Experience (QoE)-oriented routing strategy to enhance user experience by selecting routing paths based on application-specific QoE. Application key performance indicators (KPIs) and dynamic link metrics are utilized to represent real-time QoE and network state. This data builds QoE models for various applications such as video streaming, VoIP, and web map, using four learning methods. The trained models are implemented in a software-defined networking (SDN) controller for optimal QoE routing. Evaluations using the Mininet network simulator reveal that the proposed QoE routing strategy can select the best path 78.4% of the time which is almost 20% more than the top-performing state-of-the-art. This results in measurably higher application performance, proving the efficiency of the proposed approach in improving the application’s QoE.

Keywords

• Link metrics
• machine learning
• quality of experience (QoE)
• routing
• software-defined networking (SDN)