Application-layer QOS metrics to aid network performance monitoring and diagnostics for encrypted traffic

Most of the network statistics do not capture the impact of network conditions on user's applications and their quality of experience (QoE). Active measurements can help but their results may not accurately reflect the QoE at end users. In this work, a low complexity non-intrusive method is proposed to estimate application-layer Quality of Service (QoS) metrics experienced by end-users. The method passively and continuously monitors user's traffic from a vantage point within the ISP network. These metrics accurately capture the network conditions experienced by end user's devices or internet applications. The approach is designed for encrypted traffic but can also be applied to unencrypted traffic. The metrics, together with existing set of network statistics, provide more accurate network performance monitoring. We also demonstrate that the metrics can be used for detecting degraded network performance and finding the root cause. Our top-down approach from user perspective complements the usual bottom-up technique using network component statistics and alerts.