Understanding the quality of Experience (QoE) of web brows- ing is key to optimize services and keep users’ loyalty. This is crucial for both Content Providers and Internet Service Providers (ISPs). Quality is subjective, and the complexity of today’s pages challenges its measurement. OnLoad time and SpeedIndex are notable attempts to quantify web performance with objective metrics. However, these metrics can only be computed by instrumenting the browser and, thus, are not available to ISPs. We designed PAIN: PAssive INdicator for ISPs. It is an automatic system to monitor the performance of web pages from passive measurements. It is open source and available for download. It leverages only flow-level and DNS measurements which are still possible in the network despite the deployment of HTTPS. With unsupervised learn- ing, PAIN automatically creates a machine learning model from the timeline of requests issued by browsers to render web pages, and uses it to measure web performance in real- time. We compared PAIN to indicators based on in-browser instrumentation and found strong correlations between the approaches. PAIN correctly highlights worsening network conditions and provides visibility into web performance. We let PAIN run on a real ISP network, and found that it is able to pinpoint performance variations across time and groups of users.
