Modeling and analysis of web usage and experience based on link-level measurements

Internet traffic monitoring and analysis have been playing a crucial role in understanding and characterizing user behavior on the web. In particular, ON-OFF models capture the essential phases of user communication with web servers. The OFF phases reflect both deliberate and accidental gaps in the traffic flow. In this paper, we present a passive monitoring and analysis method devised to assist in the identification of such traffic gaps that may result in the degradation of Quality of Experience (QoE). Our first contribution consists in a revised ON-OFF model to cater for OFF times reflecting accidental gaps which are induced by the network. Second, a wavelet-based criterion is proposed to differentiate between the network-induced traffic gaps and user think times. The proposed method is intended to be implemented in near-real-time as it does not require any deep packet inspection. Both web service providers and network operators may use this method to obtain objective evidence of the appearance of QoE problems from link-level measurements.