Characterizing webpage load from the perspective of TCP connections

Abstract—Over the last years websites have evolved rapidly incorporating new content types and becoming more and more dynamic. Users today are able to access a wide variety of content and services through their web browsers. As a consequence, web traffic has become increasingly complex and, from a network perspective it can be difficult to ascertain which websites are being visited by a user, let alone which part of the user’s traffic each of them is responsible for. Although there is an extensive literature on the new charac-teristics of web traffic, few works have focused on a connection level perspective even if this kind of data is easily available for network administrators. In this paper we offer a characteriza-tion of webpage download using connection level metrics. This description is a first step in developing techniques able to identify individual webpage downloads in real traffic. We have captured an extensive dataset of more than 20,000 webpage downloads that we study in order to provide different connection level based metrics. We study how these metrics vary between different webpages of different popularity and complexity. In the end, we attempt to provide a general modelling of a normal webpage download. I