Traffic properties, client side cachability and CDN usage of popular web sites

Abstract

Web traffic measurement and modeling have contributed to understanding the effect of Web traffic on Internet resources since the 1990s. In the past years, a number of newWeb features have gained more and more importance, e.g. content delivery networks (CDNs), increased amount of advertisement, personalization, usage tracking, client scripting and Web 2.0 style "mashups". This paper uses active Web measurements to assess the efficiency of client side caching for modern Web sites, investigating some Web features in detail. As expected, we see that more than 50% of the average downstream traffic volume is saved when loading a page using client side caching. More unexpected results comprise the actual distribution of cache effectiveness, varying between extreme and no reduction of traffic, the cachability of "Web bugs" and the variance between sites in cachable image pixels and CDN based files. © Springer-Verlag Berlin Heidelberg 2010.