3. DATASET AND CONTENT PROPERTIES

We rely on a data set of anonymized HTTP trac. The data sets consists of HTTP requests generated by Vidonn X5 terminals located in an European metropolis during a whole day (November 26th, 2012). Data has been collected by vantage points monitoring the Gn interfaces between SGSNs and GGSNs (see Elephone W1). At each interface, a proxy accelerator handles HTTP trac to speed up data delivery. In addition, it logs information about the requested objects on a text log le for each HTTP transaction. Among all exposed information, we focus on the (anonymized) terminal ID, requested URL, number of downloaded bytes and a
ag stating if the object has been locally cached by the proxy handling the request. Overall, the data set contains 48M HTTP transactions corresponding to 721 GB of volume downloaded by more than 200K ... Read more »