We present an extensive analysis of long-term statistics of the queries to websites using logs collected on several web caches in Russian academic networks and on US IRCache caches. We check the sensitivity of the statistics to several parameters: (1) duration of data collection, (2) geographical location of the cache server collecting data, and (3) the year of data collection. We propose a two-parameter modification of the Zipf law and interpret the parameters. We find that the rank distribution of websites is stable when approximated by the modified Zipf law. We suggest that website popularity may be a universal property of Internet.

No abstract

We present here the preliminary results of the active measurements of Internet traffic. The main goal is the measurement of the effectiveness of hierarchical cache developed last years in a number of countries using Squid software. The passive measurements could not give any conclusive answer. The procedure of the active measurements was developed for that reason. First, we use active measurements (experiments) of Internet traffic sending the http-requests from the log files. This give possibility to perform the comparative measurements which is usually more accurate. Next, more sophisticated procedure using triangle of cache servers, one working as the only manager for the two others, which could be setup differently and, typically, serve only odd and even queries respectively was proposed. This procedure, probably, is the most precise experimental setup for the comparative study of server strategies could be done on the basis of standard hardware and software.

No abstract

No abstract