Survey Data Mining:   Home | FAQ | Archive | Glossary
Free Reports

You are viewing an outdated report. The latest version of this report was published on July 1st, 2024

Web Server Survey by IP Address
February 1st, 2002

Web server surveys have been around for a long time. One of the problems with running a survey is how to produce meaningful results given the number of choices with the collection and reporting methodologies. Some examples:

  • How does one treat wildcarded DNS sites that generate unlimited number of unique hostnames, all residing on the same system?
  • How does one treat taste-testing web sites that do not have any unique content on them at all?
  • If very large number of web sites reside on a single IP address, is it a a case of search engine spamming, a large number of low traffic web sites on a single platform, or a web hosting facility using load-redirection for a large number of legitimate web sites?
  • How "greedy" or aggressive should the hunt for new web sites be?
All of the above are issues that our conventional web server survey deals with. The biggest issue that comes up time and again, however, is the fact that our conventional survey counts unique hostnames, not unique installations. This can seriously impact market share statistics. Case in point: the data set used both web server survey reports contains (for the last month's data set) 4,830,270 hostnames, but only 1,619,916 distinct IP addresses.

This particular report addresses this in part by counting unique IP addresses in an attempt at answering the question of how many installations of a particular server might exist. By using the IP address as a proxy for the physical hardware on which a server is running, we still need to be aware that it is not a perfect 1 to 1 mapping:

  • A load balanced network of servers might share a single public facing IP address, and be undercounted.
  • A single server may have multiple public IP addresses bound to one or more network cards on the system, and thus still be counted more than once.

While not a perfect solution, we feel that counting web servers by IP will give a much closer view of the number of server installs for a given technology.

Click any server to see detailed graphs and histories.

Server January 2002
January 2002
December 2001
December 2001
Error - couldn't load prev data

Note: Readers are cautioned that the Growth percentages are based on a server's market share penetration rather than the absolute number of servers we detected. Thus, it is possible for the number of servers to increase, while the growth is actually a negative value.

© 1998-2024 E-Soft Inc. All rights reserved.