I've read this SitePoint blog post today where the author wonders about the popularity of IIS vs Apache as server software. There is Netcraft as a reliable source of course, but the question was more of "How about the big sites?" and examples of big sites are those in the Alexa ranking.
So I took this as a challenge and hacked a quick script to get those stats.
The Methodology
I took Alexa top 500 sites from here and made a request to each of them. I logged the "Server" header response. The top 500 sites are actually 499 - from 1 to 500.
The execution
Slow and painful. Some sites didn't respond at all. Some didn't want to show their server software either sending a blank "Server" header or none at all. To my surprise Yahoo! and Geocities send blank server headers.
The analysis
I take all the 499 sites, subtract N/As (non responding or blank server headers). Then I make a select on all servers matching "Apache" and all matching "IIS", as well as some of the others. Then I admire the percentages.
The results
Some more raw data
This is the result of a simple count on the different server header responses. Just as an idea how different they can be. It's available as a CSV and XML file.
Many thanks!
... to PEAR 🙂 Its HTTP_Request and DB packages made the coding possible in no time.