Forum Moderators: phranque
A new survey conducted by Port80 Software of the 1000 most-visited Web sites on the Internet reports that MS Internet Information Services (IIS) serves more high traffic sites than any other Web server, including open source Apache. Based on the best available data from Nielsen/NetRatings, the global standard for Internet audience measurement, Port80 Software analyzed the top 1000 highest traffic sites to determine which Web server software they are using.
Port80's Top 1000 Corporations' Web Servers Survey, analyzing US Fortune 1000 Web sites, has been intensely debated since its inception in January 2003. The company challenges the relevance of the popular Netcraft Web Server Survey for business decision makers, asserting that the Port80 surveys better reflect the large enterprise Web server market. This new Top 1000 High Traffic Web Servers Survey was conducted to illuminate market share among those sites that demand the most out of their Web server software: the most visited destinations on the Internet.
"Netcraft canvases millions of sites, whereas Port80 focuses on selective samples of high traffic and major corporate Web sites," said Joseph Lima, COO of Port80 Software. "In response to our Fortune 1000 survey, many critics suggested that sampling the highest traffic sites would make for a more enlightening survey. Each survey provides a different picture of the Web server market and, while there is no definitive answer, we hope that this high traffic survey adds a significant new perspective to the debate."
Port80 Software's January 2004 survey of the Web server software used by the 1000 highest traffic Web sites revealed the following market shares:
Microsoft IIS: 43.1%
Apache: 39.7%
Netscape Enterprise: 8.7%
Other Web servers: 8.5%
For details on Port80 Software's new monthly Top 1000 High Traffic Web Servers Survey, methodology, and a tool that checks a site's Web server type in real-time, visit:
http*//www.w2knews.com/rd/rd.cfm?id=040209RN-Port80_Software
I could see a good situation where a Corp has an internal MS Network and a Linux based DMZ for conection to the internet.
h**p://www.port80software.com/support/p80tools
Put www.apache.org in the box, and they correctly identify the server as using Apache on Unix - followed by
Note:
No matter what the above results show, this company may be running Microsoft IIS and protecting its Web server identity with (ProductName).Try (ProductName) FREE for 30 days. Download Now!
Buy (ProductName) for only $49.95 today!
Put microsoft.com, and they follow up with a similar ad for their product (which changes the header to hide the server name), with "Protect your IIS server from hackers!" Even if you don't believe they're biased, their agenda is clear - the survey is just a marketing ploy.
They are trying to make a name for themselves by taking on Netcraft and attacking the latter's methodology. Netcraft tries to poll everything (including millions of parked domains), whereas Port80 looks at a very select group. Of course, both approaches are valid, but neither is ideal. Specifically, Port80's first survey of Forture 1000 companies (which also showed IIS in the lead) was evidently US-centric, and included a large number of holding companies' little-visited sites. Their latest survey is more interesting, but only relevant to very large content providers making software selections, not the rest of us. I don't think that the likes of Amazon would pick a software platform/web server combination on the back of this survey anyway.
An interesting point is that they sorted their list along "unique audience", which probably means what other surveys call "reach". This doesn't necessarily say much about performance requirements, but pushes up the microsoft properties msn.com, microsoft.com, passport.com, passport.net, windowsmedia.com, msnbc.com, expedia.com, and possibly others I have missed.
But the main problem is really what adamas points to: With high traffic sites, the software you see from the outside is really only a reverse proxy. That's a gateway that will forward all requests to other internal systems, and tells us nothing at all about the platforms that do the actual work. Add port80's own camouflage software to the picture, and the figures become completely meaningless.
I believe bird is saying that the results are meaningless because you can have one domain handled by multiple servers.
I was getting at the fact that you can have multiple domains handled by one server. In that case it may be that none of the individual domains would be a '1000 highest traffic site' but the server(s) may be handling more requests than the server(s) of a '1000 highest traffic site'.