The Giants Didn’t Win Super Bowl XLII – The Website Wars

The New York Giants did not win Super Bowl XLII in my book. I’m not much of a sports fan myself, but I certainly can enjoy having a few beers with friends and family to sit down and watch the Super Bowl. I’ve never lived in a city that has gone into the post-season so that might explain things.

As such, I was rooting for the commercials, which until the 4th quarter were winning, although not by much. What I want to know after watching all these commercials is who won – Microsoft and IIS or *nix and Apache.

What would an inquisitive person do? They would probably visit each company website and keep score using a Firefox extension called Live HTTP Headers which makes it “easy” to determine what operating system a web server is using. They also might utilize Netcraft to lookup their server information along with historical data and finally probably check out NMAP for a though-and-through analysis.

That’s exactly what I did. I watched 65 commercials all over again, lasting more than 30 minutes of advertisements. Am I a consumer whore? You bet! *wink*

There are few technicalities that need to be addressed before getting an accurate score. For example, many companies are using web-acceleration platforms which may skew results. One such company that provides a caching platform is Akamai who claim they serve various media (video, advertising, etc) for over 50% of the companies in the Super Bowl. So at least 50% of company websites use Linux, right? Not quite so fast.

If a company uses Microsoft IIS and Akamai caching which is a fairly common technique, using both Microsoft and Linux – what should get counted? One? Both? Count .5 for each? Furthermore, how should the number of companies be counted? Even Microsoft’s own website serves up content using the Akamai caching network that consists of tens of thousands of Linux servers located around the globe running Linux.

Should we even consider counting all the Fox/News Corporation ads (who runs a Microsoft-only ship), or just count them once? Undoubtedly if a different TV network such as NBC or CBS were televising the Super Bowl, they would undoubtedly tip the scales the other way with ads for their own TV shows and movies towards Linux or FreeBSD, respectively.

Similarly, Budweiser ran 6 commercials – should they get counted six times? Probably not. I must raise the question of if “Joe Sixpack” will ever visit budweiser.com to figure out what kind of beer to buy. Mr. Sixpack will hit up the nearest convenience store and never even use the computer to find out about micro brews available in his locale.

Would Joe Trader visit etrade.com? If they’re in the market (pun intended) they might. Should they get counted twice, once for each appearance of the puking baby?

I think the best way to count is by looking up every company or product website that was advertised in the Super Bowl, regardless of if the domain name was mentioned in the advertisement, then figure out what’s running the server. That decision thins the crowd a bit. Instead of 65 websites with duplicates included, we’re down to 41 domains. I’m not even going to count NFL.com which is completely powered by Akamai/Linux.

There were some interesting bits found in my research. etrade.com runs Ubuntu. They’re actually the only server that replied with an OS by name in the HTTP Header. Everyone else had generic responses like “Apache”, “Apache (Unix)”, “Microsoft-IIS/6.0″, or “Sun-ONE-Web-Server”. Victoria‘s Secret replies with a simple “Webserver” which is a front for Akamai. Kudos to E*Trade for pimping Ubuntu.

The last check I ran was using nmap with an input file containing the list of domains to scan. It took about 20 minutes since a few servers decided to slow down the scan, undoubtedly after an IDS identified the scan attempt.

One such domain that had EVERY PORT OPEN and delayed a response on each port was www.whitehousedrugpolicy.gov. They were in the list for the commercial about the out-of-work drug dealer because kids are scoring from their parents’ bathroom cabinets. Oops, I wasn’t expecting to port scan a government website. Nmap did not give any additional insight into the operating system running on any server. Netcraft is pretty good. :)

In the end, the best option is to score on what is reported by the HTTP headers. In this case we have two sites that report Akamai and are indeterminable – Life Cant Wait (Sunsilk) and Victoria’s Secret. Otherwise, the major players are Apache, IIS, Sun, and finally IBM – which is actually based on Apache.

So who actually wins?

http-headers.PNG
(click to enlarge)

At the end of regulation time, it’s a tie! Both Apache and Microsoft are powering 17 servers! If we were to count IBM which is Apache based, then Apache wins. Add in Akamai which uses Linux, and I think it’s a safe bet to say that Apache and *nix win by a hair.

These results are impressive. Who thinks we can round up enough people and/or vendor support to advertise a Linux product in the Super Bowl next year?

  • Anthony A.

    Great post! (as usual)

    What software did you use for the graph? And what’s that font? Together they look pretty impressive.

  • http://www.fsckin.com/ Wayne

    The font is Calibri.

    No shame here, but I’ve been running Excel in a virtual machine – don’t worry, I’ve got a free license – the Beastmaster hasn’t made a buck off me in years, discounting licenses purchased at my day job.

    I haven’t found a good replacement for a binary newsgroup reader for Linux and I’ve been using the VM for that also.

    I wrote an article about using Office/Outlook in a VM awhile back, and I’ve been dogfooding that since then.

  • PredatoryFern

    Wayne, I have a quick nit pick. 5th paragraph:
    “For example, many companies are use web-acceleration platforms which may skew results.” Shouldn’t it be using?

    I think VM’s are going to get a lot use bridging the gap between Windows and WINE/*nix. I know people like to choose 1 or the other — pick a side as it were. Personally, I’m enjoying running a bunch of them.

  • http://www.fsckin.com/ Wayne

    PredatoryFern: Fixed! That’s the problem with writing and editing being done by one person. :)

    I tend to use what is familiar, and slam my head repeatedly against the learning curve of freedom software when time presents itself.

  • Anthony A.

    Haha- What would rms say about that, Wayne? Just kidding- I use VirtualBox + XP + Office 2007 regularly, especially when I need professional looking documents (such as official resumes and such). It works so perfectly for me, I’m probably going to remove my dual-boot setup during my next format.

    As for my plotting stuff, I use octave/gnuplot/latex all the time (I’m an engineer, so my plots aren’t usually bar graphs). Openoffice plotting is intolerable at best, so bit the bullet and slammed my head repeatedly against latex for a while.

  • Vadim P.

    I’m learning R for plotting :/

    Great post though!

  • http://www.tophatstuff.co.uk/ Ben

    I’m surprised that the results were so close actually – I expected Apache servers to win by a far larger ratio.

    I suppose Apache appears more dominant when I’m browsing with the ServerSpy Firefox extension because it’s cheaper when it comes to shared hosting, and is therefore more often used – there could be more *physical* microsoft servers, but vastly more individual Linux powered websites, even if they’re crammed 1000-to-a-box.