17 August 2008

The Olympian Apache

Here's a nice reminder that open source - in the form of Apache - has been head of the field for more than 12 years, despite what certain companies would have us believe:

In 1996 the World Wide Web was truly in its very early stages. The Olympics took place less than a year after Netscape went public, which many consider the key event marking the transition of the Internet from a research network used primarily by the technical community to the commercial behemoth that it went on to become.

The new World Wide Web had the feeling of magic, but, in 1996, it was pretty primitive magic. To begin with, the vast majority of people accessing the Web at the time were doing so over slow dial-up modems with bandwidths of 56 kilobits per second or less. Only at work, if you were lucky, did you have access to faster broadband speeds. It wasn't until years later that broadband usage in the home became commonplace.

As we were planning the IT infrastructure for the Olympics website, hardware was not an issue. We used IBM's SP family of parallel supercomputers which we were confident would provide us with all the computing power we could want.

But the software for web servers was quite immature. Netscape's web software was the most widely used in those days, and while it was adequate for small workloads, its scalability was suspect. We could not use it. Instead, we used the open source Apache Server as the basic web server, and custom built the extensions needed to support its content, applications and other capabilities.

We were pretty sure that the Atlanta Olympics website was the largest such web project anyone had undertaken so far. Because it was all so new, we did not know how many people would come to our website and what features they would use once they got there. We were well aware of the considerable risks inherent in doing such a complex, new project on such a global stage. We knew, for example, that beyond a certain number of users, the response time would start to degrade, and if sufficiently stressed beyond its capabilities, the system could become unstable and crash.

...

Our Olympics website worked quite well, except for some unduly slow response times when traffic got very heavy. Overall, the site handled 187 million hits – that is, individual pieces of information served to users. We learned a lot about the requirements for building and operating large, complex websites. All in all, it was a very successful experiment.

8 comments:

Gen Kanai said...

Glyn, I'm wondering if you could consider a future post or article about the troubling trend since early 2005 where Apache has been losing ground to IIS.

http://news.netcraft.com/archives/2008/05/06/may_2008_web_server_survey.html

Glyn Moody said...

I've already written it a year ago, here:

http://www.linuxjournal.com/node/1000248

Unfortunately, LJ messed up the move to a new platform and have eaten the article.

If you're interested, I'll email you a copy - just let me know at glyn.moody@gmail.com.

Dr. Roy Schestowitz said...

Ouch, Maybe it was DRM there in LJ.

Yum!

Anonymous said...

Too bad IBM doesn't do the Olympic sponsership any more. Now you have to have Microsoft Silverlight installed to watch Olympics on the web.

Glyn Moody said...

@Roy: no, nothing nefarious, just a botched job.

Glyn Moody said...

@Freeman: good point, the Silverlight stuff is worrying.

Gen Kanai said...

More stuff on Silverlight and the Olympics here:

http://www.nytimes.com/2008/08/10/sports/olympics/10stream.html?partner=rssuserland&emc=rss&pagewanted=all

and on Chris Blizzard's blog here (comments instructive too):

http://www.0xdeadbeef.com/weblog/?p=545

Glyn Moody said...

Thanks for those links.