Not surprisingly, Apple’s websites were hammered yesterday on news of the Mac Mini, iPod Shuffle, and iWork:
The Apple web site, which runs on Mac OS X, experienced some slowdowns but was largely available. Apple’s online store (also on Mac OS X) struggled, however, experiencing outages and lengthy response times. Faring even worse was the official site for MacWorld Expo, which runs on Windows Server 2003, and was offline for hours following the show’s keynote address by Apple CEO Steve Jobs.
The article at Netcraft includes graphs to illustrate the response times (or lack thereof). I don’t think anyone needs me to point out the irony in running a Mac site on Windows… or the ability of the “small” Mac audience to take down a Windows-based server.
This is not entirely fair – as Apple uses (and owns part of) Akamai to do distributed serving/cacheing of their traffic, and there would probably a a number of inverse proxy caches sitting in front of their webservers. These tend to be transparent to the end user, but will offload at least 95% of the traffic from the webservers themselves.
So you’re really monitoring the response time of the cache’s for any infrastructure of any size. If I was running Apple’s sites I would reduce the freshness factor of any material in the cache, and let them take all traffic at such times (i.e. preload the cahce’s)
The irony of the windows server is till valid though..
Good point. I don’t know why I didn’t think of it, or why Netcraft didn’t mention it—they’re certainly in a position to know! Sure, they weren’t focusing on the server’s software so much as the sites’ availability, but they’ve talked about sites using Akamai before.