I hate to gripe, but I'm kinda obligated to jump on this one. Google did *not* invent this; it's been thought of independently multiple times, including by myself as well as by Brewster Kahle of the Internet Archive. See Cringely's 24 November article at http://www.pbs.org/cringely/ pulp...it20051124.html, for instance. Please be careful using the word "invent" when writing for public consumption, particularly when crediting large corporations, because it has legal connotations, and folks with billions in market cap sure don't need the public's help defending patents on ideas that are not original.

The idea itself just makes sense; shops spend a great deal on building out their data centers in class A space, raised floors and everything, then they inevitably need to move. In my own case, my family business has spent years building racks, power, and cooling in a leased industrial park space. This locks us into our lease in a way that is very bad. A while back I decided that, if we move or scale up, we need to build any new data centers in refrigerated shipping containers, as sealed units with internal redundancy, because then we'd be able to just drop them off and plug them in. If enough parts die inside of one, then we just ship it back home for overhaul. Because you need to be able fail any container over to another, you need to mirror the data between containers; it'd likely be cheaper to connect them to dark fiber than to a local ISP, otherwise you'd be murdered on bandwidth costs.

As far as I can tell from my own math so far, this is cheaper in the long run than operating in class A data center space, and encourages better systems architecture in terms of redundancy and reliability as well.
Steve Traugott | Homepage | 04.16.06 - 12:00 am | #