An interesting article in the New York Times about how Google builds it’s Supercomputer Infrastructure suggests many of the ideas I had about Google for atleast two years (since I heard someone from Google speak at my workplace - the same day we had the Blackout on the entire East Coast - I think it was August 14th - or was it August 17th, 2004?  Don’t remember for sure).

A couple of years ago I realized that Google has saved a great deal of money by building their entire computer systems and network datacenters from the ground up - using a lot of cheap equipment by setting them up in several clusters - so no matter how many servers failed - or how many disks - there would be plenty of copies of the data on alternate systems - and Google’s response time would actually be better than anyone else’s.  And so it is - and we now have some basic numbers on what building their systems from the ground up saved Google - about 30% of their total budget!

"Mr. Reynolds estimated that Google’s computing costs are half those of other large Internet companies and a tenth those of traditional corporate technology users. "

The Times article brings up an interesting question - at what point does building your own infrastructure cease to be worthwhile?   It can be argued that the same approach that causes Google to succeed in Engineering may also cause it’s lack of focus in developing marketshare for it’s products outside of Search, as I wrote up in Google a one trick pony?