There was a thread on one of the SQL newsgroups where we were discussing Google's databases and the hardware they use to run them.  Someone pointed us to a link by one of Google's chief engineers describing their proprietary database and indexing systems designed in-house (as it turns out most of their systems are not SQL-based), and the "throw-away" servers they use to run it all.  Interesting stuff.

I'll see if I can locate the articles and post the links here later.