Google machines Douwe Osinga is working for Google on the Google File System architecture. If you don't know what is the Google File System (GFS) you should read this paper called The Google File System (PDF 269 KB), by Sanjay Ghemawat, Howard Gobioff, and Shun-Tak Leung.

The paper describes how Google stores data and goes into some detail here. Basically Google runs one distributed filesystem over more than a thousand machines using thousands of disks in order to manipulate hundreds of terrabytes at a go. Data is safely duplicated and can be checkpointed. "It is really quite astonishing to work on stuff like that from the inside and you'd think that if you're going to keep stuff secret, this would be one thing. But it is not, it is a public paper.", Douwe Osinga says.

The Google File System Architecture

Douwe Osinga announces that "There's another paper out there about some of the stuff we're doing here that will really blow your mind"...