A Deep Look In How Google Search Works
ZDNet has a fascinating article based on a talk by Google's VP of engineering detailing how Google manages searches and servers all across the globe. The level of detail is amazing but still easily understood, and it is definitely worth the time it takes to read. Some tidbits:

• Google's last complete system failure was in February 2000, and the current system makes a future one virtually impossible.
• Google's servers are so cheap, they just refer to them as PCs.
• Google File System, or GFS, will never be useful for regular computers, since it has 64 megabyte clusters, as opposed to the usual 2KB.
• Google's spell checker is based entirely on computational power; it is self-learning, and Google is trying to bring that to contextual results.
• Google would like to give you a "remove all commercial results" button, but cannot do it as a company, at least at the moment.

(via Greg Linden)

- posted by Nathan Weinberg @ 12/01/2004 04:08:00 PM