When they say "grep", they appear to literally be running grep on four billion documents[1]. I'm assuming that is a significantly more expensive operation than

When they say "grep", they appear to literally be running grep on four billion documents[1]. I'm assuming that is a significantly more expensive operation than their normal search algorithm as it won't use whatever search tree they've got set up. 1. http://blog.blekko.com/2011/09/13/get-ready-to-grep-the-web/ http://blog.blekko.com/2011/09/13/get-ready-to-grep-the-web/