https://www.facebook.com/notes/facebook-engineering/tao-the-power-of-the-graph/10151525983993920
www.theregister.co.uk/2013/06/27/facebook_tao/
QuoteIt has been built to deal with over a billion reads per second across a data set "of many petabytes," Facebook said. Tao was designed by Facebook to better link together data kept in its main data store (MySQL) and caching layer (memcache), while being able to deal with unpredictable queries on objects.
FB are releasing a more detailed paper in a few days that gets more technical about it. edit, here it is http://0b4af6cdc2f0c5998459-c0245c5c937c5dedcca3f1764ecc9b2f.r43.cf2.rackcdn.com/11730-atc13-bronson.pdf
I've been reading up on graph DB's & triplestores the past few days... maybe FB one's could be called a tripestore ;o)
The scale they've achieved is impressive to say the least.
And Facebook search is still crap.
It's almost impossible to find people you aren't directly connected with. Am trying to find old friends from my old home town - it's in my profile, but when I search their names, FB still returns people from the US.
If FB implemented a search based on location and/or likes, school, etc, it would be a killer.
Instead, it's great for people near you, useless for those who aren't.
You can search for schools etc here
https://www.facebook.com/find-friends/browser
& of course their new graph search
https://www.facebook.com/about/graphsearch
But right enough, to search outside your circle, graph search is no good.