I don't have time to watch the whole thing right now, but I don't buy 12TB/day for a second. Facebook has what, on the order of 1e8 users, doing a handful of updates per day which are text snippets of a dozen bytes each. Not even close.<p>Maybe they're including things like posted image data? That's schemaless and static, and not really a "data warehouse" problem as such. Someone enlighten me, because this doesn't add up.