>In the Big Data world, we didn’t see much talk about columnar data until Google’s Dremel paper in 2010<p>Try 2005:<p><a href="http://db.csail.mit.edu/projects/cstore/#papers" rel="nofollow">http://db.csail.mit.edu/projects/cstore/#papers</a><p>We had a very large Vertica installation up in 2009.<p>But as Abadi has noted it's an in-memory format, and it removes a lot of bottlenecks at the CPU-Memory interface. In that sense it's quite new; we've been wrestling with tricky in-memory formats for some time.