The Design and Implementation of Modern Column-Oriented Database

Just come across this article - http://db.csail.mit.edu/pubs/abadi-column-stores.pdf 

Anyone have comment about this?

array based software has been around for some time - it seems that folk are starting to catch on.

the ‘tradeoff’ with appending to arrays is losing locality (say, seek+write to different arrays on disk or memory).  but kdb+ seems to handle significant load (at least adequate for millisecond reaction).

i wonder for what scenarios it’s good to append to rows (rather than to columns) in file/memory.  a big win with arrays is the data can be similar, so compression is more effective.