There's already lots of different types of "database systems" floating around. Map/Reduce approaches to large-scale data processing has been getting most of the hype in the F/OSS space and columnar approaches in the enterprise spaces.<p>An area that is finally starting to get some love are approaches like Mark Logic's "XML contentbase" approach -- native XML data store that's wicked fast. Of course, I'm biased but I think MarkMail (<a href="http://markmail.org/" rel="nofollow">http://markmail.org/</a>) is a pretty slick, public example of it in action.
This would be easier to understand if Aster Data's website (<a href="http://www.asterdata.com/" rel="nofollow">http://www.asterdata.com/</a>) didn't seem like a bunch of marketing jargon.