Today Slashdot has picked up a storycolumn-oriented databases. The story claims that one size fits all approach does not work well for the current data warehousing requirements and that the organizations should explore other options beyond legacy RDBMS. The post says "Hence, my prediction is that column stores will take over the warehouse market over time, completely displacing row stores."
The fundamental assumption here is that somehow the data warehousing solutions are drastically different than OLTP ones and that's why has different storage, or I should say access, needs. What the post is missing is that many modern OLTP applications require real time analytics side-by-side and cannot really depend upon a separate data warehousing. The technology such as in-memory databases and materialized views that run on top of OLTP RDBMS make it feasible for an application provider to just have one hybrid system - OLTP or data warehousing, whatever you want to call it. This was obviously not the case few years back and you could get shot if you propose to run your analytics on a production (OLTP) database. I do believe that there is a need for special purpose databases that are different in architecture for very specific kind of applications but RDBMS is far from being obsolete. I heard the similar arguments in the past when object oriented database vendors claimed that RDBMS would become obsolete when people would switch over to object-oriented programming languages. Deja vu all over again!