Greg Linden has published an interesting analysis of the column in The Database Column comparing MapReduce with DBMSs.
Most relevantly, he notes:
The most compelling part of the post for me is their argument that some algorithms require random access to data, something that is not well supported by GFS, and it is not always easy or efficient to restructure those algorithms primarily to do sequential scans.
Which I would fully agree with. If you require random access to data, MapReduce is not likely to be the best tool to use and they realize that.
