Key points are not available for this paper at this time.
MapReduce advantages over parallel databases include storage-system independence and fine-grain fault tolerance for large jobs.
Dean et al. (Mon,) studied this question.