MapReduce advantages over parallel databases include storage-system independence and fine-grain fault tolerance for large jobs.