A survey of mapreduce based parallel processing technologies

Along with the increasing Big Data challenges, the MapReduce based systems are extensively welcomed, because of their remarkable simplicity and scalability. However, from the first day MapReduce is proposed, its argument with parallel DBMSs never stops, as it over-focuses on the scalability but overlooks the efficiency. Accordingly, extended systems are proposed in order to improve the performance on the limited scale clusters. In the meantime, traditional RDBMS technologies like structured data model, transaction, SQL, etc. are also getting more attention. This paper reviews such systems, from Google and also the third parties, trying to indicate the directions for the future research.