Abstract: Because MapReduce lacks support for high-level ETL specific constructs,this paper presented a parallel dimensional ETL framework based on MapReduce (MapReduce Distributed ETL--MDETL),which exhibits the data processing to the composable property(the processing of dimensions and facts),directly supports high-level ETL-specific dimensional constructs．This paper evaluated its performance on large realistic data sets.The experimental results show that MDETL achieves very good scalability.

Key words: ETL,MapReduce,MDETL,Dimensions,Facts