ETL　is　a　key　link　in　the　construction　of　data　warehouse.　On　the　base　of　analyzing　the　mainstream　ETL　tool　Datastage,　the　data　extraction,　transformation　and　loading,　proposes　a　ETL　framework　based　on　data　processing,　and　the　realization　method　and　steps　are　discussed　in　detail.　The　framework　uses　HIVE　as　a　data　processing　station,　improve　the　operating　efficiency　of　the　file;　data　task　according　to　the　E,　T　and　L　three　parts　and　hierarchical　partitioning,　conversion　of　data　users　to　better　grasp　the　process;　development　data　using　the　configuration　file　of　the　task,　the　development　personnel　free　out　from　the　heavy　code,　will　to　shift　the　focus　of　the　work　to　the　data　logical　task,　which　has　greatly　improved　the　efficiency　of　development　personnel　data　processing.