Research In Data Warehouse Technology

In the year of 90s of 20th century, development of traditional DBMS got into low tide. Database providers of the entire world were obliged to find new technology and economic incremental point by cruel market competition. One of the hotspot was Object-RDBMS such as Oracle9i. The other was BI which was based on Data Warehouse technology.The technique of OLAP and DataMining became more and more mature, so that the application market of Data Warehouse was built a good foundation. After the year of 2000, applications of OLAP and DM based on DW were developped in traditional data dense trade as finance, insurance, negotiable securities, telecom and tax. However, since the basic information equipments of China were not consummate and versatility and agility of data source system did not meet the requirment, when new demands and increment services based on the present data warehouse were brought forward, supply of original data begame bottle-neck of persistent development of the whole system.Redesign and realize the clear up and upfront work of data source according to new demands is unbearable on waste of time and manpower.The article described development and analysis status of UBISP(United Business Information Service Platform) system, analyzed the Extract strategy and optimized algorithm of ETL rules, discussed the data capacity and granularity selection in DW modeling, raised an idea of how to sub divide dimension table and fact table's granularity by compound primary key aimed at above question. This system is operated on the platform of UNIX (IBM AIX), using PowerDesigner as the modeling tool, based on the DB storage of Oracle9i, carried out ETL works with DataStage, built multi-dimension cube and laid out the result upfront by Cognos. Design and research of the whole DW system was deeply and particularly set forth here, this may make help to Unicorn's DW project period 2 on theory.