The Research Of Metadata Management System In Data Integration

In data integration system, metadata is the most important, which provides a data map for the whole data. By metadata, the data structures, data sources and data targets in the system can be known.The existing metadata management tools often focus on the part of data integration process, such as the metadata of ETL tools and data warehouse tools. However, with more and more data, data integration becomes more and more complex. If we only know a part of metadata, it is difficult to meet the current demand of metadata management. So it is necessary to bring forward a metadata management framework, which can make us know all data structure of data sources, ETL processes and data targets and can exchange metadata between metadata databases. In this paper, a metadata management framework is proposed. It is a metadata management prototype, based on federation metadata database and CWM (Common Warehouse Metadata) standard and combined with the fact of the share database system in university data integration.The extensibility of metadata management tools is very important. With the development of business operation system, data integration rule will be changed possibly. So we must adjust our tools to adapt the change. In the metadata management tools, it is most difficult to minimize the change while to realize the integration rule. This paper provides a dynamic keyword method to meet the extensibility of tools. The tool is used to meet the dynamic demand for metadata by the maintenance of a keywords list and the path of dynamic link library (DLL). The DLL is matched for keywords to realize the certain function.ETL process is an important step during data integration which executes data integration driven by metadata. Thus, that how to describe, store and manage metadata is very important. In this paper, an ETL process is described logically and a simplified store model of ETL is presented.In order to increase data reliability, the cause and effect must be comprehended in data integration. It is explained in example that how to realize the tracing of a data pedigree in this paper.