Research On CWM Based Metadata Management

Data warehouse is a novel technology. It has brought great benefits and its importance is being recognized more and more. Metadata play a key role in data warehouse because it is the foundation of comprehending the operations between different tools. However, venders nowadays adopt different metadata standards and metadata management strategies, which result in much difficulty of integrating and sharing metadata in data warehouse. Aiming at this problem, OMG puts forward CWM .CWM describes all aspects of data warehouse to the best of its abilities. Furthermore, the MOF metamodeling architecture where CWM locates affords CWM profound application background and technology support. Now many companies declared to support CWM and will develop their next-generation products based on CWM.On the basis of the research on MOF metamodeling architecture, we analyze CWM comprehensively, examining the content and makeup of metadata and illustrating the dependency among packages in CWM, based on which we bring forward a metadata management solution WMMS. WMMS integrates modeling tool, metadata transformation tool, searching engine and storage interface etc. The purpose of WMMS is to provide powerful ability in metadata management and can be well adapted in distributed environment.In this paper, we mainly design the modeling tool and metadata transformation tool in WMMS and implement the latter. Modeling tool takes advantage of the natural relation with UML and redevelops the Rational Rose based on REI to support CWM modeling. We also expatiate the basic function and modules in modeling tool, advance a series of model-verifying mechanisms to ensure the correctness of metadata as much as possible.Metadata transformation tool adopts XMI as the metadata exchange standard. Its function is to transform the metadata produced in modeling tool to XMI format. We define their mapping in accordance with XML Document production rule and bring forward some optimized data structures and two transformation algorithms in order to transform efficiently. Finally, in view of the limit of CWM, we discuss how to express and transfer the metadata that are not based on CWM as the basis of next development of modeling tool and metadata transformation tool.