Study And Application Of ODS Construction Strategy Based On Meta Data Technology

With the informatization construction for many years, there are a large number of heterogeneous OLTP systems being run at one time in an organization, and a mass of historical data were accumulated. Tri-level architecture DB-ODS-DW was came forth in order to supply global and consistent information view and satisfy different layer users' analysis requirement. The ODS takes a connecting link functions between the preceding and following in this tri-level architecture. The construction of ODS makes not only the global and consistent information view be possible to founded, but also the up-to-the-second OLAP and global OTAP can be implemented in it.Firstly, the definition, characteristic, types, difference and relation with DB and DW of ODS are introduced, and some foundation knowledge of meta data are described, such as definition, type, function etc. Then the generation, development, hierarchical model, and the meta data architecture advanced by CWM of Common Warehouse Metamodel are introduced especially.Therefore a method of constructing ODS system is proposed in the background of special equipments ODS system implementation. The method involves three models: metamodel, logical model and physical model. Metamodel is used to store relational physical model, depends on CWM, expands some CWM classes properly and then generates ODS metamodel with refering CWM classes. Metamodel is transformed to physical structure needed in fact via Java metadata interface program. In logical model, the problem of metadata storage and management is considered especially. After comparing and analyzing metadata storage and management technology, metastore technique is adopted in logical model. Physical model describes the process of constructing ODS system in detail. A general data exchange mechanism based on XML and a data transmission strategy based on WAN and heterogeneous environment is applied to realize data extracting in ODS system ETL function. In addition, the functions and implementation method of data transformation and data load are described.