Abstract: Here is database integration in which a relational table and an analytic view are generated and configured in a relational database for acceleration and optimally populated from summarization hierarchies derived from a dimension in a multidimensional database. This approach is a new way to transfer online analytical processing (OLAP) metadata from a multidimensional database to a relational database. Shared members may cause this approach to generate a directed acyclic graph (DAG) that organizes the metadata. The OLAP content generated in the relational database has a novel data topology that is unconventional because relational data normalization strategies such as third normal form (3NF) are avoided. This novel data topology achieves unprecedented data locality of metadata in a relational database. Due to optimization techniques such as graph coloring for detection and resolution of conflicts caused by shared members, the accuracy and speed of the data integration between the two databases is increased.

Abstract: Here is database integration in which a relational table and an analytic view are generated and configured in a relational database for acceleration and optimally populated from summarization hierarchies derived from a dimension in a multidimensional database. This approach is a new way to transfer online analytical processing (OLAP) metadata from a multidimensional database to a relational database. Shared members may cause this approach to generate a directed acyclic graph (DAG) that organizes the metadata. The OLAP content generated in the relational database has a novel data topology that is unconventional because relational data normalization strategies such as third normal form (3NF) are avoided. This novel data topology achieves unprecedented data locality of metadata in a relational database. Due to optimization techniques such as graph coloring for detection and resolution of conflicts caused by shared members, the accuracy and speed of the data integration between the two databases is increased.

Abstract: Techniques are described herein for automatically discovering undeclared PK-FK relationships between dimension tables and a fact table in situations where: the primary keys of the dimension tables (a) are not known, and (b) are composite keys, and the foreign keys of the fact table (a) are not known, and (b) are composite keys. The techniques involve removing from consideration all fact table columns that satisfy first removal criteria, and all dimension table columns that satisfy second removal criteria. Fingerprints are generated for the fact and dimension table columns that remain. Matched pairs (e.g. two fact table columns and two dimension table columns) are generated based on the columns that have not been filtered. The matched pairs are further filtered based on third removal criteria. The matched pairs that remain are then graded, and a PK-FK relationship is established based on the matched pair with the highest grade.