Database Reference
In-Depth Information
dimensions, and hierarchies inferred from the analysis of the operational
database (Fig. 10.4 ). This led to the multidimensional schema shown in
Fig. 10.7 .
Validate Conceptual Schema with Users
The initial data warehouse schema as presented in Fig. 10.7 should be
delivered to the users. In this way, they can assess its appropriateness for
the analysis needs. This can lead to the modification of the schema, either by
removing schema elements that are not needed for analysis or by specifying
missing elements. Recall that in the source-driven approach, during the
requirements elicitation the users have not participated, thus changes to the
initial conceptual schema will likely be needed.
Develop Final Conceptual Schema and Mappings
The modified schema is finally delivered to the users. Given that the
operational schema of the Northwind database is very simple, the mapping
between the source schema and the final data warehouse schema is almost
straightforward. The implementation of such a mapping was described in
Chap. 8 , thus we do not repeat it here. Further, since we already have
the schemas for the source system and the data warehouse, we can specify
metadata in a similar way to that described for the analysis-driven approach
above.
10.4.5 Analysis/Source-Driven Conceptual Design
In the analysis/source-driven approach, two activities are performed, target-
ing both the analysis requirements of the data warehouse and the exploration
of the source systems feeding the warehouse. This leads to the creation of two
data warehouse schemas (Fig. 10.9 ). The schema obtained from the analysis-
driven approach identifies the structure of the data warehouse as it emerges
from the analysis requirements. The source-driven approach results in a
data warehouse schema that can be extracted from the existing operational
databases. After both initial schemas have been developed, they must be
matched. Several aspects should be considered in this matching process, such
as the terminology used and the degree of similarity between the two solutions
for each multidimensional element, for example, between dimensions, levels,
attributes, or hierarchies. Some solutions for this have been proposed in
academic literature, although they are highly technical and complex to
implement.
Search WWH ::




Custom Search