A Method for Data Warehouse Design - Data Warehouse Systems: Design and Implementation - page 411

￿ The integration process to determine whether the data sources cover the
users' requirements may need complex techniques.
10.8 Summary
In this chapter, we have presented a general method for the design of data
warehouses. Our proposal is close to the classic database design method and
is composed of the following phases: requirements specification, conceptual
design, logical design, and physical design. For the requirements specification
and conceptual design phases, we have proposed three different approaches:
(1) the analysis-driven approach, which focuses on analysis needs; (2) the
source-driven approach, which develops the data warehouse schema on the
basis of the structures of the underlying operational databases, typically
represented using the entity-relationship or the relational model; and (3) the
analysis/source-driven approach, which combines the first two approaches,
matching the users' analysis needs with the availability of data. The next
phases of the method presented correspond to those of classic database
design. Therefore, a mapping of the conceptual model to a logical model
is specified, followed by the definition of physical structures. The design of
these structures should consider the specific features of the target DBMS
with respect to the particularities of data warehouse applications.
10.9 Bibliographic Notes
Given the lack of consensus about a data warehouse design methodology,
we comment in some detail the most well-known approaches to this topic.
Golfarelli and Rizzi [ 65 ] presented a data warehouse design method composed
of the following steps: analysis of the information system, requirements
specification, conceptual design workload refinement and schema validation,
logical design, and physical design. This method corresponds to the one used
in traditional database design, extended with an additional phase of workload
refinement in order to determine the expected data volume. Lujan-Mora and
Trujillo [ 119 ] presented a method for data warehouse design based on UML.
This proposal deals with all data warehouse design phases from the analysis of
the operational data sources to the final implementation, including the ETL
processes. Jarke et al. [ 96 ] proposed the DWQ (Data Warehouse Quality)
design method for data warehouses, consisting of six steps, focusing on data
quality concepts.
Regarding requirements specification following the analysis-driven
approach, Mazon et al. [ 130 ] propose to include business goals in data
warehouse requirements analysis. These requirements are then transformed
into a multidimensional model. Kimball et al. [ 103 , 104 ] base their data