Database Reference
In-Depth Information
warehouse development strategy on choosing the core business processes to
model. Then, business users are interviewed to introduce the data warehouse
team to the company's goals and to understand the users' expectations of
the data warehouse. Even though this approach lacks formality, it has been
applied in many data warehouse projects.
There are several methods for requirements analysis based on the source-
driven approach: Bohnlein and Ulbrich-vom Ende [
15
]proposedamethod
for deriving logical data warehouse structures from the conceptual schemas
of operational systems. Golfarelli et al. [
68
] presented a graphical conceptual
model for data warehouses called the Dimensional Fact Model and proposed
a semiautomatic process for building conceptual schemas from operational
entity-relationship (ER) schemas. Cabibbo and Torlone [
23
]presenteda
design method that starts from an existing ER schema, deriving a multidi-
mensional schema and providing an implementation of it in terms of relational
tables and multidimensional arrays. Paim et al. [
153
]proposedamethodfor
requirements specification consisting of the phases of requirements planning,
specification, and validation. Paim and Castro [
152
] extended this method by
including nonfunctional requirements, such as performance and accessibility.
Vaisman [
210
] proposed a method for the specification of functional and
nonfunctional requirements that integrates the concepts of requirements
engineering and data quality. This method refers to the mechanisms for
collecting, analyzing, and integrating requirements. Users are also involved
in order to determine the expected quality of the source data. Then, data
sources are selected using quantitative measures to ensure data quality. The
outcome of this method is a set of documents and a ranking of the operational
data sources that should satisfy the users' requirements according to various
quality parameters.
As for the combination of approaches, Bonifati et al. [
16
]presenteda
method for the identification and design of data marts, which consists of three
general parts: top-down analysis, bottom-up analysis, and integration. The
top-down analysis emphasizes the users' requirements and requires precise
identification and formulation of goals. On the basis of these goals, a set of
ideal star schemas is created. On the other hand, the bottom-up analysis aims
at identifying all the star schemas that can be implemented using the available
source systems. This analysis requires the source systems to be represented
using the ER model. The final integration phase is used to match the ideal
star schemas with realistic ones based on the existing data.
10.10 Review Questions
10.1
What are the similarities and the differences between designing a
database and designing a data warehouse?
10.2
Compare the top-down and the bottom-up approaches for data
warehouse design. Which of the two approaches is more often used?
Search WWH ::
Custom Search