Database Concepts - Data Warehouse Systems: Design and Implementation - page 12

Designing a database system is a complex undertaking typically divided
into four phases, described next.
￿ Requirements specification collects information about the users' needs
with respect to the database system. A large number of approaches for
requirements specification have been developed by both academia and
practitioners. These techniques help to elicit necessary and desirable
system properties from prospective users, to homogenize requirements, and
to assign priorities to them. During this phase, the active participation of
users will increase their satisfaction with the delivered system and avoid
errors, which can be very expensive to correct if the subsequent phases
have already been carried out.
￿ Conceptual design aims at building a user-oriented representation of the
database that does not contain any implementation considerations. This
is done by using a conceptual model in order to identify the relevant
concepts of the application at hand. The entity-relationship model is one of
the most often used conceptual models for designing database applications.
Alternatively, object-oriented modeling techniques can also be applied,
based on the UML (Unified Modeling Language) notation.
Conceptual design can be performed using two different approaches,
according to the complexity of the system and the developers' experi-
ence:
- Top-down design: The requirements of the various users are merged
before the design process begins, and a unique schema is built.
Afterward, a separation of the views corresponding to individual users'
requirements can be performed. This approach can be dicult and
expensive for large databases and inexperienced developers.
- Bottom-up design: A separate schema is built for each group of
users with different requirements, and later, during the view integration
phase, these schemas are merged to form a global conceptual schema
for the entire database. This is the approach typically used for large
databases.
￿ Logical design aims at translating the conceptual representation of the
database obtained in the previous phase into a particular logical model
common to several DBMSs. Currently, the most common logical model
is the relational model. Other logical models include the object-relational
model, the object-oriented model, and the semistructured model. In this
topic, we focus on the relational model. To ensure an adequate logical
representation, we specify a set of suitable mapping rules that transform
the constructs in the conceptual model to appropriate structures of the
logical model.
￿ Physical design aims at customizing the logical representation of the
database obtained in the previous phase to a physical model targeted