Goal-Oriented Conceptual Database Design
A successful information system is the one
that meets its design goals. Expressing the goals and subsequently
translating them to a working solution is a major challenge.
In the past 15 years, Requirements Engineering (RE) research
has more and more recognized the important role that goals
play in the RE process. This recognition has led to a new
stream of research on goal modeling, goal specification and
goal-based reasoning for multiple purposes, such as requirements
elicitation, specification and verification. At the same time,
the state-of-the-art for designing the database part of an
information system hasn't made much progress. The process
of database design shares many similarities with that of software
development as both are dependent on an analysis of user requirements.
The proposed research aims at incorporating goal-orientation
in data modelling and offers methodological support that leads
to a goal-oriented database design methodology, which will
enjoy many of the benefits of the Goal-Oriented Requirements
Engineering (GORE) approaches for software development. These
benefits include systematic exploration of alternatives and
explicit traceability of rationale.
Design Process
The first step of this work is to validate
our hypothesis that goal-orientation in database design results
in better design in terms of coverage of stakeholder goals
and generates schemas with rich and explicit data semantics.
We provided evidence that this hypothesis holds in the practice
by conducting a case study on the design of a real-world,
industrial biological database. The result was reported in
RE'06. Moving
forward in the direction, we have proposed a goal-oriented
conceptual database design process which extends traditional
methodologies with an early analysis of stakeholder goals,
and derive from them a regular conceptual schema in a systematic
way. This process was reported in RE'07.
Design Dimensions
One way to view a design problem is by taking
the analogy of a search of the ˇ°optimalˇ± solution in a multidimensional
space, where each design dimension corresponds to
one type of design issue to be addressed, and gives
rise to a set of design alternatives with different
degrees of support for that issue. The entire design space
is the cross product of all the dimensions, and a point in
design space corresponds to a complete design. Various
types of design criteria are used to evaluate and
select among design alternatives.
After surveying a number of design dimensions
including time, unit, quality, provenance and privacy, in
this work we focus on the data quality (DQ) dimension. DQ
itself is normally viewed as a multi-dimensional and a hierarchical
concept, including sub-dimensions such as accuracy, timeliness,
consistency. We proposed a general quality design process,
which extends our previously proposed goal-oriented conceptual
database design, to offer an integrated framework for address
both application specific data requirements (AppData)
and quality assurace data requirements (QAData),
as shown in the following diagram. This quality design process
has been reported in ICIQ'07.
Prototype & Evaluation
The next step is to implement a prototype
design environment to support the proposed design process
and to evaluate of the methodology using either analytical
or empircal approach (work in process).
|