Database Reference
In-Depth Information
CURRENT DATA WAREHOUSE
DESIGN APPROACHES
modeling from an E/R diagram. Their method
produces a fact schema by: 1) defining facts from
relationships or from frequently updated entities;
2) building an attribute tree for each fact, useful to
construct a fact schema; 3) pruning and grafting
the attribute tree in order to eliminate unneces-
sary levels of detail; 4) defining dimensions; 5)
defining measures; and 6) defining hierarchies.
The third, fourth and fifth steps are supported by
the preliminary workload declared by decisional
users.
Also starting from a given E/R diagram,
(Cabibbo, & Torlone, 1998) propose a method to
build a multidimensional database in four steps: 1)
identifying facts, dimensions and measures through
a thorough, manual analysis of the given E/R dia-
gram; 2) restructuring the E/R diagram to describe
facts and dimensions in a more explicit way. The
produced version of the E/R diagram can be directly
mapped onto the source data model; 3) deriving a
dimensional graph that succinctly represents facts
and dimensions from the restructured E/R diagram;
and finally, 4) translating the dimensional graph
into the multidimensional model. In this method,
the first step which is the most crucial is manually
done. In this method, the designer must have do-
main expertise to correctly identify all potentially
needed multidimensional elements.
(Moody, & Kortnik, 2000) propose a three-
step method to design a DM/DW also from an
E/R data model: 1) entity classification which
classifies entities into three categories: transaction
entity (describes an event that happens at a point
in time), component entity (directly related to a
transaction entity via a one-to-many relationship)
and classification entity (related to component
entities by a chain of one-to-many relationships)
2) hierarchy identification using to-many rela-
tionship and 3) data mart schema development
for each transaction entity. Each schema is a star
schema represented through a fact table and a
number of dimension and sub-dimension tables.
Separates star schemas can be combined to form
constellations or galaxies.
Data-driven (or bottom-up) development ap-
proaches rely on the analysis of the corporate data
model and relevant transactions (List, Bruckner,
Machacze, & Schiefer, 2002), cf., (Golfarelli,
Maio, & Rizz, 1998), (Cabibbo, L., & Torlone, R.
1998), (Moody , & Kortnik, 2000), (Prat, Akoka ,
& Comyn-Wattiau, 2006), (Zribi, & feki, 2007),
(Golfarelli, Rizzi, & Vrdoljak, 2001), (Vrdoljak,
Banek, & Rizzi, 2003), (Jensen, Møller, & Ped-
ersen, 2001). Data-driven approaches were justi-
fied by Bill Inmon (Inmon, 1996) by the fact that
unlike transactional systems whose development
lifecycle is requirements-driven, decision support
systems (DSS) have data-driven development
lifecycle. In addition, Inmon argues that require-
ments are the last thing to be considered in a DSS
development; they are understood after the data
warehouse is populated with data and query results
are analyzed by the decision makers. Hence, data-
driven approaches enjoy a double advantage: they
reduce the task of decision makers by proposing
potential analytical needs and they guarantee that
the enterprise's information system can feed the
selected needs with pertinent data.
Considering the advantages of data-driven
approaches, we elected to propose a data mart
design method within this category. Hence, we
next limit ourselves to over viewing works perti-
nent to data-driven approaches where the majority
starts from conceptual schemas modelled through
E/R, XML models, or UML cf. (Prat, Akoka, &
Comyn-Wattiau, 2006) and (Zribi, & feki, 2007).
We focus on those works that start from E/R and
XML models since they are more pertinent to
our method.
E/R Diagram-Based
Design Approaches
(Golfarelli, Maio, & Rizz, 1998) propose a semi-
automated method to carry out a DM conceptual
Search WWH ::




Custom Search