GeoBI Architecture Based on Free Software: Experience and Review - Geographical Information Systems: Trends and Technologies - page 255

Logical Design

The logical design phase focuses on the transformation of the conceptual
multidimensional schema into a logical schema and in the more detailed
specifi cation of the ETL processes. This schema is model dependent, but
tool independent, i.e., it can be used for implementing in different DBMSs
that rely on the same model, e.g., relational model. The mapping rules
from the MultiDim model into the relational model consider differently the
conventional and spatial levels (Malinowski and Zimányi 2008). For each
conventional level, e.g., Gender , Time , and Cancer type, a relation is created
that contains all attributes of the level and an additional attribute for a
surrogate key. To avoid having relations with only one or two attributes
that may affect system performance, we denormalize them, i.e., include the
level attribute(s) in the relation representing a fact relationship. This kind
of table is commonly known as degenerate or fact dimensions (Jensen et
al. 2010; Luján-Mora et al. 2006; Malinowski and Zimányi 2008).
The mapping of a spatial level requires the addition of an attribute
representing geometry. Furthermore, for representing relationships between
hierarchy levels traditional mapping rules are applied that consider
existing cardinalities. An example given in Table 2 shows a District level
with an additional attribute representing geometry and two foreign keys
for indicating its hierarchical relationship with the County and Health area
levels. This mapping gives a snowfl ake schema where normalized tables
are used for representing each level. However, this representation can be
modifi ed for performance reasons, using a star schema.
Finally, a table representing a fact relationship is mapped in the same
way as n -ary relationships in the ER model, i.e., it includes foreign keys of
all participating levels and attributes representing measures. However, not
all measures can be stored in the SDW. Some of them, e.g., the Incidence Rate
in the Cancer fact relationship in Fig. 4, that requires special calculations,
should be defi ned during SOLAP cube creation, as we will see in the
corresponding section.

Table 2. Logical representation of a District level.

Dimension
District

District
Key

District
ID

Partial
Code

Name Area Geom County

FK

Health
area
FK

Restrictions PK Unique Not null,

>0 and
< 20

Not null >0 Multipolygon FK FK

Data type Integer Integer Integer Character

(20)

Decimal
(10,2)

Geometry Integer Integer