From Conventional to Multiversion Data Warehouse - Evolving Application Domains of Data Warehousing and Mining - page 47

Figure 2. Different levels of abstraction in mul-
tiversion data warehouse

Figure 3. Version creation framework

frAMeWork of
verSIon creAtIon

evolution operations

It is not possible to clearly anticipate market and
business trends before the development of a data
warehouse (Golfarelli, 2004; Mitrpanont, 2006).
Therefore, it is difficult to capture all the evolving
requirements on data warehouses in advance. In
reality, changes in business strategies, changes in
business markets, establishing new departments,
independence of countries and redevelopment of
boundaries or other events may trigger evolution
operations. These are the operations which often
lead to development of new versions of the data
warehouse. These operations are (Morzy, 2004;
Balaschka, 1999): insert level, delete level, insert
attribute, delete attribute, connect attribute to
dimension level, disconnect attribute from dimen-
sion level, connect attribute to fact, disconnect at-
tribute from fact, insert classification relationship,
delete classification relationship, insert fact, delete
fact and insert dimension into fact. The following
brief introduction to these operations is adapted
from Blaschka's work (Balaschka, 1999):

As soon as a business change occurs, change in
operational sources is triggered and, as a result, opera-
tional sources evolve. However, the data warehouse
is not designed to handle operational source changes.
Therefore, a framework is needed that facilitates
adjustment of the data warehouse in order to adapt
to changes in the operational source.
The framework takes the dimensional schema
and evolution operations as input and produces
a new version of the data warehouse, while ex-
tending the versioning relation. The framework
called 'Versioning framework', is composed of
evolution operations, versioning graph and adapta-
tion process, as shown in Figure 3: a) Evolution
operations are responsible for triggering changes
in the data warehouse; b) Versioning derivation
graph shows high-level relationship between ver-
sions and c) Adaptation process is a phenomena
that can be used for adaptation of change in the
data warehouse.