Data Warehouse Maintenance, Evolution and Versioning - Data Warehousing Design and Advanced Engineering Applications - page 186

warehouse, maintenance methodologies can also
be used to facilitate so called what-if-analysis.
Bebel, Wrembel and others (2004, 2007) present
an approach for the management of multiversion
data warehouses. They differentiate between real
versions and alternative versions. The former are
used to historicize data warehouse modifications
resulting from real world changes.Alternative ver-
sions facilitate to create several versions, each of
them representing a possible future situation and
then apply what-if-analysis on them. Addition-
ally, alternative versions may be used to simulate
data warehouse modification for optimization
purposes.
Shahzad, Nasir and Pasha (2005) present
a similar approach for enabling evolution and
versioning in data warehouses that supports the
simulation of business scenarios with alterna-
tive versions. Based on the formal model of a so
called versioned multidimensional schema, they
introduce a set of operations on the schema and
instance level. All the operations are defined by
a changing-algebra defined upon the multidi-
mensional model. The versioning function sup-
ports both versioning and evolution. To support
simulation of business scenarios so called real
versions and alternative versions may be created.
The problem of data transformation following the
change operations is not dealt with.
Another instance of data warehouse main-
tenance is the so called View Maintenance. The
classical data warehouse maintenance deals with
structural changes, and often assumes the data
warehouse structure to be rather independent from
the underlying sources. View maintenance, on
the other hand, interprets the data warehouse as
a materialized view of the sources. Thus, changes
in the sources directly affect the data warehouse.
For instance, Zhuge, Garcia-Molina, Hammer and
Widom (1995) present an approach to synchronize
changes in the source data to the materialized
view. The main problem here is to decide whether
to update or to recalculate the view from scratch.
But as this, and also similar, approaches only deal

with changes of transaction data they are out of
scope for this chapter. Bellahsene (2002) presents
an approach for structural view maintenance, i.e.
updating the view definition with changes from the
underlying data sources. The presented operations
only cover addition and deletion of attributes of
the source schema. Also here the main question
is, whether an update of the view is possible and
cheaper than recalculation.

FUTURE TRENDS

Current commercial systems assume that the data
warehouse structure is defined at design time and
does not change afterwards. Therefore the support
of structure modifications is rather limited. On
the other hand, real-world systems are used in
evolving application domains. Thus, the demand
for modifications is present, because the system
has to be consistent with its application domain
and requirements. Despite the fact that more effort
is put into integrating maintenance capabilities
into commercial data warehouse systems, cur-
rent products are still not well prepared for this
challenge.
Schema and instance maintenance are quite
elaborated in current research approaches, but
efficient cell data transformation between differ-
ent structure versions is still subject to research.
The two main problems with data transforma-
tion are first of all to define semantically correct
transformation functions, which express the
user's requirements and expectations. The second
problem is the huge amount of data, which has
to be dealt with. Related to data transformation
are also multiversion queries. The problem with
such queries is again their semantic definition, i.e.
whether and how to include cell data related to
elements which are not valid of the whole period
of consideration.