Data Warehouse Maintenance, Evolution and Versioning - Data Warehousing Design and Advanced Engineering Applications - page 184

Approach of Golfarelli,
Lechtenbörger, Rizzi and Vossen

Approach of SAP AG

Besides these scientific approaches, there are
also two commercial products including sup-
port for changing structures in data warehouses.
SAP AG (2000) presents an approach capable of
producing four different types of reports: a report
using today's constellation, a report using an old
constellation, a report showing the historical truth
and a report showing comparable results. This ap-
proach is limited to basic operations on dimension
members and does not allow to transform cell data
between structure versions.

Golfarelli, Lechtenbörger, Rizzi and Vossen (2004;
2006) present their approach for schema version-
ing in data warehouses. Based on a graph model
(called schema graphs ) of the data warehouse
schema they present their algebra for schema
modifications. This approach supports version-
ing, therefore, past versions are not lost. Based
on those schema versions the authors describe a
mechanism to execute cross-version queries, with
the help of so called augmented schemas. For cre-
ating such an augmented schema, an old schema
version is enriched with structure elements from
a subsequent version, so that the data belonging
to the old schema version can be queried as if it
followed the new version.

Approach of KALIDO

The KALIDO Dynamic Information Warehouse
(2004) supports some aspects of data warehouse
maintenance. Dealing with changes is realized by
the so called Generic Data Modeling. The data
warehouse model consists of three categories of
data: the transaction data , describing the activi-
ties of the business and the facts associated with
them, the business context data , which is the
analog to the instances, and the metadata , which
among others comprises parts of the schema.
With evolving the business context data, instance
evolution is supported.

Approach of Malinowski and zimanyi

Malinowski and Zimanyi (2006) present an
approach for representing time-varying data in
dimensions. Based on their MultiDimER (Ma-
linowsky & Zimanyi, 2004) they present a set
of temporal extensions that allow conceptual
representation of time-varying levels, attributes
and hierarchies. They describe the versioning
of levels and hierarchies. When describing time
varying levels, they actually refer to modeling
changes in the members contained in this level.
With respect to hierarchies, the authors distinguish
between temporal levels and non-temporal rela-
tions between them, temporal levels and temporal
relations between them, and non-temporal levels
and temporal relations between them. Based on
these scenarios the authors extend their Multi-
DimER metamodel to include the capability of
expressing these changes.Additionally, means for
transforming MultiDimER into a classical Entity
Relationship model are proposed.As the approach
is more concerned about conceptual modeling of
changes, considerations about dealing with cell
data are not included.

Comparison of Approaches

After having presented the various approaches,
Table 2 shows a classification of them with respect
to the following features:

1. Level of Maintenance: Does the approach

support schema maintenance, instance
maintenance or both of them?

2. Type of Historization: Does the approach

support versioning or evolution?

Of course, the desired feature combination
would be an approach supporting versioning on
schema and instance level, because this offers