Data warehouses provide sophisticated features for aggregating,
analyzing, and comparing data to support decision making in
companies.

The most popular architecture for data warehouses are
multidimensional data cubes, where transaction data (called cells,
fact data or measures) are described in terms of master data (also
called dimension members). Master data is hierarchically organized
in dimensions, where the facts of the upper levels are computed
from the facts of the lower levels by some consolidation
functions.

Available OLAP (On-Line Analytical Processing) systems are
therefore able to deal with changing measures, e.g., changing
profit or turnover. Surprisingly, they are not able to deal with
modifications in dimensions, e.g., if a new branch or division
is established, although time is usually explicitly represented as
a dimension in data warehouses.

The COMET approach is a data warehouse metamodel which allows to
represent not only changes of transaction data, but also of
schema, and structure data. The COMET model can then be used as
basis of OLAP tools which are aware of structural changes and
permit correct query results spanning multiple periods and thus
different versions of dimension data.

This thesis consists not only of a formal metamodel, but also of
an implementation of the COMET approach. This implementation
serves as a layer between the data warehouse sources and a
standard front-end like Hyperion Essbase, Oracle Express or Cognos
PowerPlay. Therefore, it can be easily integrated into an existing
data warehouse architecture.