| Previous | Table of Contents | Next |
MDC-OIM provides an OLAP Schema metamodel for describing the use of multidimensional database technology within the enterprise
in support of advanced business analytics and decision support capabilities. OLAP technology has broad applicability, both
within the data warehousing environment, specifically, and across the enterprise, in general. Hence, both CWM and OIM have
a requirement for representing OLAP and multidimensional metadata.
The CWM and MDC-OIM OLAP metamodels have many similarities, but many fundamental differences, as well. Perhaps the most fundamental
difference is in the overall orientation of the two metamodels.
The CWM OLAP metamodel is a pure, semantic model of general OLAP concepts, and does not define any particular logical or physical
deployment constructs of its own. This is done for two reasons:
• OLAP and multidimensional concepts (what the user sees) tend to be rather abstract in nature and very broad in applicability; for example, notions such as “dimension? and “dimensioned variable? are concepts that span the enterprise and really aren’t specific to any particular technology that provides computational support for such concepts.
• OLAP concepts may be implemented in many different ways, depending on the objectives of the enterprise and the technologies available. For example, OLAP applications are often implemented using either relational database technology (ROLAP), multidimensional database servers (MOLAP), or some hybrid mixture of both relational and multidimensional technologies.
So the CWM OLAP metamodel defines generic OLAP concepts only and leverages the CWM Transformation metamodel to map OLAP metaclasses
to metaclasses of other packages that could be used to describe logical models of implementations (for example, the CWM Relational
and Multidimensional metamodels). Those logical models, in turn, rely on the Software Deployment metamodel to describe their
actual, physical deployments.
The MDC-OIM OLAP model, on the other hand, is largely derived from the OIM Database Schema model (in the same manner that
the Data Transformation model is). For example, Cubes and Partitions are ultimately derived from ColumnSet. This may have
the effect of restricting the usage of the OIM OLAP model to the representation of relational-OLAP constructs only.
The OIM OLAP model also includes a number of logical and physical deployment metaclasses, such as OLAPServer, DataSource,
and Connection metaclasses, plus DeployedOLAPDatabase and LogicalOLAPDatabase subclasses, in keeping with the OIM’s overall
dichotomization of the concepts of logical versus deployed subclasses. As stated earlier in the discussion on the relational
Database Schema, there is no need for the CWM OLAP metamodel to include these kinds of metaclasses, since logical descriptions
are implicitly defined by transformation mappings of OLAP semantics to more logical constructs (for example, relational),
and the physical deployment metaclasses are provided within a single, Software Deployment metamodel.
Areas where the CWM OLAP and OIM OLAP metamodels are mostly (though not completely) similar include the following:
• Cubes and Dimensions. Both metamodels support the concept of Cubes and Dimensions being separate from one another and both contained within an OLAP Database (called Schema in CWM). Both support the special designation of a Time Dimension, although the CWM OLAP metamodel further defines a Measures Dimension. Both metamodels also support the concepts of virtual versus physical Cubes, as well as the concept of a Cubes being composed from sub-cubes (called Cube Regions by CWM and Partitions by OIM). However, OIM includes the notion of an Aggregation metaclass, which represents pre-calculated aggregations in relational stores, generally what one might find in a typical, relational Star-schema deployment of OLAP. CWM provides no such concept, because this is regarded as being an implementation detail that would be addressed at the model instance level.
• Levels and Hierarchies. Both OLAP metamodels support the concept of Hierarchy as being a separate entity from its owning Dimension. Both metamodels support the concept of multiple Hierarchies per Dimension. Both metamodels also support the concepts of Dimension Levels and the association of Dimension Levels with Dimension Hierarchies, and both also define mapping constructs that enable Hierarchies and Levels to be mapped to logical deployment structures. However, within the OIM OLAP metamodel, these deployment mappings are explicitly geared toward a relational database (and optionally Star-Schema) deployment, whereas the CWM OLAP contains mapping constructs that derive from more general CWM Transformation mapping metaclasses and, hence, can be used to specify deployment mappings to any conceivable logical structure that might be supported elsewhere within the CWM metamodel.