Metadata Standards considers various standard efforts for modelling, implementation and interchange as well as storage of metadata within a business intelligence (BI) environment. Furthermore, it responds to the question: What kind of methodology standard is most suitable for metadata designing and integration? This leads particularly to the CWM which provides a seamless system integration and interoperability of metadata in BI solutions. Based on the fact that CWM is built upon the core metamodelling technologies Meta Object Facility (MOF), Unified Modeling Language (UML) and XML Metadata Interchange (XMI), this section gives also an abstract about the CWM reference model.

Content:

• 1. The CWM solution
• 2. An architectural overview of CWM
• 3. The CWM components

The CWM specification is a reference model for integrating shared metadata in a standardised data warehouse (DWH) and BI analysis environment. It is comparable with a “set of construction plans”, but in a higher abstraction level [OMG02, page 78]. With regard to metadata representation and interchange, IBM and Oracle support particularly the CWM standardisation efforts [OMG02, page 233]. A general characteristic of the CMW standard is the embedding of a “model-based approach to meta data integration and interoperability” [OMG02, page 36] as well as interchange and reusability in conjunction with multivendor capability. Furthermore, the CMW accomplishes the requirement of syntactical and semantical integration into a homogeneous metamodel “to construct meta data (for example, models or instances of the metamodel) describing all components” [OMG02, page 36] such as information objects in BI layers (cf. Five-Tier BI Architecture). The OMG standard fulfils the requirement to present shared meta-data in a uniform format. In addition, the CWM standard is based on the established UML notation (cf. UML) which enables to abstract complex correlations in the software engineering. Thus, the CMW standard uses the “object-oriented concept of inheritance to extend the modeling elements” of UML [OMG02, page 38] to specify the complex structures and relationships of DWH and BI requirements.

The CWM reference model is available as a set of UML classes and divided into 21 separate metamodel packages [OMG02, page 81] as shown in their assigned layers in figure 13. Further, the reference model is organised into five thematic layers which correspond to an important functional BI area [OMG02, page 36]. Also, each of these thematic layers includes CWM packages for a specific domain of a BI solution. A package includes classes, associations and constraints and can be depend on another packages [OMG02, page 81]. Figure 13 contains 22 packages, which is due to the fact that “the Object package in the Resource Layer does not exist as a separate package” [OMG02, page 82].

Figure 1: CWM metamodel layer architecture [OMG02, page 37].

• Object Model: In the Object Model layer is the modelling language defined, on which every other layer is set up.
• Foundation: The Foundation layer defines metamodels for the specification of basic modules like data types, expressions, type casting and index constructs as well as model elements for business information or software system architectures.
• Resource: Within the scope of the Resource layer, it contains packages for the definition of various structures of data resources, and target systems comprising a BI solution. Regarding this aspect, “note, that object-oriented databases are not explicitly represented, because the Object layer as a subset of UML is replete with object-oriented modeling concepts” [OMG02, page 39].
• Analysis: The significant Analysis layer embraces modelling constructs of data transformation processes and various forms of multidimensional data analysis. It describes the mapping rules between the data source and the target system.
• Management: Finally, the Management layer focuses on the entire BI processing and contains elements that are needed for controlling and monitoring of technical transformation operations.

In general, the large scale integration of the reference model is based on the fact that every CWM class is integrable into an inheritance hierarchy. New constructs are only added when corresponding elements are not available.