XMI and the Many Metamodels of Enterprise Metadata

Track: Metadata and Semantics, Core Technologies, Case Studies

Audience Level: High Level/Technical view

Time: Tuesday, November 15 16:45

Author: Joram Borenstein, Unicorn Solutions

Author: Joshua Fox, Unicorn Solutions

Keywords: Application Architecture, Business Model, Data Management, ECommerce, Enterprise Applications, Enterprise Content Management, Information Management, Information Model, Integration, Knowledge Management, Markup, Metadata, Ontology, Repository, Schema, Semantic Model, Semantic Web, Semantics, UDDI, W3C XML Schema, Web Services, XML Schemas, XML, XSD


Enterprise metadata appears in many languages and formats. XML provides a standard and consistent language for metadata, simplifying both interchange and parsing. But simply storing metadata as an XML file (be it XSD, BPEL, WSDL, J2EE EJB descriptors files, or any of dozens of proprietary formats) does not automatically and formally capture the full richness of the given metadata language. Even if XSDs are used to constrain syntax, they cannot define all possible structures and relationships, nor can they express the meaning of metadata in its business context.

The XML Metadata Interchange standard (XMI) was developed by the Object Management Group (OMG) to answer the need for exchanging and storing metadata in a variety of different metadata languages. In XMI, the metamodel of the given language is described according to the Meta-Object Facility (MOF), a metamodeling standard closely based on Unified Modeling Language (UML). Each metamodel is codified in a special XSD or DTD which shows the structure that a given metadata artifact can take. For example, the relational database metamodel include tables, columns and foreign keys, while a business process metamodel includes inputs, outputs, events, and conditions.

Using XMI metamodels, development and runtime tools share metadata in the form of complex object models; these models can then be stored in a universal metadata repository that is open to any new metadata format that already exists or that may come into existence in the future. Indeed, the open-source Eclipse project has successfully integrated tools for Java, EJB, relational databases, and other formats through the medium of XMI. Many other UML modeling tools have also adopted XMI as their interchange format.

But XMI goes beyond metadata interchange and storage to achieve the business goals of enterprise architecture, by enabling the metamodel-driven automation of metadata processing. With an XMI metadata system based on flexibly-defined metamodels, all enterprise metadata is understood in the same way across an organization and across industries. The formally-captured metadata lends itself to having its semantics captured as well, by mapping to a central ontology model, which encodes the organization's real-life business concepts.

XMI and semantic mappings can be used to comprehensively support enterprise architecture in the use of business processes. With XMI and semantic mapping, business processes across an organization's many departments and business lines can reuse and share metadata. For example, if IT managers want to decommission a legacy data store, they can turn to their universal metadata repository to discover the interactions of the data store with other IT assets. In the course of this process, they may find that a legacy COBOL application draws data from the data store, with a J2EE application in turn relying on the COBOL application. By reference to the semantic mappings, the managers then use the repository to precisely determine the business meaning of the data. This gives IT managers the information they need to decide whether to decommission or not, and if so, what other data stores can provide the relevant data.

XML and XSD provide standards for structuring data and syntax respectively. XMI goes beyond them in allowing the interchange of metadata in any of the many existing or future metadata languages. With XMI-based metadata in place, semantics can be added through ontological mappings, rounding out the picture of automated Enterprise Metadata Management.

This presentation will outline the standards and architecture for Enterprise Metadata Management based on XMI, illustrating with case studies from enterprise deployments.

Copyright: © 2005 Unicorn Solutions, Inc.