Metadata Architecture options are covered in the Metadata Overview, and gives a good back-ground and supplement this field tip. We are going to talk here about which architecture to use when, and their pros and cons.
The brief back-ground is-
In the past, companies have failed to make an enterprise wide physically integrated repository. This is because of the lack of metadata exchange standards, diversity of technologies and business stamina to undergo the change management. However- Recently, given the XML exchange standards and vendors coming out with new age products, one can restart looking at the possibilities.
In real world one has independent metadata repositories, serving a given community or interest group. All metadata architectures assume that one has to live with this reality. Therefore the architecture options on working on how to integrate these repositories while letting them to function for their interest group.
One broad assumption is that Metadata repositories are the subject of relevance more to the large enterprises (say global 20000), as that’s where the cost and effort of metadata repositories will have a business case.
Following are the options, their pros & cons and the scenarios in which they should be used:
Bridged or federated approach:
In this case one creates point to point solution between the metadata repository, (using primarily XML standards) or create a hub, which acts as a routing point. Assuming that you have different technologies used in the independent repositories, this bridging approach will need some level of manual effort to take care of this diversity. Different technologies have different semantic model and capabilities (some may have versioning and some may not have versioning).
This kind of bridging approach should be used by the organizations, which have a limited enterprise level metadata integration needs for integrating these repositories through bridges. One way to minimize the risk linked to the bridging is to enforce a common technology base for metadata repositories as they are created by different interest groups.
Physically integrated Metadata warehouse:
This essentially means that you create a Metadata warehouse (like you create data warehouse to integrate data). This will mean a truly enterprise level initiative and implementing all the consistencies, standards and processes that go with it.
The metadata warehouse technologies expensive, and will need significant management commitment (as it is not only an IT issue). You will need to set-up a dedicated organization under data steward to drive this initiative. The recommendations for successful implementation are:
- The organizations with experience: This approach to be tried by companies, which had a good experience in implementing independent repositories and some level bridging success.
- Go for warehousing in a phased approach: Just like data warehouse, don’t go big-bang. Identify a logical grouping of repositories and let it be integrated and mature.
|