Last up for me today were some folks from IBM on IBM Metadata Strategy – An Information Management Perspective. IBM’s focus is Information on Demand – getting information about of the data management layer and into an integration layer from which it can be delivered as business intelligence and performance management. I, of course, would add that delivering information to automated systems is also critical.
Now IBM has tried lots of approaches to metadata before and is very focused on trying to fix the problem effectively. Metadata is growing in importance as things like SOX create legal risks for companies that can be addressed by better metadata. Three key challenges come up:
- Getting agreement on what metadata actually is!
- Find the right strategy for specific project requirements
- Finding the strategy that works, technology that works and getting people to conform
There is a promise that metadata will fix traceability, impact of change and much more. The reality is that there is a big offline repository that does not work because systems are still little islands and because metadata is not aligned to business needs. As a result IBM has a very specific focus for a unified integration and delivery platform. IBM pulled together its various integration engines as well as loosely coupled UIs for different kinds of users. Ultimately a single integration platform repository is what is needed. This led to IBM Information Server with unified deployment and metadata across the integration lifecycle – including across third party modeling and BI tools relevant to integration. They don’t pretend to have a general purpose repository for enterprise metadata and don’t manage third party metadata explicitly (can consume and link metadata from modeling tools, BI tools etc). Do have a common metadata server for their integration projects – technical (tables, columns, datatypes), business (stewards, categories, terms), project and operational (job execution etc) metadata. Third party data is exchanged by MITI tools (the ODBC of the metadata world)
IBM is focused on a small number of repositories, focused on the various brands like Rational or Websphere as well as a federated approach across brands. Not trying to link everything to everything so much as support business processes that require specific cross-brand /brand-third party queries. They are using common underpinnings like OWL, XPath/XQuery etc.
For instance, can use the integration server and deploy a job as a service. The service registry picks up some of the metadata from the integration server and allows a federated query can bring the service registry and do lineage and impact analysis of a service. Which services impact this database column? Which columns are used in this service?
Clearly metadata is a topic I will be exploring more in the context of decision management for impact analysis considerations, if nothing else.