How to integrate 2 data marts

[rajeshkadiri]

[sud]

Put em in a mixer.

[Ultramundane]

How to integrate 2 data marts

The Monte Carlo method should work well (especially for steep slopes) to establish several random points that you can use to sum the parts.

[chulett]

And here I was going to say simply bus them to the same school. Oh well.

[DSguru2B]

Put em in a mixer.

with strawberries and sugar.

[ray.wurlod]

Define integrate in this case.

Step 1 is planning, usually with the inputs of business analysts and/or subject matter experts.

Step 2 is more planning, taking the results of Step 1 (which determine what is required) and designing the ETL as a set of target-from-source mappings.

Step 3 is the physical set up of the target tables.

Step 4 is acquisition of the metadata associated with those tables, and construction of any other needed design components (Transforms, Routines, Data Elements).

Only then should you contemplate doing any actual ETL development.

[John Smith]

perhaps you could provide more details on what you're trying to achieve and clearly define integrate?
i guess the underlying metadata has to be consistent for any integration work.you can't integrate when both the data marts speak different languages if you know what i mean.

[ray.wurlod]

Wanna bet?

I'm involved right now in a Japanese & English mix.

Yes, I understand what you're saying - that there has to be some measure of compatability - and business agreement on common terms - before data marts (or any other silos of information) can be merged in any meaningful sense.

IBM's Business Glossary is a great way to get the business analysts and subject matter experts to produce their part of the bargain, and the common Repository in version 8 means that is available to database architects (at least if they're using Rational Architect) and to DataStage developers.