Design Advice

[kelso]

Hello

Maybe I just need another point of view

I need to take data and elements from a corporate cube which is in another country, This cube need the corporate data and the posibility for users to create new elements and include other hierarchies, also some new rules. For example new versions to be compared with actuals, some new regional prducts etc.

I guess I have five options to accomplish this,

- Replication and then customize a new cube.
- Copy all tm1 files to the server, starting up the cubes and make lots of TI processes to create my customized cubes
- Ask for ASCII files and create new models with TI
- ODBO which creates some Unique code to be able recreate all the hierarchies
- Mix some of this steps

Does somebody replicated a cube into another location? giving local users the change to customize their cube from an original version... I would like a design advice because the problem is that it has to be syncronized...

Not expecting a detailed solution..

Cheers,

[Steve Vincent]

If the central data is in need of being refreshed more than on just the initial build then i can't see how replication will be of much use in your case. It's a 2 way process so anything they do locally will be replicated back to the corporate source, not just from corporate to local. if the corporate view is a bit like a high level, less detailed view of the local then it sounds like an asciioutput / asciiinput solution is best. It is how i'm managing a similar situation here.

[kelso]

thanks..

how do you update the dimensions elements on hierarchies on both schemas?

[Solanna]

How many elements are we talking about here?

Certainly you don't care about the C level elements in the Corporate Cube unless the data from the local cubes is coming back into the Corporate Cube

So focusing on your local cubes... from your first post it seems like your users will have admin authority to manage the dimensions (not advisable)
Using a TI process, simply add any of the new N level elements from the Corporate ASCII file and put them to a C level element called "Undefined Parent" for example
This will give your users visibility into the new elements that have been added during the loading of the ASCII file
Then let your local users figure out how they want them to roll up

As for syncronization, what exactly do you mean? That the same data in the the Corporate Cube is in all the other cubes at a specific time?
Certainly you don't mean reconciliation given the fact that users will have admin rights and can easily add, delete elements and data and that actuals can also be changed
There seems to be a major lack in control of the business processes which really should be your first concern
Not how you will replicate the data...

[kelso]

Thanks,

"Undefined Parent" is exactly what I was thinking to use to solve this, but 5 cubes are planned to be build where this change needs to be done (between 900 and 200 elements max and min) how I can be sure the user will do it right, some elements have negative weights or have two parents, I am not sure this what I would like to use.

By syncronization I meant that both models (corporate and local) have the same data, having the chance the modifyng the local cubes.