Very Large Dimensions - Consolidations

670648633

I am trying to create a total of all of the elements in a very very large dimension. It has over 1.5 Billion elements. It builds electronically but when I try and put a simple consolidation in to get a total for the dimension it won't build. Is this a know problem or can it just not be done. I have already tried building the dimension first then adding an update for the total, I have tried building the dimension with out attributes and still trying to build one consolidation but the server will crash each time.

We have a large development box on 64 Bit. Any Ideas are welcome!

thanks

Rose

hezi15826738412

You probably already tried this, but couldn't you establish the consolidation when you are automatically building the dimension?

1)Create the parent
2)Create each element (1.5B)
3)Assign each element to consolidate into parent.

mmkjzz

rmelillo wrote:It builds electronically but when I try and put a simple consolidation in to get a total for the dimension it won't build.

Building with what. A TI process, dimension editor, or dimension worksheet?

sf656

tomok wrote:[quote]rmelillo wrote:It builds electronically but when I try and put a simple consolidation in to get a total for the dimension it won't build.

Building with what. A TI process, dimension editor, or dimension worksheet?
[/quote]
Presumably via TI, I think that goes without saying as I can't imagine it would be possible in the dimension editor and it categorically isn't possible with XDIs.

What in the design is it that requires such a large dimension? Is it required for a cube that is transactional in nature? If so why and what other options have been considdered?  TM1 isn't really built to handle transactional data pure and simple.

The theoretical maximum dimension size is somewhere around 2 billion. I have worked with dimensions of several hundred thousand members and up to 1.5 million which TM1 has handled quite well complete with multiple hierarchies and attributes and very good query response time (just beware of the properties pane in the subset editor.)  I have heard of models with dimensions of up to 10 million members but a dimension of 1.5 billion elements is breaking new ground as far as I'm aware.  Although it may be within the theoretical boundaries it is certainly pushing the boundaries.  Are you doing a theoretical exercise or attempting to build a ganuine application? I would recommend you get yourself some expert advice in how to design your application and also to try and follow up with IBM technical support as a dimension with 1 billion elements is probably a first.