TM1 Forum

We have a Matrix cube for costs allocation, since the source cube is very large and dimensions in allocation matrix cube are many, so the performance is poor with normal feeder setting.

I can set the special flag in allocation matrix cube with rule as below.

---------------------------------------------------------------------------
SKIPCHECK;
['Allocator'] =N: if ( condition statements
, 1
, continue
) ;

['Other Allocated Costs'] =N: ...

FEEDERS;
---------------------------------------------------------------------------

then I wish the 'Allocator' will act as feeder for other allocated costs in this matrix cube.

But I found the 'Allocator' itself is unfed by above statement, is there any way that I can have the 'allocator' to be fed, and to feed other allocated costs in this matrix cube?

There is no better way to feed 'allocator' with other cubes. If I had such way, I can directly feed those costs to be allocated as well.

Regards,
Chris

wang_chris wrote:We have a Matrix cube for costs allocation, since the source cube is very large and dimensions in allocation matrix cube are many, so the performance is poor with normal feeder setting.

I can set the special flag in allocation matrix cube with rule as below.

What "Special Flag" are you referring to?

wang_chris wrote: SKIPCHECK;
['Allocator'] =N: if ( condition statements
, 1
, continue
) ;

['Other Allocated Costs'] =N: ...

FEEDERS;
---------------------------------------------------------------------------

then I wish the 'Allocator' will act as feeder for other allocated costs in this matrix cube.

But I found the 'Allocator' itself is unfed by above statement, is there any way that I can have the 'allocator' to be fed, and to feed other allocated costs in this matrix cube?

You haven't shown the Feeders that you have used to try and feed "Allocator"? Are there any?

No, I want to set the 'allocator' as a feeder to other costs, since normal feeder's performance is poor.

so at the place I wish to feed other costs, I set the 'allocator' as 1, however, those setting by rules itself is not fed. Should I write a script to insert '1' into the cells. I prefer to use rule rather than script to set the 'allocator'.

Regards,
Chris

Here's the deal: At some point you have to have actual values in a cube somewhere with which to feed from. They can be in the cube you are writing the rules in, or some other cube that shares at least some sort of dimensionality with the target cube so writing the feeder will make sense. You can't just make up some values, like you did, via a rule and use that to feed something else, UNLESS THE MADE UP NUMBER HAS ITSELF BEEN FED. Comprende? The reason the feeder performs poorly for allocations is probably because feeding allocations is inherently tricky and usually results in significant overfeeding. If you have some way of thinning out the feeders and using a TI to populate "allocator", then that's probably going to be your best bet. I have used this strategy on multiple occasions with good results as long as you understand the limitations (the TI will need to be run any time the "allocator" values should change).

I've dealt with this in the past by taking the rule driven flags and writing them to a static slice or cube. (The flag rules themselves can't be skipchecked.)
@tomok, I agree that feeding allocations can be tricky - but I also think that doing the allocations themselves within TI typically loses all traceability, which rather destroys the point of them. Also, I would challenge that you have 'overfeeding' if the feeders you write are those necessary to do the job.
I do tend to freeze the results of the allocation into a static cube for performance, thus getting the best of both worlds.