TM1 Forum

Hi,

Would just like to check with the community on the best practice for using Feeders. As I understand it, Feeders are used to optimise cube performance. You have to be aware not to underfeed feeders to ensure that the consolidated value is correct. Over feeding can greatly affect the performance of the cube.

However, if my consolidated levels have a value that is rule derived, do I still need feeders?
e.g. If I am forcing the cube not to consolidate (by putting 0 for all consolidated levels), can I just add SkipCheck; to optimise the performance of the cube?

[Edit: There are N level elements that are rule derived but we do not need these cells to consolidate, we don't want any cells to consolidate]

Regards,

Magnus

Seems quite an unusual setup.
The aspect you haven't mentioned is zero suppression. Unfed values disappear in zero suppressed views. If you have a unfed, ruled cell at n level then I don't believe the parents will show up in a zero suppressed view.
I'll let Lotsaram post here with his (to me) 'unusual' approach to having unfed cells as artificial parents - it's not worked in my models, but that doesn't mean it is never appropriate.
And, by the by...
If you are suppressing all C levels, why are they there?

Whether you need feeders or not is dependent on what the rule at the consolidation nodes need to work. For example, if your consolidation rule looks like this:

['Measure'] = C:12*5;

then no feeder would be needed because 12 times 5 is not dependent on any other area in the cube to calculate. However, if your rule refers to another node in the cube, at the C level, it's not going to work unless that other C level node has been fed. I suppose it's possible that the first scenario could happen but more than likely your rule is dependent on another C level node, which means not feeding it will result in the rule not working.

mags88 wrote:
Would just like to check with the community on the best practice for using Feeders. As I understand it, Feeders are used to optimise cube performance. You have to be aware not to underfeed feeders to ensure that the consolidated value is correct. Over feeding can greatly affect the performance of the cube.

However, if my consolidated levels have a value that is rule derived, do I still need feeders?
e.g. If I am forcing the cube not to consolidate (by putting 0 for all consolidated levels), can I just add SkipCheck; to optimise the performance of the cube?

[Edit: There are N level elements that are rule derived but we do not need these cells to consolidate, we don't want any cells to consolidate]

Then I would assume that the consolidations are just there for grouping purposes.

Just write your rules at N level, and weight each member of the consolidation as zero in the dimension.

If you are suppressing all C levels, why are they there?

We have dimensions that are shared across different cubes, in some of these cubes consolidations are needed (and hence feeders have been added). However in other cubes that also has the shared dimension, consolidation of the value does not make sense due to the business logic. Also as Alan pointed out, they are there for grouping purposes.

In the cubes where we do not want consolidation, the consolidation value does not rely on rule derieved N level elements. We have hardcoded the value to be 0 and so to clarify, we do not need feeders for this particular cube?

Thanks again for all your help

Magnus

mags88 wrote:In the cubes where we do not want consolidation, the consolidation value does not rely on rule derieved N level elements. We have hardcoded the value to be 0 and so to clarify, we do not need feeders for this particular cube?

You certainly don't need feeders for the C: 0 hardcoded rules! In general if you don't need N rules to consolidate then you don't need feeders for those rules. However there are exceptions to the generalization where you would still need a feeder, 1/ if you have a requirement for rule calculated leaf cells to show up in a zero suppressed view or 2/ if you need to process data while skipping blank values but include calculated values then you would still need to feed the leaf cells.

Thanks again for all your help, I was just checking best practice.

Merry Christmas everyone!

Magnus