Page 1 of 1

Tricky Feeders

Posted: Mon Aug 16, 2010 3:43 pm
by tryinghard
Hello,

I need some help on the feeders:

Source - Cube 1 has 5 dimesions - A,B,C,D,E

Intermediate Cube 2 has 6 dimensions - (As cube 1) B,C,D New (F,G,H)

Target cube 3 - this is final cube in which the dimensions are same as cube 2. In this cube I have calucation that takes data from cube 1 Total of dimension A and one element of E, rest of the diemsions are common.

Basically it is an allocation where Cube 1 has data, cube 2 has allocation profiles and cube 3 is the final cube.

I am trying to create feeder in cube 1 that feeds cube 3; however the problem is that there is no dimnsion F,G and H in Cube 1 and I am targeting leafs of dimsions F,g and H and not one element so how do I write the feeder...

The calculation in Cube 3 is woring fine but as expected very slow since it does not have the feeder.

cheers..

Re: Tricky Feeders

Posted: Mon Aug 16, 2010 4:13 pm
by mce
Assuming that all leaf level elements in dimensions F,G, and H of Cube3 are dependant on the data in Cube1, you can feed the top parents in dimensions F,G, and H of cube 3 from cube 1, which will mean feeding all leafs in those dimensions.

Re: Tricky Feeders

Posted: Mon Aug 16, 2010 6:31 pm
by tryinghard
mce wrote:Assuming that all leaf level elements in dimensions F,G, and H of Cube3 are dependant on the data in Cube1, you can feed the top parents in dimensions F,G, and H of cube 3 from cube 1, which will mean feeding all leafs in those dimensions.
Thanks Man, it does work, but it is only slightly better than No Skipcheck, with skipcheck it takes about 1.5 minutes to build the view and without it takes 2.5 minutes. I think this Feeder is making the Skipcheck almost redundant since it is feeding every cell, including five levels in the hierarchies of dimension F, or does Feeder apply to the leafs only when targeting the top level?

Re: Tricky Feeders

Posted: Mon Aug 16, 2010 10:47 pm
by David Usherwood
To do allocations you are inevitably changing the shape of the data, splitting out (eg) department costs over customer.
Allocations boil down to
['Output'] = n:
['Input','High Level'] *
['Driver'] \
['Driver','Total']
You have the choice of using either the input or the driver for the feeder. I suggest you try both and see which is lighter. Make sure you have the cubes properly sequenced as well.
It's likely that the feeder will be a fair size which ever route is lighter. That's life in Allocation Alley - get used to it.