Can I map different levels of a hierarchy to different sources

Question

Hi,
I want for AWM not to aggregate the upper levels of a dimension by itself. Instead, i want to feed the data for upper levels also through different sources.
Its becauese, the measure values of my dimension can not be aggregated from lower level. The value has to be calculated separetely.
I know that the AWM will not aggregate when i specify the dimension agg rule as 'NON ADDATIVE' or something like that. I did it and i was done too. There were the data only for base level for that particular dimension. The other levels were empty.
Now i want to feed the data for other empty levels from different source.
Does anyone knows if it is possible in AWM 10.2.0.1 or can i do it from any other way?
If one knows how to do it, i would be very much grateful to know it clearly.
with regards,
subash

Answer 1

The following method assumes that you are assuming responsibility for managing all of the aggregate level data on your own:
1. Define the cube as (a) uncompressed (the aggregate space of a compressed cube is always managed by the multidimensional engine), (b) with aggregation methods of SUM for all dimensions and (c) with only the detail levels selected on the 'Summarize To' tab of the cube definition. This will set up the cube so that it will not aggregate data when it is loaded.
You will have a formula used to represent the measure and it will look something like this:
DEFINE SALES_CUBE_SALES FORMULA DECIMAL <TIME CHANNEL CUSTOMER PRODUCT>
EQ aggregate(this_aw!SALES_CUBE_SALES_STORED using this_aw!OBJ1819687276)
You can see from the AGGREGATE command in the formula that it is set up to aggregate data at runtime if necessary. The formula reads from a variable that will look something like this:
DEFINE SALES_CUBE_SALES_STORED VARIABLE DECIMAL WITH AGGCOUNT <SALES_CUBE_PARTITION_TEMPLATE <TIME CHANNEL CUSTOMER PRODUCT>>
2. Load the data. You can load the base data through AWM if you wish or you can load it into the variable of the measure using OLAP DML code. You cannot map tables to summary levels using AWM, so you will need to use OLAP DML code to do this. An example program follows.
DEFINE LOAD_SALES PROGRAM INTEGER
PROGRAM
vrb _errortext text
trap on HADERROR noprint
sql declare c1 cursor for -
select -
to_char(MONTH_ID), -
to_char(ITEM_ID), -
to_char(SHIP_TO_ID), -
to_char(CHANNEL_ID), -
SALES -
from GLOBAL.SALES_FACT -
where -
(MONTH_ID IS NOT NULL) and -
(ITEM_ID IS NOT NULL) and -
(SHIP_TO_ID IS NOT NULL) and -
(CHANNEL_ID IS NOT NULL) -
order by -
CHANNEL_ID, -
SHIP_TO_ID, -
ITEM_ID, -
MONTH_ID
sql open c1
if sqlcode ne 0
then do
_errortext =  SQLERRM
goto HADERROR
doend
sql import c1 into -
:MATCHSKIPERR TIME_MONTH -
:MATCHSKIPERR PRODUCT -
:MATCHSKIPERR CUSTOMER -
:MATCHSKIPERR CHANNEL -
:UNITS_CUBE_PRT_TOPVAR(UNITS_CUBE_PRT_MEASDIM 'SALES')
if sqlcode lt 0
then do
_errortext =  SQLERRM
goto HADERROR
doend
sql close c1
sql cleanup
return 0
HADERROR:
trap on NO_ERROR noprint
sql close c1
NO_ERROR:
trap off
sql cleanup
END
3. Change the formula of the measure so that it does not aggregate using OLAP DML commands. E.g.,
consider sales_cube_sales
eq SALES_CUBE_SALES_STORED
The formula will then simply get data from the variable rather than call the aggregate system.
Since you are writing to the physical implementation of the AW you will need to check this code each time you upgrade the database because the physical implementation changes from time to time. (Usually major releases such as 10.1 and 10.2, not usually maintenance releases, but you never know.)