Five issues about OLAP making me confused.

Question

1. Why do we need to map level twice in both dimension and cube? I know when mapping in cube, we only map the lowest level of
a hierarchy. And what if a same level member has been mapped with different column from source table?
2. There could define multiple hierarchies for a dimension, then why do we need to set one of them as default hierarchy? In
what scenario we need to define multiple hierarchies?
3. When creating a Cube, there is an option in Aggregation->Precompute->Cost_base Aggregation- >Percentage, what does
"Percentage" used for? For performance?
4. When I do some operation in AWM, sometimes it comes up with some errors. But it is really harsh to understand what it is
meaning. Like the error message below,
The transaction is not committable: "An error has occurred on the server
Error class: Express Failure
Server error descriptions:
DPR: Unable to create server cursor, Generic at TxsOqDefinitionManager::generic<CommitRoot>
INI: XOQ-01600: OLAP DML Error "Analytic workspace object OLAPTRAIN.SALESTRACK!
SALES_CUBE_PRODUCT_ALERT does not exist." while executing DML "SYS.AWXML!R11_MANAGE_CUBE
('SALES_CUBE.CUBE' 'DELETE')", Generic at TxsOqStdFormCommand::execute
at oracle.olapi.transaction.BaseTransaction.commit(Unknown Source)
at oracle.olapi.transaction.BaseTransactionProvider.commitCurrentTransaction(Unknown Source)
at oracle.olap.awm.dataobject.DatabaseDO.commitOLAPI(Unknown Source)
at oracle.olap.awm.dataobject.aw.WorkspaceDO.commitOLAPI(Unknown Source)
at oracle.olap.awm.dataobject.olapi.UModelDO.commitOLAPI(Unknown Source)
at oracle.olap.awm.dataobject.olapi.UModelDO.delete(Unknown Source)
at oracle.olap.awm.dataobject.olapi.UCubeDO.delete(Unknown Source)
at oracle.olap.awm.navigator.node.DeleteThread.run(Unknown Source)
When it said "Analytic workspace object OLAPTRAIN.SALESTRACK!SALES_CUBE_PRODUCT_ALERT does not exist.” I can't even find out
what SALES_CUBE_PRODUCT_ALERT is and where I can find the more detail information for me to solve the problem. I mean is here
a manual or reference Oracle provided for helping developer?
5. Does it matter the order in Join Condition when we mapping a cube to resouce, like a=b or b=a? which kind of Join is it, LEFT JOIN, RIGHT JOIN, OUTER JOIN or INNER JOIN?
Thanks,
Satine

Answer 1

Satine wrote:
Hey Shankar,
Thanks a lot for your help.
I was out of office until this week, so I have to read your reply now. I think your explanation is very clear, but there is something I am not sure about.
1.
Shankar S. wrote:
However this is quite similar to loading DAY level in Cube and marking DAY and MONTH levels as stored levels, only relational side does the additional work of precalculating MONTH level summaries for loading into MONTH level of Cube. Quarters get calculated on basis of MONTH level info.First, I have to apologise for my bad english cause I don't understand it very well. Could I say any calculation for the data, which will be loaded to cube, is done on the source data side? Like from your example,Yes, calculations are required on the source side also... specifically the calculation over day level information to get Max at Month level information should be done on the source (relational) side.
Relational:
TBL1: Fact table containing DAY level data
TBL2: Fact table containing MONTH level data. This can be an Materialized View with the requisite logic built into the MV definition or this can be a normal summary/aggregate table which is loaded via ETL e.g. a pl/sql procedure.
Cube:
Dimension TIME has levels DAY, MONTH, QUARTER and YEAR. Use Implementation Details tab option of creating dimension members using keys from data source (dont generate surrogate keys in the AW).
Aggregation: Sum along TIME dimension.
In Mapping Editor for Cube: use option: Other (not star or snowflake)
Use both tables TBL1 and TBL2 in the mapping editor and associate/map DAY level with TBL1 and associate/map MONTH level with TBL2.
Load each day's Daily Total Withdrawal for each Customer/Customer Account at DAY level.
Load the Maximum (MAX) amongst the day level Total Withdrawals at MONTH level.Does it mean the calculation of "Total Withdrawal for each Customer/Customer Account at DAY level" and "Maximum (MAX) amongst the day level Total Withdrawals" both done on source data side before loading to cube? If it does, it means we have to balance the granularity and loading time while we mapping the cube, right?Shankar: Yes, its not optimal but the requirement/business case is not typical either. We'll be executing two loads (from 2 fact tables) to load a single cube.
2.
Shankar S. wrote:
Cube has aggregation set as SUM along TIME, SUM along CUSTOMER but by loading DAY level info into DAY and MAX(day) level info into MONTH, we achieve something like below which is not possible to do out of the box.
DAY level info loaded as is, MAX aggregation operator at MONTH level, SUM aggregation operator at QUARTER, YEAR levels of TIME
SUM along CUSTOMER from Customer Account upwards.After we deifined the MONTH level as "Maximum (MAX) amongst the day level Total Withdrawals", does it mean we can't do SUM aggregation operator at MONTH level just like we do on QUARTER or YEAR?As mentioned above, we dont define Max anywhere within the cube (structure or definition). We define Sum along Time dimension for the Cube but due to the fact load of data at 2 levels using the max operator, the cube will not perform sum of day level to get data for month level but display the loaded data values at month level "straight off" from the cube. Quarter and Year levels would get calculated as SUM(month) dynamically (if not stored levels). If they are stored levels, then the post load cube aggregation process will calculate and store them.
NOTE: Post the load (and internal cube aggregation process) the cube appears to be fully solved... all the data is available at all the levels and we just read them off the cube from the appropriate dimension levels using the relevant access mechanism (sql, olap dml, olap api etc.)
I hope I could express my confusion clearly :).
Thanks again for your help!
Satine