Hints on this model: Deciding on Cube/ODS & types of dimensions

Question

Hi,
We have 4 ODSes which are fed data from 4 different flat files:
ODS1—loaded daily-key data-field1, field2, field3, field4
ODS2—loaded twice per week-key data-field1, field2, field3, field4
ODS3—loaded daily-key data-field1, field2, field3, field4, field5
ODS4—loaded twice per month-key data-field1, field2, field3, field4, field5, field6
This is a shipping, delivery, finance reporting scenario.
1. What considerations need to be made with respect to the differences in the key fields as the data are pushed up in the data flow?
2. What considerations need to be made with respect to the differences in the key fields as the data are pushed up in the data flow?
3. If the goal is to summarize the data in the cube so that the details are kept in monthly buckets, can you provide me with some steps on implementing it?
4. What if for performance reasons, we want to keep the summarized data in buckets of months, but we need the capability to drill to the details in the ODSes.
Can you provide me with some steps on implementing it?
Thanks

Answer 1

I picked ODS4 as main ods because the key on this ods represents as key for all the other ODSs.
Example.
ODS1: Key1, Key2, Key3, Key4, data1,data2,data3
ODS2: Key1, Key2, Key3, Key4, Key5, data3,data4,data5,data6
ODS3: Key1, Key2, Key3, Key4, Key5, Key6, data6, data7, data8.
Main ODS (Consolidated ODS)
ODS3: Key1, Key2, Key3, Key4, Key5, Key6, data6, data7, data8.
in your update rule, only map the field you need from each ODSs, the other fields need to be no update if the field is getting populated from other ods. That is what I mean in respective fields from each ODSs.
Yes all the loads should come to the ods and gets updated to the cube, I am not sure if your extractor is supporting delta. You need to think about the reporting requirements, should you report on the daily load if even the bi-weekly, weekly and bi-monthly loads get loaded? But one meaningful record will be created on the consolidated ods (main ods) and depending on how granular your reporting requirement, you could either put the key of the ods in your cube. If you have all the key in your cube, then your cube is very granular and there is no need to do jump-query to the ods.
For creating a dimension, just keep in mind that you have create as many dimension as you could (max 13) instead of creating one big dimensions. Create a relevant char in the same dimension where there is one to one relationships.
please don't forget to say thank you by assigning points.
thanks.
Wond