with 5 cubes and a plan to design a ODSes, what is the best approach to design ODSes? i.e. with the datasources, including R/3 and flat files, any recommendation for the best design?
In other words, knowing dimensions, theier characteristics and key figures for the fact table, how do you approach ODS designs;
I also plan to use multiproviders to unite some of the cubes? With respect to the multiproviders, what are the determining factors to choose particular cubes to unite.
When you plan to use ODS there must be a valid reason why you want to use them in the system. Like you are using it for detailed records, delta enabling or some other reason.
As ODS are flat file structures and hence not useful for analytical reporting. So would suggest use it only when there is a valid reason for the same. Moreover the key fields in the ODS identify the record uniquely and it calls for quite a bit of effort determining and testing the same depending on the functional area.
You can always have Infocube ODS combinations wherein you have summary data in Infocube and jump to a detail report on the ODS using RRI, and could model your system accordingly.
Hope it helps.
1. Basically it all depends on your business and reporting needs. Typically an ODS would have more detailed data hence more no. of fields than a Cube. If you are going o load 1 cube from 1 ODS typically the ODS will have all the fields(equal or more) as that of your cube. You can refer to the best practices docs in SDN.
2. A multiprovider uses a union operation to fetch the data from different data targets.Here, the system constructs the union set of the data sets involved. In other words, all values of these data sets are combined.
Refer this link: