Currently in my project I encountered the case where 1 BW system is shared by 2 source systems (R/3 and ECC6) and the master data from the 2 source systems are overlapped (eg. same material no from R/3 and ECC6 have different meaning).
One of the suggestions is to concat master data value with source system ID as prefix. For example:
Material 4711 from ECC6 Source System = AB_4711
Material 4711 from R/3 Source System = CD_4711
I found the following relevant post in previous topic (viewtopic.php?f=16&t=148940&p=1046167#p1046167)
However, I found that some master data having data type NUMC (eg. 0EMPLOYEE) that:chetan wrote:not_yet_a_bw_guru wrote:
Having said that its more of a disign than anything else.
You only use the concatinated object as the key buy you still have the original as an attribute and thats what you use in your queries.
There will also be a little training involved. But it works like a charm
I am speakng of my own experience and with consultation of SAP AG.
cannot name the client but these Instalation i refer to is TOP 5 BW installations in USA,quite huge and extensive cutting all the SAP modules
Hope it Helps
â€¢ We need to use numeric value so that the prefix is consistent to all master data, hence we can no longer use symbol like dash, underscore to separate prefix and master data value. This requires thorough observation to all master data values, for example if we use prefix '99', whether this is safe and doesn't overlap with existing master data that is not prefixed (we might only prefix master data for the newly added R/3 source system, and if there is other new source system added in future; so that the existing master & transaction data that have been live for 2 years using ECC6 source system are not interrupted).
â€¢ Some master data may have been populated with values with maximum length, thus to add the prefix we need to extend the length of InfoObject (eg. 0COMP_CODE is CHAR(4) and normally uses values: 1000, 1100, 1200, etc.). To be able to extend the length we have to delete all records in master data, and consequently we have to delete transaction data in cubes that use master data as well, thus it still impacts existing data.