Data Model Best Practices
New data integration content available
You're currently viewing a data integration topic that has been replaced and will eventually be removed from our documentation.
For the latest data integration content, see: Data Integration.
General Best Practices
Whenever possible, name tables after their names in the source system.
Connect tables as you joined them in the transformations.
Only join the case table and the master data for activities to the activity table.
For memory footprint reasons, use tables for data representation with complex logic instead of views.
For performance reasons, only use views of master data tables.
Keep the data model as simple as possible.
We recommend keeping the row number in your tables below 800 million. Table with row number larger that this might take extra time to process, or, in case of high network traffic, might fail to be processed.
Foreign Key Relationships
When connecting tables using foreign keys in a data model, consider:
Tables with a 1:n relationship are correctly joined automatically.
For tables with a 1:1 relationship, the join direction has to be considered for pull functions to work properly. You can only pull from the fact to the dimension table. In the example below, you could only pull from BSEG to LFB1, e.g. PU_COUNT(LFB1, BSEG.LIFNR).
 |