Depending on the selection conditions, different deletion strategies are used for the data volume to be deleted as well as the properties from the database tables.
- If the complete InfoProvider is deleted based on the selection conditions from the archiving run, then the affected tables in the database are deleted with a DDL command and are recreated.
- If the section of records to be deleted has exceeded a specific threshold value (for example, ten percent), then a copy of the table is created with the selection conditions that complement the archiving run. The table is deleted afterwards, and the new table is renamed to the name of the old table.
- If the section of records to be deleted remains below the threshold value, it is selectively deleted in the database table with DML commands (Data Manipulation Language).
- If the fact table for an InfoCube is partitioned, the deletion occurs under specific conditions by deleting the complete partitions (drop partions).
The same strategies apply for InfoCube aggregates as they do for the fact tables with one difference: If you delete aggregates, only the deletion conditions for the characteristics that are also used in the aggregate are shifted. As a result, too much is deleted from the aggregate in some cases. This is corrected when the aggregate is completed by means of aggregating the difference quantity from a more detailed aggregate or, if no such aggregate exists, from the fact table.