In data warehouse concepts, there may be two table types: fact tables and dimension tables.
The dimension table size is relatively small, so we can and must persist the latest version of every record, for lookup purpose. So it's OK to add TTL based on _valid_to.
But for fact tables, the table size is usually huge and most likely the records won't be updated after inserted. So if the TTL is still based on _valid_to, no records would be purged and the table grows infinitely.
So it's suggested to add TTL based on _valid_from, to purge old data for fact tables.
In data warehouse concepts, there may be two table types: fact tables and dimension tables.
The dimension table size is relatively small, so we can and must persist the latest version of every record, for lookup purpose. So it's OK to add TTL based on
_valid_to.But for fact tables, the table size is usually huge and most likely the records won't be updated after inserted. So if the TTL is still based on
_valid_to, no records would be purged and the table grows infinitely.So it's suggested to add TTL based on
_valid_from, to purge old data for fact tables.