Archives of the TeradataForum

Message Posted: Fri, 26 May 2000 @ 11:16:39 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

I guess that the price for dropping the date from the primary index is to change the table definition from set to multi set table. In this case you avoid the duplicate checks in the loads / Insert/Select and have the better primary index.

It has some risk to do that and I wouldn't do it (if I could avoid it) in tables which are populated via Multiload. You have to be more careful in the population or load process for this table. If you populate the table via insert/select make sure that every insert/select leaves an id for the run which adds the row into the table. The generation of the id and the insert/select have to be covered as one transaction.

On the other hand it is also possible that you drop the date from the primary index if you are running V2R3 with the new hash algorithm. Hash synonyms are handled more efficient in case I correctly understand the documentation.

	<Prev	Next>		<<First	<Prev	Next>	Last>>