Archives of the TeradataForum

Message Posted: Tue, 29 Nov 2005 @ 20:34:04 GMT

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

Please provide the average row size and the DATABLOCKSIZE for the table. 150 row per index value should not cause much of a duplicate row check (NUPI dup problem). Don't forget to account for two bytes on VARCHAR and presence and compression bits in the average row size. The rule of thumb is that the number of row with same index hash should fit into one block without any problems. I have found it to be much higher.

Some other possible problems are secondary or join indexes. I would also avoid updates if possible.

Short on pros and cons on MULTISET.

PROS: no duplicate row checks (this can also be avoided with an unique index)

Allows duplicate rows if desired

CONS: Your load routines have to perform Referential Checks so you don't insert the data twice. If you get data in there more than once this will cause duplicate row counts on queries which can be a nightmare.

Thanks,

Bob Diehl

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