Archives of the TeradataForum

Message Posted: Mon, 27 Oct 2003 @ 21:16:24 GMT

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

In your situation, it is very unusual to have an SI on the 'small' table (Table B). If you want to use an SI it would be better to have it on the large table, i.e., Table A. I don't think the SI on Table B can help at all with any type of join. In fact it might rather confuse the optimizer, which is what seems to be happening in this case.

What happens if you drop the SI on table B and rerun the query? What is the run time of the query?

If you want to speed it up you might opt to have an SI on Table A, but please allow for the extra space for it. If the performance gain for your query in the presence of the 'large' SI is not significant you probably don't need any SI at all. Just make sure you have fresh statistics on both tables for the columns you join on. In some rare (bug) cases the query will run faster without any statistics at all.

Regards,

Victor

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