|
Archives of the TeradataForumMessage Posted: Mon, 04 Nov 2002 @ 19:16:19 GMT
A lot of good answers so far, as expected. I would attack it in the following order: a) decide if you REALLY need the index, then prove it by EXPLAIN analysis. Also, try the same query without the index (dropping stats should achieve this) and see if you can live with the query performance hit v ETL performance gain. b) check the data demographics of the USI/NUSI column(s). c) consider removing FALLBACK if present. d) use dummy values in place of default USI/NUSI values (e.g. -1 * well_distributed_column). e) consider adding another column to the index e.g. date/store/sku instead of store/sku, for instance. Hope this helps, Paul Johnson.
| ||||||||||||||||||||||||||||||||||||||||||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||
Copyright 2016 - All Rights Reserved | ||||||||||||||||||||||||||||||||||||||||||||||||
Last Modified: 15 Jun 2023 | ||||||||||||||||||||||||||||||||||||||||||||||||