|
Archives of the TeradataForumMessage Posted: Fri, 03 Jun 2005 @ 15:28:11 GMT
Hello again, Collecting statistics on the partitioning column is always a good idea, and could be helpful. However, it does not provide as specific information to the optimizer as collecting stats on partition will in V2R6.1. For example, the optimizer cannot use statistics on the partitioning column to determine whether a partition is actually empty or not, while statistics collected on partition can help here. You may run into no difficulty at all carrying a large number of empty partitions, but it is not the recommended approach because of potential trade-offs. To quote from the PPI Orange Book, page 40: " Consider having only as many date ranges as needed currently plus some for the future. Limiting ranges to currently needed ones helps the Optimizer cost plans better and allows for more efficient primary index access, joins, and aggregations when the partitioning column is not included in the primary index. A good rule-of-thumb may be to define 10% or less of the partitions to be empty partitions for future dates. " Thanks, -Carrie
| ||||||||||||||||||||||||||||||||||||||||||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||
Copyright 2016 - All Rights Reserved | ||||||||||||||||||||||||||||||||||||||||||||||||
Last Modified: 15 Jun 2023 | ||||||||||||||||||||||||||||||||||||||||||||||||