Archives of the TeradataForum

Message Posted: Tue, 03 Apr 2001 @ 22:14:55 GMT

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

A couple of points:

"This latter approach would allow us to order the columns in such a way that the most frequently filtered upon columns appear first in the key."

There is no order in the columns used in a Teradata primary index. These are hash distributed, not value ordered.

"Oracle uses data range partitioning to separate data in the fact table based on the partitioning key." You can have value-ordered secondary indexes on 4 byte columns -- Dates are the most often used -- to help with data range queries.

In general, try to choose an index where the columns are almost always included in the query conditions. As long as the distribution is adequate (no more than 2% variation among AMPs), you should be OK.

John

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