Archives of the TeradataForum

Message Posted: Tue, 04 Dec 2007 @ 18:54:22 GMT

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

Kamin:

You have not provided any level of detail with your questions. However, think about this: if it were to use two JI at the same time, it would have to join them. Would it really make any sense to join them when one or the other of them covers the query? I think not.

As for choosing the slower one, do you have statistics collected on the JIs? If they are badly skewed based on the PI you selected and there are no statistics present, when the PE does the AMP sampling it could think that it is the smaller of the two JIs.

When a JI does not cover the query, then it will need to be joined to at least one other table to abtain the requested columns not in the JI. Is this what is happening? If you create a view using the SELECT that created the JI, this cannot happen. If the missing column(s) are used often, maybe they should be included in the JI.

Just my thoughts on your posting.

Michael Larkins
Certified Teradata Master
Certified Teradata SQL Instructor

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