Archives of the TeradataForum

Message Posted: Wed, 18 Oct 2000 @ 10:48:32 GMT

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

Hi Jaz,

I would expect the issue to be the fact that the fact table is very small! only 6 million rows! and the optimiser is right to chose that way. What size Teradata system have you got(is it the 4 node 5200?)? To scan 6million records on a reasonable sized system would be hardly anything.

Depends on what how selective the selection criteria you are using for the query against the dimension tables, e.g. how many of the dimension tables you are selecting a value from.

If you have 8 dimension tables 50k plus 7 that are 50 say then the number of rows this would generate would cause more work than the inclusion of the fact table at an earlier stage. The optimiser has assessed the cost and chosen that way. If the Fact table was bigger as is the case most of the time or the dimensions smaller then the access profile would probably be different.

Example 7 dimension tables of 10 rows each - 10 * 10 * 10 * 10 * 10 * 10 * 10 = 1million product join spool tab le result if all rows are selected.

Cheers

John
Senior Principal Consultant,
Relationship Technology Solutions for Financial Services,
NCR Limited

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