Archives of the TeradataForum

Message Posted: Mon, 10 Jul 2006 @ 18:32:05 GMT

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

<-- Anonymously Posted: Monday, July 10, 2006 14:25 -->

The Dimensional Model does not change or require special considerations because of the JIs. The JIs enable me to get the data closer to the report w/o touching the fact table.

For example the simple Star schema of CLAIMS would have a fact table named INSURANCE_CLAIM with the dimensions AGENT, DATE, YEAR, MONTH, STATE, TYPE, CATEGORY, etc. For this example lets estimate the number of rows in the fact table at 1300 million for 10 years.

I use the JI's to create aggregates at different levels of domain granularity on the FACT table only!!!!!.

A report that seeks to print loses for each STATE on LAST YEAR MONTH vs THIS YEAR MONTH by TYPE and AGENT would use the join Index with the right granularity and skip the FACT table. That join index would have a lot less rows to scan than the actual fact table (maybe 100 to 150 M rows). I also save a ton of CPU doing SUMS and GROUP Bys too,

Now, if I could only get PPIs on a Join Index I would be a happy man.

All of this done automagically, of course.

Anonym

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