Archives of the TeradataForum

Message Posted: Wed, 29 Jun 2005 @ 14:50:50 GMT

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

thanks Srini. I dont have any stats yet because I am in the design phase.

Is it an industry standard to use join indexes on large related tables? the distribution and collision cannot be calculated without actually loading the data. But based on what I know about the data, the collision will be more than 10. The singularity of the indexed columns on the child table cannot be gauranteed to be less than 100 as suggested in the manuals. Hence, there is not gaurantee that NUPI might work.

Is it a standard practice to determine hash collision values after designing the physical model and then customize/finalize the PI structure by monitoring the real time performance?

swapnil

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