Referencial Integrity in an EDW


Referencial Integrity in an EDW

I'm curious to know how may people who have actually developed a relational data warehouse on a Teradata platform actually enforced RI during their loads? I know there is a huge cost in doing this and am curious how/if it has actually been used in the real world. Anybody have any strong opinions on the use of RI in a Teradata warehouse?



Re: Referencial Integrity in an EDW

I don't have numbers to share, but I can vouch it's very nasty...

so in general I would say, keep soft RI in your master tables (which is good enough for the optimizer).

Use Hard RIs on the staging table (which is dropped before load and then turned on to get bad records)

On second thoughts, it's a lot easier to run a custom SQL in the staging tables to weed out baddies than to use RI ...

I wouldn't go near it if I had an option. :-(