Composite Key vs. Surrogate Key

[NorbertKrupa]

I did a test - and its a fact and nothing else, I do not write my opinion. Its your point doesn't prove nothing - because its an opinion.

Could you kindly share these results? Specifically timings, data set sizes, and execution plans.

[id10t]

Hi!

[DELETED]

[NorbertKrupa]

This isn't a competition. We're trying to understand your argument and are asking how you came to that conclusion (by showing us some empirical results).

I would recommend reading section 5.2 of materialization strategies for column-oriented DBMS and materialization strategies in Vertica.

[scutter]

I haven’t done a performance analysis of composite keys vs surrogate keys. But I have seen a case where a customer didn’t have a primary key in a table so was using all columns in the join to identify differences between record sets. So there were probably ~20 columns used in the join, using the <=> operator as well. This join was horrifically slow, even as a merge join.

—Sharon