Huge table meets star schema

Hello everyone,

I have a general question.

I was wondering if it's acceptable to combine a star-schema with one or more huge table(s). Where those huge tabels would be a plain dimension like dim_huge_x and a fact_table would be just a fact_table.

So I was thinking of a design like this:

dim_date -> fact_table
dim_time -> fact_table
dim_person -> fact_table
dim_huge_x -> fact_table
etc...


What are your thoughts on this ?

I don't follow... do you mean to say derive a big flat dimension by denormaling all related dimensions into one table... creating a schema made up of one fact and one dimension table?

I wouldn't do that... performance would be terrible.

ngalemmo wrote:I don't follow... do you mean to say derive a big flat dimension by denormaling all related dimensions into one table... creating a schema made up of one fact and one dimension table?

I wouldn't do that... performance would be terrible.

I mean derive a big flat dimension by denormaling some dimensions, not all. So that way you will have a normal star schema, but with one or two big flat dimensions.

It really depends. If you mean creating a junk dimension for a handful of low cardinality codes and descriptions, fine.