January 24, 2010 at 1:40 pm
A modeling question that I was looking for a few opinions on:
We are enhancing our data warehouse with a few new fact tables where the majority of the dimension tables are already in place and in use. After we got done modeling one of the fact tables (that will eventually hold very large amounts of data) we ended up with 4 measures and 15 foreign keys.
My question is if there is such a thing as having too many foreign keys on a table where it would cause performance issues. I know a case could be made on the loading side (with the # of potential indexes) but our concern is more with the subsequent performance of queries via Analysis Services. Of course in this case all of the connected dimension tables are needed, so there will need to be concessions made somewhere if we need to change what our model currently looks like.
Thanks!
January 24, 2010 at 8:43 pm
FK's should not be a problem for applications that read data (Analysis Services only reads data for most of its activities). Indexes on the FK columns on your fact table would be useful.
As you already are aware of - FK's can be a concern with your data import. You can temporarily disable the FK's during the load to help with that.
Viewing 2 posts - 1 through 2 (of 2 total)
You must be logged in to reply to this topic. Login to reply