Newbie Clustered Index question - when to NUSE?

2018-10-19 05:30:22

I have a bit of a newbie question, which I feel I should really know having dealt with databases for so long, but...

I know the general advice for indexes is NUSE, but a lot of my tables have a non-unique column that is a not-quite-foreign-key from another table, and will be 90% of the predicate key for any queries. The ID will never be used (because of awesome design going back years).

I've attached an example, with horrible anonymising.

So, my question is, if the ...eID is not Unique, and not Ever-Increasing, would I be better off moving my Clustered Index to a PK that is never used for anything other than identity/index, or would I be better using a combo of my ...eID/PK for a Unique Clustered Index (replacing the uniquifier with my actual pk int)?

As you can see, there has been multiple NCIX made on the one column I'd be proposing, and aggregating them, and leaving some space in my FILLFACTOR sounds reasonable to me, which might lead to higher IO, but I'm unsure