Given that a unique constraint essentially creates a unique non-clustered index on the table, does the unique constraint store the row identifiers of the "non-clustered index" at the leaf level of the b-tree (I'm assuming the unique constraint is in a b-tree?). I've read blogs that suggests that the query optimiser treats unique constraints and unique indexes the same, which suggests to me that the underlying structures should be the same, but is this the case?