The leaf level of the non-clustered index would point, not to the heap table, but to the clustered index. The concepts still hold. You would make a covering index if your clustered did not satisfy your query. Regarding Heap Tables. Personally, I only put a clustered index on a table if there is a reason to, not just because there is a mandate. I prefer to have clustered indexes unique, although with 2005+ there is a uniquifier column just in case your index is not unique by key. If there are primary/foreign key relationships, then you have a clustered index.