Excellent article. This coupled with Brad's article is very helpful in design considerations. I was especially looking for the size of the row pointer. It's quite a consideration for the choice of a clustered key. For example If I don't have a clustered index I get 8 byte pointers. If I used an INTEGER row ID (auto number) I get 4 byte pointers. If I use a US phone number as my clustered key I get 10 byte pointers. (I use data compression techniques).

You said:

But that is not so true for those of us working on the Mobile. Every byte is precious.

I'm looking forward to the other installments.