Luis Cazares - Tuesday, February 21, 2017 2:12 PM
Let me second what Luis said. There is an IDENTITY and it makes an excellent clustering key. It's what's known as an artificial key because it has nothing to do with the data. The other, a natural key, identifies the row just by being what it is. Granted, the natural key you pick may not be any good a few years down the road and it'll almost certainly add more than 4 or 8 bytes like Luis said. When you consider that every row of every nonclustered index inherits the bytes of the clustering key, keeping the clustering keys small becomes important. Remember, you have to read them and they have to be in memory to be able to use them.
I speak practical a lot better than I speak theoretical.