In Part 1 of this article we looked at a specific use case (probably exaggerated) that gave us an idea about how CPU and IO performance might be affected by a bad page split. We continue this analysis looking at what really happens to the data on the leaf pages of a clustered index with a single insert into an already filled page (fill factor = 100%).
Warning: the dialogue below is well into the geeky intellect containing both T-SQL code and numbers. Side effects might include a nap or grabbing another BEvERage.
Approach
In this analysis, I used the