Excellent. I've been preaching this [VM page faults kill performance and are all to often ignored] for years, though not as eloquently. Years ago, I had a very heated argument with a PhD on our staff who refused to understand this. No matter how many real world numbers I showed him, he kept insisting on using the theoretical "Knuth" numbers (I wasn't the only one he plagued with his theory crap, just the first. It eventually got him fired.)