It probably didn't come across in my post (I see on re-reading it) but I wasn't trying to shit on people who haven't had that kind of experience. I think VM/con

It probably didn't come across in my post (I see on re-reading it) but I wasn't trying to shit on people who haven't had that kind of experience. I think VM/container performance in modern software ecosystems (and especially on shared hardware, which is typical "in the cloud") is so bad that lots of people just genuinely don't realize what kind of workload the hardware itself is cabable of serving, and so are consistently surprised when a single server—much faster than a typical server circa, say, 1999—is happily serving a not-even-that-large workload. It's not their fault, just a difference in experience and some... very good marketing, I'd say.