It's interesting to see someone actually put numbers on the I/O penalty of (even optimized) virtualization. I wonder if this means that the improvements from K

It's interesting to see someone actually put numbers on the I/O penalty of (even optimized) virtualization. I wonder if this means that the improvements from KVM/Nitro just aren't that good after all. I'll look forward to that followup article. (In other news, bare-metal VPS is coming back into fashion. The old is forever new.)