If I wanted to run 100s (maybe even 200) Windows 7 VMs on a 64 core AMD machine CentOS 6.x machine, but only say 40 or 50 were actually being used at any one time by end-users, would the Centos OS and KVM be able to automatically handle that kind of load? Specifically, would Centos KVM be “smart enough” or capable enough to realized that 100s of VMs at any time are not actually being used, so its memory can be paged and CPU should drop to essentially zero; or, do I have to manually shut down those VMs or hibernate them until they’re needed? I’m hoping that since the 200 or so VMs essentially are based off of only 5 or so “original” VMs, the Linux OS (Centos) and KVM would automatically handle the load by paging / sharing memory load and so forth of those VMs that are not being actively used. I’m furhter assuming that swapping the memory out of paged disk is a “one-time” effort and that after that occurss, a VM would perform “normally”. For the purpose of this question, you can assume that there’s enough RAM to handle 50 VMs running concurrently, but there’s definitely not enough RAM to handle all 200 VMs running concurrently. Any advice/discussion/thoughts/brainstorming is appreciated.
Hmm, The CentOS KVM can be smart enough, but even though they’re not actively in use, Windows clients are always active…
Did you ever put this scheme into operation? How did it go?