[plug] Windows guest VM's

Brad Campbell brad at fnarfbargle.com
Thu Jan 31 00:46:22 UTC 2013


On 31/01/13 08:04, Kevin Shackleton wrote:
> While it's handy to know that one client Windows OS works better than another on a Linux VM 
> server, what is the technical reason for this difference and does this difference extend to 
> different incarnations of Linux VM clients running in the same environment?  What's the best bang 
> per buck?
>

That's a really good question Kev. I've got a number of VM's here with either Debian and Ubuntu in 
them. A mix of 32 & 64 bit VM's on a 64 bit host.
I don't find a quantifiable performance difference among any of them really. If I were serious about 
finding out I supposed I'd acquire a copy of 64 bit Windows XP and run that up against the 32 bit 
version, and then do a 32 bit install of Win7 against the 64 bit version. I just don't like working 
with Windows enough to want to do that.

My gut feeling is 10 years of OS kernel improvements mean Windows 7 just performs a bit better. 
Maybe it just works better with the hardware qemu virtualises.

For the VM's I leave open all the time (Blackberry BESX, Office 2007) the XP VM uses less resources 
on the host, but for hungry apps like Autocad Win 7 gives a much better user experience.

I'm running all my VM's on KVM with Libvirt. A Debian stable host with qemu from git and libvirt 
from recent sources (I upgrade it about once a year) and kernel from whatever is current 
linux-stable on the rare occasions I reboot.

It is a pretty well specced host though.


More information about the plug mailing list