Virtualization without extensions?

[Patton Echols]

On 03/17/2011 09:12 AM, Kent Borg wrote:
> Patton Echols wrote:
>> My server is still running 9.10, so it looks like I could build 
>> rebuild qemu from source and enable kqemu, but that defeats my purpose.
>
> I don't understand the details there, but it reminds me of an 
> advantage of compiling your own qemu: it is possible to have more than 
> one version on your computer at the same time. So one can test a new 
> version without disturbing existing VMs. One can migrate VMs one at a 
> time. (Simple shutdown-restart migration in this case, no fancy 
> live-migration.)
>
>> There is a comment at the top of that page that suggests that qemu 
>> would run without either kvm or kqemu -- just that it would run 
>> poorly. Since I am just interested in testing some web apps, poor 
>> performance might not be much of an issue. Any thought on that?
>
> I think performance is orders of magnitude slower to emulate 
> instructions one-by-one without kqemu. Qemu is impressive in that mode 
> (typical Android development is to run Arm code emulated in an x86 
> qemu...it does work...eventually), but there are limits, it is not fast.
>

Interesting stuff .  Thanks for the lesson!

>> I suppose my other alternative would be to do my testing in a VM 
>> under VirtualBox on my desktop.
>
> You get the equivalent of kqemu with Virtualbox. (Also as a module, 
> like kqemu.)
>

Yes, but I have to run it on a machine with a GUI, precluding the 
server.  I'd wanted to use the server because I can allocate more 
resources to the VM with no realistic impact on performance.  But given 
what I need to do, the desktop machine will work ok.

Thanks again for helping me puzzle through this.