On Tue, 02 Sep 2008 18:24:52 +0100, Orchid XP v8 wrote:

>> Quote, from the wiki page you linked
>> 
>> "VMware and similar virtualization software for the x86 must employ
>> binary translation techniques to trap and virtualize the execution of
>> certain instructions. These techniques incur some performance overhead
>> as compared to a VM running on a natively virtualizable architecture"
> 
> Right. So it only emulates some instructions, instead of all of them?

Not really; at least as I understand it, if you had to emulate the 
instructions, you'd have to recompile.  What VMware and programs like it 
do is trap them and execute them within their 'sandbox'.  It's not really 
the same as emulation, which tends to be more of a translation process.

Of course, if you have a chip that supports VT (for example), then that's 
entirely different.  Or a dual-core processor/MP system, which I think 
VMware also can take advantage of, by using one core for the virtualized 
system and one for the host.  Not sure how they do that, I don't imagine 
it's well documented publicly.

I do know when I played around with virtualization a few years ago, there 
was a need to recompile kernels that were running (one of the issues that 
prevented NetWare from running natively in XEN, for example - NW kernel 
would've had to be rewritten in order to deal with the fact that it is a 
nonpreemptive OS - from what I understand from people who have worked 
closely with the NW kernel developers, that makes virtualization (at 
least as implemented in XEN 2) more difficult to achieve.

Jim

Post a reply to this message