ESXi and Xen white box

[Shankar Viswanathan]

On Mon, Oct 26, 2009 at 8:48 PM,  <jay-R5TnC2l8y5lBDgjK7y7TUQ at public.gmane.org> wrote:
> Well I've been studying up and it does seem the phenom II x4 is the better option for a vm server.  The athlon II x4 is the same cpu minus the shared cache. But shared cache is good for vms since they generally don't stick to a single core for long.  Thus if the data is cached there is only a 1 in 4 chance of the proces being able to access it.  L3 cache is shared across all processors.  Unfortunately amd also uses an exclusive cache, and I have not found any info on if this has been optimized some how with amd-v.

I'm not sure I understand what you mean by "exclusive cache", but as
you mention the L3 cache is shared across all the 4 cores on the
Phenom x4 and Phenom II x4. You are also correct that a shared cache
is useful for VMs if the hypervisor doesn't employ some sort of core
affinity for each guest.

But to really say if the shared L3 will provide a big performance
benefit, I'd have to know what kinds of applications will be running
inside each guest instance. If those applications are memory bound and
the working set can fit in the L3, then you'll see a measurable
performance difference.

-Shankar
(* disclaimer: I work for AMD)