This is an archived copy of the Xen.org mailing list, which we have preserved to ensure that existing links to archives are not broken. The live archive, which contains the latest emails, can be found at http://lists.xen.org/
Home Products Support Community News


RE: [Xen-users] maximum domU performance

> -----Original Message-----
> I'm not intentionally being thick, but I want to cover all
possibilities :-)

No worries, I can appreciate wanting to know.  Whenever I have the
luxury of time, I like to experiment  with settings I have not tried
before, or to otherwise "push the envelope".  I usually learn something
new, useful or not...

> To extrapolate from that, if I wanted to avoid having dom0 competing
> for a core, I would have to pass through any devices the domU needed
> to access, or have hardware that supports sr-iov?

Yes, I think so.  But I have no experience with this--I've used standard
PV drivers for all my domU hosts.

> Just for the sake of argument, if I could eliminate all hardware
> proxying requirements of the dom0, would a domU be able to hold on to
> all the cores, or would the dom0 or  hypervisor still periodically
> "steal" a core?

Theoretically, if your domU were standalone and didn't rely on dom0 for
drivers or anything else, the dom0 could cease to operate entirely.
Though  I don't know if that's possible or practical with the Xen
software as it is currently released.  (Can you even pause a dom0?
Never tried it...)

As I understand Xen's hypervisor architecture, a dom0 isn't all that
special, it's just more privileged than any other domain.  If a dom0
shuts down, I assume it must take down the hypervisor with it.  But if
the dom0 were just  really, really quiet, there's no reason I can think
of that a domU can't use most or all of the available hardware CPU
resources.  But keep in mind, as long as there is interrupt activity
(e.g. timer0) there is at least some CPU activity on dom0.


Xen-users mailing list