[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: [Xen-users] Xen, SMP and Opterons with different stepping


  • To: "Richard Jones" <rich@xxxxxxxxxxx>, xen-users@xxxxxxxxxxxxxxxxxxx
  • From: "Petersson, Mats" <Mats.Petersson@xxxxxxx>
  • Date: Mon, 27 Feb 2006 11:45:00 +0100
  • Delivery-date: Mon, 27 Feb 2006 10:46:35 +0000
  • List-id: Xen user discussion <xen-users.lists.xensource.com>
  • Thread-index: AcY6PEAl3Dmx8eytRkyc5mrV3v6MQgBTIVLA
  • Thread-topic: [Xen-users] Xen, SMP and Opterons with different stepping

> -----Original Message-----
> From: xen-users-bounces@xxxxxxxxxxxxxxxxxxx 
> [mailto:xen-users-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of 
> Richard Jones
> Sent: 25 February 2006 18:48
> To: xen-users@xxxxxxxxxxxxxxxxxxx
> Subject: [Xen-users] Xen, SMP and Opterons with different stepping
> 
> 
> We have a single Opteron server running Xen.  For various 
> reasons we need to add a second CPU.  Although we can get the 
> same model Opteron, it's unlikely that it will have the same stepping.
> 
> Is this going to be a problem, or do we have to buy matched Opterons?
> 
> Can we avoid this by somehow pinning the domains to CPUs so 
> that they will never unexpectedly move to a different stepping CPU?

I recommend that you use the same stepping and speed on both CPU's. Yes,
I know, it's a pain to have it that wat, but it assures that the CPU's
really are doing the same thing when they are supposed to - as a
representative of AMD I can't recommend you do anything else. Chances
are also that the BIOS actually requires this to match... 

As to Xen, if you pin to one CPU, then that guest will run on that CPU.
However, Xen kernel itself will run on both CPU's (potentially
simultaneously), which is where the worry could be - there are
mechanisms in Xen that rely on the processor doing things in a
particular fashion, and although this is guaranteed by the processor
architecture, there's no guarantee that mixes of steppings will not have
subtle changes that means that one CPU doesn't quite understand what the
other one is doing - which would cause VERY interesting effects. 

This said, I think the likelyhood of such changes are actually quite
small - it's just that if you're relying on the processor to do things
in a particular way, it doesn't take much to make things "not work" -
even speeding up/slowing down a particular instruction by a single
clock-cycle could cause things to start going wrong - something that
DOES happen from one revision to another.

--
Mats
> 
> Rich.
> 
> --
> Richard Jones, CTO Merjis Ltd.
> Merjis - web marketing and technology - http://merjis.com 
> Team Notepad - intranets and extranets for business - 
> http://team-notepad.com


_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.