Re: [Wg-test-framework] Inquiry from SoftIron.co.uk (SoftIron Overdrive 3000)

[Julien Grall <julien.grall@xxxxxxx>]

(Adding Thomas and Steve)

On 09/08/2016 13:12, Lars Kurth wrote:
> Ian, (adding Julien)
Hi Lars,

> On 09/08/2016 11:48, "Ian Jackson" <ian.jackson@xxxxxxxxxxxxx> wrote:
>
> > (Adding the Xen Project Test WG mailing list.)
> >
> > Background for the list: We've been talking to SoftIron about getting
> > a couple of SoftIron Overdrive 3000 1U ARM64 servers for the colo.
> > All the technical details seem good.
> >
> > We like to have machines in (at least) pairs, because that makes it
> > much easier to distinguish hardware failures from host-specific
> > software bugs.  Also, given that these will be our only arm64-capable
> > machines, having a spare would be essential.
> >
> > Lars, do we have another possible vendor in the pipeline too ?  I
> > think an architecture tested in our colo only two hosts is possibly on
> > rather thin ice.  Obviously having two is better than none, and if
> > there is some problem we could degrade back to the situation where we
> > had none, but it would be nice to have more than two.
>
> I was also going to get two Cavium boxes for this release cycle, which
> are also ARM 64. So we need to get moving on this.
>
> See http://b2b.gigabyte.com/products/list.aspx?s=92&p=190&v=1029&ck=102
> Possible candidates are
> - R120-T30 (rev. 100) 1U Cavium ARM Rackmount Server
> - R150-T60 (rev. 100) 1U 10-bay Cavium ARM Rackmount Server
> - R150-T61 (rev. 100) 1U 4-bay Cavium ARM Rackmount Server
> - R270-T61 (rev. 100) 2U 24-bay Cavium ARM Rackmount Server
> - R270-T60 (rev. 100) 2U 12-bay Cavium ARM Rackmount Server
Do we have more details about the core? More specifically the revision
of the cores?

> > This is especially true given that these now boxes are hosts which we
> > might want to do some 32-bit testing on too, even if only of 32-bit
> > guests, so we test 32-on-64.
> >
> > AFAICT from our records we have 6U spare in the rack.  I wouldn't want
> > to use all of it and in any case one of the spaces is inaccessible due
> > to other equipment.  But 3 (if identical) or 4 (if not) ARM64 boxes
> > would be nice.
> That was the plan: I am still waiting for Cavium to confirm which one
> would complement SoftIron best from a testing perspective and which
> US redistributor to use. As far as I understand the key difference
> between those machines on the list above are that some are 1 vs 2
> sockets.
For what is worth, the 2 sockets server would allow us to test NUMA in
osstest when the support will be added in Xen.

> Also, the tests will not work on Cavium servers until we have GICv3,
> which is coming. Julien (CC'ed) should be able to confirm. So we have
> a bit of a chicken and egg problem.
Andre Przywara is working on the ITS support for Xen. I expect to see
the feature in 4.8. However, bringing up Xen on Cavium may require more
work such as implementing workarounds and some platform specific code.

Lars, Thomas, you may want to reach out Cavium to see if we can get the
platform specific code in Xen 4.8.

Regards,
IMPORTANT NOTICE: The contents of this email and any attachments are 
confidential and may also be privileged. If you are not the intended recipient, 
please notify the sender immediately and do not disclose the contents to any 
other person, use it for any purpose, or store or copy the information in any 
medium. Thank you.


_______________________________________________
Wg-test-framework mailing list
Wg-test-framework@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/cgi-bin/mailman/listinfo/wg-test-framework