Poor performance with FreeBSD 10.1 under Xen 4.2

[Andrew Daugherity]

> On Mar 28, 2015, at 8:16 AM, Roger Pau Monné <roger.pau at citrix.com> wrote:
> 
> Hello,
> 
> Thanks for the detailed description of the issue. I've reproduced your
> benchmarks and with the hardware I've used (Core i3-5010U) I'm not able
> to see this performance issue, below are the figures in my case:
> [...]
> 
> I'm Ccing feld because IIRC he found something similar on one of his
> boxes, that also had VTx but no EPT (just like yours). Would it be
> possible for you to try the same set of tests on a different hardware?

I think you're on to something.  I copied this FreeBSD 10.1 VM to a system running the same version of Xen (and same SLES in the Dom0), but with an Opteron 2360SE CPU (which has both SVM and NPT), and it is *much* faster (and feels more responsive too):

Forked, executed and destroyed 5000 processes in 5.216613 seconds.
[Subsequent runs are consistently between 5.1 and 5.5 seconds.]

sssd configure: 22.210u 8.751s 0:30.39 101.8%	4915+236k 0+300io 31pf+0w
sssd build:	176.556u 78.846s 2:15.77 188.1%	5534+217k 49+44io 8pf+0w

I'll run some more benchmarks, but I don't recall Linux VMs being noticeably different in speed between the Opteron and the Xeon systems -- I'm pretty sure it's not a case of a raw CPU power advantage.

> Also, if even FreeBSD 10.1 compiled without XENHVM shows this issue it
> means there's something in the generic code that doesn't work well when
> running virtualized on this specific hardware, but I'm afraid figuring
> it out is not trivial. One place to start would be asking on
> freebsd-hackers and freebsd-virt.

I suppose this performance delta with presence of EPT/NPT vs. lack thereof means it's time to take it to those lists?  My next step will be to test 10.1 under KVM on the Xeon to confirm whether it's a Xen issue or strictly EPT.

Thanks for the tips!


-Andrew