Re: 2nd round of xen benchmarking

[Manuel Bouyer <bouyer%antioche.eu.org@localhost>]

On Wed, Aug 21, 2024 at 04:24:27PM +0200, Manuel Bouyer wrote:
> I've run a similar benchmarck:
> /usr/bin/time sh -c "make; make clean" in pkgsrc/devel/m4, with
> WRKOBJDIR in tmpfs make MAKE_JOBS=1
> This is on a Xeon E5-2637, which does have EPT, the domU has 8 vcpus and
> 8GB RAM.
> 
> PV (netbsd-XEN3_DOMU.gz from the NetBSD 10.0 release)
>       155.66 real        64.30 user        91.96 sys
>       153.36 real        63.57 user        92.71 sys
>       153.79 real        64.68 user        92.20 sys
>       153.77 real        64.89 user        91.99 sys
> 
> PVH (netbsd-GENERIC.gz from the NetBSD 10.0 release)
>        64.08 real        38.36 user        19.98 sys
>        61.28 real        38.80 user        20.05 sys
>        61.31 real        38.79 user        20.07 sys
>        61.38 real        39.08 user        19.83 sys
> 
> As expected, with EPT PVH gives a good performance boost over PV.
> I can't test bare metal as this is a production system running several VMs.

Here's some more results, from an unused machine this time.
Hardware is a Xeon X5650, which has VT-X and EPT.

Bare metal (netbsd-GENERIC.gz from the NetBSD 10.0 release)
       63.03 real        42.56 user        15.40 sys
       60.31 real        43.01 user        15.14 sys
       60.36 real        42.63 user        15.60 sys
       60.37 real        43.12 user        15.10 sys

dom0 PV (netbsd-XEN3_DOM0.gz from the NetBSD 10.0 release)
      366.66 real        58.14 user       320.63 sys
      347.21 real        58.67 user       306.10 sys
      350.79 real        57.97 user       311.53 sys
      347.43 real        59.33 user       306.22 sys

domU PV (netbsd-XEN3_DOMU.gz from the NetBSD 10.0 release)
      204.93 real        81.24 user       126.66 sys
      202.91 real        81.98 user       125.97 sys
      203.11 real        81.08 user       127.13 sys
      203.21 real        80.95 user       127.36 sys

domU PVH (netbsd-GENERIC.gz from the NetBSD 10.0 release)
       85.40 real        51.07 user        25.71 sys
       80.12 real        50.95 user        26.07 sys
       80.47 real        51.78 user        25.58 sys
       80.53 real        51.77 user        25.62 sys

I've yet to find why PV domU is faster than dom0.

So definitively, you want a system with EPT to run virtual machines these
days, unless you don't care about performances. PV was good when i386 was
the norm, but for 64bit systems the performance hit is important.
PVH is good on hardware that can run it efficiently (i.e. at last with EPT
support). A PVH dom0 is probably something to consider, on hardware where it
can run (AFAIK Xen still considers this as experimental).

-- 
Manuel Bouyer <bouyer%antioche.eu.org@localhost>
     NetBSD: 26 ans d'experience feront toujours la difference
--