Am Mon, 5 Jul 2021 14:01:07 +0100
schrieb Andrew Cooper <andrew.cooper3@citrix.com>:

> > Unfortunately, I'm not able to prove the reported gain with the systems I have today.
> > I'm waiting for preparation of different hardware, right now I have only a pair of CoyotePass and WilsonCity.
> >
> > I'm sure there were NUMA effects involved. Last years libvirt was unable to properly pin vcpus. If I pin all the involved memory to node#0 there is some jitter in the logged numbers, but no obvious improvement. The fist iteration is slightly faster, but that is it.  
> 
> Oh - so the speedup might not be from reduced data handling?
> 
> Avoiding unnecessary data copies is clearly going to improve things,
> even if it isn't 25%.


For HVM the only notable improvement is the initial iteration.

On average with 4 migrations of a single domU from A to B and back from B to A, transfer rate goes up from ~490MiB/s to ~677MiB/s. The initial transfer time for the 4194299 domU pages:

with plain staging:
36.800582009
32.145531727
31.827540709
33.009956041
34.951513466
33.416769973
32.128985762
33.201786076

with the series applied:
24.266428156
24.632898175
24.112660134
23.603475994
24.418323859
23.841875914
25.087779229
23.493812677


Migration of a PV domU is much faster, but transfer rate for each iteration varies with or without the patches being applied.


Olaf