[vbox-dev] [PATCH] [SVN] Please Merge 7 VirtualBox Kernel Modules into the Linux Kernel

[Stephan von Krawczynski]

On Wed, 8 May 2013 22:51:58 +0200
Frank Mehnert <frank.mehnert at oracle.com> wrote:

> On Wednesday 08 May 2013 22:31:37 Stephan von Krawczynski wrote:
> > On Wed, 8 May 2013 22:02:11 +0200
> > 
> > Frank Mehnert <frank.mehnert at oracle.com> wrote:
> > > On Wednesday 08 May 2013 21:50:05 Stephan von Krawczynski wrote:
> > > > On Wed, 8 May 2013 21:11:03 +0200
> > > > 
> > > > Frank Mehnert <frank.mehnert at oracle.com> wrote:
> > > > > On Wednesday 08 May 2013 20:41:40 Stephan von Krawczynski wrote:
> > > > > > Alexey, please ask yourself why noone answers questions about the
> > > > > > networking... We _do_ comparison and I have not seen vbox with
> > > > > > multi-cpu guest-to-guest reaching
> > > > > > 
> > > > > > [  5] local 192.168.4.83 port 5001 connected with 192.168.4.23 port
> > > > > > 49808 [  5]  0.0-10.0 sec  5.47 GBytes  4.69 Gbits/sec
> > > > > > 
> > > > > > (vmware iperf example)
> > > > > > 
> > > > > > ... and I really like vbox.
> > > > > 
> > > > > Please don't give up. :) Bad performance with networking in SMP
> > > > > guests is a known issue and there is work going on to fix several
> > > > > related bugs.
> > > > > 
> > > > > Frank
> > > > 
> > > > Hi Frank, nice to read you.
> > > > 
> > > > I did not read the corresponding code so far, but it sounds reasonable
> > > > to me that using one queue per guest cpu with affinity should let it
> > > > fly.
> > > 
> > > It's a bit more complicated :)
> > > 
> > > Frank
> > 
> > Have you tried? :-)
> 
> No, because we are aware of other problems which we want to fix first.
>
> Frank

Hope you don't waste time with distro-kernels' problems already solved in stock ;-)
Give 3.9 a try.

-- 
Regards,
Stephan