[Lxc-users] 10GbE Interface

[Mayur Gowda]

Hi Robert,

Tried few combinations of disabling the netfilter calls and noticed that it
does seem to make a small difference of around 100Mbps, increasing the
thruput to around 2.8-2.9Gbps, but its still far off from the optimum range
for 10Gig interfaces.


Regards
Mayur

On Mon, Jul 11, 2011 at 1:50 AM, Robert Kawecki <thewanderer at gim11.pl>wrote:

> Dnia 2011-07-10, nie o godzinie 19:59 +0100, Mayur Gowda pisze:
> > Hello All,
> >
> >
> > As planned I got 10GbE interfaces(with linux TOE) on my server(long
> > wait...!!) and got two debian containers up with following
> > configuration:
> >
> >
> > >Set macvlan on the 10GbE physical interface
> >        sudo ip link add link eth0 address 00:00:00:ab:00:01 macvlan0
> > type macvlan mode bridge
> >
> >
> > > Setup containers with macvlan
> > lxc.network.type = macvlan
> >     lxc.network.macvlan.mode = bridge
> >     lxc.network.link = eth0
> > The containers can ping eachother but the problem is they are no way near
> the 10gig mark (throughput IPERF < 1Mbps). Am I missing any configuration
> here? my goal is to bridge the containers @ 10gig using macvlan. The veth
> bridging yields 2.8Gbps ,with MTU tweak, but thats the maximum I can get to.
> Need your expertise on pushing the rates !
> >
> > Thanks & Regards
> > Mayur
>
> Hi,
> I'm not able to provide any hints on macvlan, but have you tried
> disabling Netfilter calls on your bridge with veth? The control files
> should be in /proc/sys/net/bridge - turning some of them off might
> improve performance if you haven't done so already, I think.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linuxcontainers.org/pipermail/lxc-users/attachments/20110711/1479c4ac/attachment.html>