[ovs-discuss] Performance of openvswitch versus linux bridging with XEN

[Vijay Chander]

On Fri, Jan 20, 2012 at 5:08 PM, Jesse Gross <jesse at nicira.com> wrote:

> On Fri, Jan 20, 2012 at 4:26 PM, Vijay Chander <vijay.chander at gmail.com>
> wrote:
> > We are using the latest hot fixes with xencenter 6.0 including the one
> > mentioned below.
> >
> > What we noticed was that the openvswitch was running in user mode (for
> the
> > default xencenter 6.0)
> > installation !
>
> Why do you say that?  I don't believe that is true; even when run in
> kernel mode, OVS has both userspace and kernel components.
>

What I meant was there was no openvswitch kernel module installed !
and ovs-vswitchd in uspace was pegged at 99% while running traffic.

When we made sure that openvswitch_mod (KLM) was installed, then
we see netback/ kernel threads using CPU which is what we were
expecting.



> > We manually pulled in ovs sources and installed it as kernel module and
> > things are much
> > better now.
> >
> > Can someone pls point me to a high level architecture of the ovs data
> path ?
> > Looking at the
> > ovs-1.3 code to understand this meanwhile :)
>
> There's some information in the porting guide:
> http://openvswitch.org/cgi-bin/gitweb.cgi?p=openvswitch;a=blob;f=PORTING


Thanks for the pointer.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://openvswitch.org/pipermail/ovs-discuss/attachments/20120120/c2229bec/attachment.html>