[ovs-discuss] patch port performance

Ben Pfaff blp at nicira.com
Tue Mar 4 17:01:40 UTC 2014


Prior to 1.10, the penalty of going through a patch port is to
essentially double the latency of flow setup (because each packet goes
through two flow setups instead of one).  This can be insignificant in
some cases (a small number of flows) or a big penalty in others (a large
number of short flows).

On Tue, Mar 04, 2014 at 10:20:01PM +0530, murali reddy wrote:
> Thanks Ben. How bad it can get in versions prior to 1.10?  I want to build
> a solution with patch ports with XenServer 6.2 which has 1.4.6 OVS version.
> 
> 
> On Tue, Mar 4, 2014 at 9:48 PM, Ben Pfaff <blp at nicira.com> wrote:
> 
> > On Tue, Mar 04, 2014 at 09:42:16PM +0530, murali reddy wrote:
> > > Are there any known performance issues (latency or throughput) with using
> > > patch ports? I would like to interconnect multiple bridges (corresponding
> > > to different subnets) with patch port to a bridge configured with the
> > flows
> > > to acts as router. I am wondering if there would be any performance if i
> > > use patch port.
> >
> > On OVS 1.10 and later, there should be little or no performance penalty
> > for using a patch port.
> >



More information about the discuss mailing list