[ovs-dev] [PATCH] bfd: Make the tp_dst masking megaflow-friendly.

Ben Pfaff blp at ovn.org
Thu Oct 4 01:27:57 UTC 2018


On Wed, Oct 03, 2018 at 03:30:48PM -0700, Han Zhou wrote:
> On Wed, Oct 3, 2018 at 3:27 PM Ben Pfaff <blp at ovn.org> wrote:
> >
> > On Wed, Oct 03, 2018 at 03:02:54PM -0700, Han Zhou wrote:
> > > On Wed, Oct 3, 2018 at 2:05 PM Ben Pfaff <blp at ovn.org> wrote:
> > > >
> > > > On Mon, Oct 01, 2018 at 02:12:29PM -0700, Han Zhou wrote:
> > > > > From: Han Zhou <hzhou8 at ebay.com>
> > > > >
> > > > > When there are tunnel ports with BFD enabled, all UDP flows will
> have
> > > > > dst port as match condition in datapath, which causes unnecessarily
> > > > > high flow miss for all UDP traffic, and results in latency increase.
> > > > > For more details, see [1].
> > > > >
> > > > > This patch solves the problem by masking tp_dst only for the leading
> > > > > bits that is enough to tell the mismatch when it is not BFD traffic.
> > > > >
> > > > > [1]
> > >
> https://mail.openvswitch.org/pipermail/ovs-discuss/2018-September/047360.html
> > > > >
> > > > > Signed-off-by: Han Zhou <hzhou8 at ebay.com>
> > > >
> > > > I think that we only need to "un-wildcard" one bit, not all the
> leading
> > > > bits.
> > > >
> > > > What do you think of the following slight generalization?
> > > >
> > > > Thanks,
> > > >
> > > > Ben.
> >
> > > Thanks Ben. Yes, your version is better! I didn't realize that one bit
> is
> > > enough to tell the unmatching. It is more megaflow-friendly. Will you
> send
> > > the formal patch?
> >
> > Oops, I just applied it, but here's a copy of formal version anyway:
>
> Thanks Ben. Do you mind back-porting to at least 2.10 and 2.9?

OK, done.


More information about the dev mailing list