That's a good point, I don't have experience with that (what's the best DSCP class), but we should make sure BFD packet, as control/monitoring traffic, is prioritised over other types of traffic.
On Tue, Oct 17, 2017 at 11:12 AM, Venkatesan Pradeep < [email protected]> wrote: > Hi Ben, > > The implementations I've worked on have used tos=0xC0 (DSCP CS6). Given > that DSCP CS6 is recommended for network control traffic, and is used by > OSPF, BGP etc., I think it is likely that many implementations do that for > BFD too. Since BFD flaps have significant impact, marking BFD packets with > the right DSCP value will allow the network to apply the right QoS for > those packets and prevent BFD flaps due to network congestion. > > Regards, > > Pradeep > > -----Original Message----- > From: Ben Pfaff [mailto:[email protected]] > Sent: Monday, October 16, 2017 11:13 PM > To: Venkatesan Pradeep <[email protected]> > Cc: [email protected] > Subject: Re: [ovs-dev] TOS for BFD packets > > On Mon, Oct 16, 2017 at 04:32:45PM +0000, Venkatesan Pradeep wrote: > > BFD packets are being sent with tos = IPTOS_LOWDELAY | > > IPTOS_THROUGHPUT which would put them in the best-effort class. It > > would be better if we send with tos = IPTOS_PREC_INTERNETCONTROL so > > that they can be appropriately prioritized in the network. > > Is that common practice for other BFD implementations? If so, then I > agree that OVS should follow it also. > _______________________________________________ > dev mailing list > [email protected] > https://mail.openvswitch.org/mailman/listinfo/ovs-dev > _______________________________________________ dev mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-dev
