On Tue, Oct 22, 2013 at 05:59:51PM +0200, Jiri Pirko wrote: > Tue, Oct 22, 2013 at 05:30:16PM CEST, [email protected] wrote: > >So for me that prompts a new question. I can easily believe that teamd > >has two or three times as many bonding features as OVS. But do you have > >a way to know or guess what fraction of users actually use the features > >in teamd that aren't in OVS? Another way to look at it is, by adding > >more features, what unsatisfied potential OVS users would then be > >satisfied? > > Following table compares team to bonding by features. I should give you > an idea what teamd/liblagg would be capable of: > https://github.com/jpirko/libteam/wiki/Bonding-vs.-Team-driver-features > > From the features that OVS does not have I would note: > load-balancing for LACP support
OVS does do LACP load-balancing. > active-backup TX policy & port priorities and stickiness > ARP link monitoring > NS/NA (IPV6) link monitoring OVS does not have the features above. > multiple link monitoring setup & separate per-port link monitoring setup I'm not sure what "multiple link monitoring setup" is, but OVS does configure link monitoring separately per-link. > I'm not sure about OVS user needs. But I know that many of users of > current bonding driver are using active backup with arp monitoring for > instance. If this user decides to migrate to OVS, he would need to > change topology. OVS already has multiple link monitoring schemes (carrier, MII, CFM, BFD, and even combinations of those). I don't think that it would be too much work to implement additional modes. OVS already has code elsewhere that deals with ARP and IPv6, so it wouldn't involve much extra infrastructure. Thanks, Ben. _______________________________________________ dev mailing list [email protected] http://openvswitch.org/mailman/listinfo/dev
