IETF 85 BFD WG 2012.07.30 1540 PDT Agenda: BFD Agenda, IETF 84 ----------------------------------- Monday, Afternoon Session II 1540-1710?? (1.5 hour) Administravia - 10 minutes (chairs) BFD over LAG - 30 minutes BFD and BFD for MPLS MIB review breakout - 30 minutes Note taker: DWard@cisco.com I. Administrivia See Slides Significant notes: MIBs ready for last call Base MIB and MPLS BFD MIB BFD Crypto mechanisms. Doc stable waiting for deployment experience Charter status: P2MP draft stable, waiting for deployment experience Bootstrapping experience from DHCP Occasional interest BFD for MPLS-TP Appear done? LC in MPLS WG awaiting publication Assist with BFD for TRILL Appear done? IETF LC passed. New work: BFD over enet LAG II. BFD over LAG presented by Sami Boutros See slides Significant notes: Use of dedicated mac in DOWN/ADMIN DOWN VLAN tag usage clarification LACP doesn't interact with LACP but, the LAG management module has the interaction instead Discovery of endpoint mechanism will be presented in the future Questions: Tom Israeli from Marvell This draft addresses single hop in layer2 so, it's important to state this in draft.?? Jeff Haas: it isn't stated so firmly because one possible path in the future will be non-directly connected neighbors. Any significant changes required will be added at this time.?? Shane Amante from Level3: Why transition from dedicated mac to src. Useful to know src addr??? Sami: Some HW not capable of programming the src addr and interface and not big enough to rx more MACs. There can be other reasons Jeff Haas: As implementations are finding, switching MAC addrs saves resources for some but, harder for other and so in future probably will see a "knob" to control the behavior. Jeff Haas: this is an open technical issue. Other technical issue is using a specific set of IP addrs. Up to config (per spec) or using the security mechanisms to accomplish one of the things LACP provides. In BFD there is no level of service to make certain all member links meet criteria to bring up LAG interface ??? like LACP. Jeff Haas: For those that read the draft can we see satisfaction with: MAC addrs:No hands Shane - need more explanation IP Addrs: No hands Sami - Auto config is an option in the other draft Shane - This draft is fine saying nothing. From an IP adds standpoint manual config is fine but, advocate using link local address than other options proposed. Please fold auto-disc draft into this draft pls. Adrian Farrel on responding to IEEE: IESG and IAB met w/ IEEE leadership to talk more generally about interactions between two SDOs. BFD over LAG emerged for 5-6 mins and IEEE is pleased with interaction. Clear discussion and early exposure to work was good move. IEEE appreciative that comments were listened to and taken onboard. Some nervousness because IEEE has own OAM mechanisms and one of the things that the WG could do for politeness is that we explain any limitations to IEEE OAM and we are working on it and would like to know if they are working on it. Are we trying to shoe horn BFD onto IEEE only links (enet only). Continued cycled of communication would be smart and next stage is "thanks for communication, we believe we've addressed your comments and we are starting to discuss how to add to our charter, here is the latest version of our draft is XXXX and please join the email list." Others talked about architecture and layer violations but, were a long way from product development but, if a good bunch of vendors wanted to do this there would be no opposition and challenges.?? Jeff: Sounds like our next response is to update our liaison response and then recharter. Adrian Farrel: Yes, that sounds right Glen Parsons: Clarification on draft. Is the dedicated MAC address is it unicast, multicast, etc Jeff Haas: Item for discussion. BT has been herding vendors along with other operators. We've had all low level discussion in the open. Some are happier with unicast vs multicast. Really it's about bootstrapping so uni vs multicast doesn't really matter but, after switchover is where it matters. Resolution will require implementation and then we'll request from assigned pool.?? Glen: It will be only one addrs right? Jeff: There can be only one. Adrian: Glen ???as an IEEE person, did you hear the same thing in the mtg Wed? Glen: Two schools of thought: this is overlapping mechanism with what exists in LAG and there were others that the architectural issue and that BFD was in the middle of LACP where there is no open interface. The real proof is where you have BFD and IEEE solution and they would destroy each other. Should be ships-in-the-night and not cause confusion. Jeff: Took a step back and asked "what is the functionality we want to provide in the presence of enet link aggregates" LACP doesn't provide failover time in the range operators were looking for. We weren't trying to detect enet component link failure. Functionality is whether or not IP forwarding was working over component links. Also, does this apply to switched networks with more than one hop in the middle.Can this same mechanism be applied to MPLS LAGs Trying to find a generic solution pointed at IP. Glen: there needs to be an understanding that we don't break each others solutions. Shane: Simplicity of configuration is a very important consideration and not to subvert that. III. BFD MIB This breakout session's intent was to audit the BFD base MIB and the BFD for MPLS MIB for open issues. We would like to last-call the BFD base MIB as soon as possible, but were unable to do so until it was determined if the BFD for MPLS MIB caused any issues. Participants: - Sami Boutros - Wenjing Cheng - Jeff Haas - Tom Nadeau Necessary changes in base BFD MIB needed by BFD for MPLS MIB: - discriminators need to be MAX-ACCESS read-create to accommodate configured case. Conformance statement needs to be updated appropriately. Necessary changes in BFD for MPLS MIB: A conformance statement is needed to impose additional restrictions on base BFD MIB objects: - bfdSessionOperMode - demand mode must be precluded as does multipoint - Address objects: IP addresses are optional and that Unknown should be used - Echo interval is not permitted to be > 0. Other changes: - Need mapping to OAM-ID MIB - GTSM use for MPLS-TP is an open question. - A number of enumerations should be moved to Textual Conventions. - There was discussion as to whether TCs should be moved to a separate document and made IANA managed MIBs. The consensus seemed to be that the additional constraints required in the DESCRIPTIONS of the objects in question would require a simultaneous update of the primary MIB. - REFERENCEs need to be liberally added to the objects. - The MplsSessEntry table needs RowStatus and StorageType objects. - The performance table needs the appropriate high capacity (HC) objects and the typical boiler plate in the conformance statement for such objects. - An audit of the MPLS-TP spec to verify that appropriate counters is recommended. -