L2VPN - Vancouver Sept 2013 Introduction Welcome from Giles and Nabil explaining it’s a full agenda and we are tight on time. 00-drafts will be at the end, and hopefully there will be time for them. Note Well Document Status There are no new RFCsince the last meeting, but there is a lot in the queue. 1 in the RFC and 4 at the ESG The inter domain draft requires ICCP and that is moving to RFC. Stewart - If you don’t anticipate that ICCP will change then you can just push the draft through. On VPLS - VPLS broadcast extensions we can move forward now that point to multipoint PWE is moved. Pim snooping is being presented here and in PIM, as we wanted feedback from pin working group. EVPN main protocol draft is almost ready to go. ETREE - the requirement is in ESG last call, the framework needed an update, once that is done it moves to the AD writeup, then we can move to the solution draft. Dave Ericsson - ICCP completed working group last call, so you could move forward with it. The miscellaneous drafts: The first one is Mac optimisation draft, this needs someone to update it as the author has moved on to another company. The second, mac flush loop detection, this has 2 authors from one place and we are looking for any implementation. This is looking like it will end up as informational. VPLS multii-homing, heading to last call. Drafts 1) Usage and applicability of BGP MPLS based Ethernet VPN http://http://tools.ietf.org/html/draft-rp-l2vpn-evpn-usage-01 Jorge Rabadan jorge.rabadan@alcatel-lucent.com 5 minutes Jorge presents the slides slide 2 - from feedback from Giles changes where made and thank you for that. Questions at the Mic Lucy - I just want to clarify what does vlan aware bundle mean Jorge - in one instance you can have multiple bridge domains the treatment is that Ali - another way to put it is that in instance you you don’t have to step up multiple vlans for each instance. Lucy - so you allow each broadcast domain to have it’s own Mac address Ali - explains the same thing again Lucy - so you have 1 route target ? Ali - you have 1 route target for all those broadband domains. Section 9.4.11 Lucy - your answer now has made me confused, you are saying that each vlan has it’s own broadcast domain Giles - not sure that asking questions about what is actually explained in the draft is a good use of time. 2) IP Prefix Advertisement in E-VPN http://draft-rabadan-l2vpn-evpn-prefix-advertisement-01 Jorge Rabadan jorge.rabadan@alcatel-lucent.com Jorge presents the slides Questions at the Mic Lucy - can you go back to use case one. I’m not convinced that we need a new route type for this use case Jorge - you need to advertise the subnet behind the network, so you can use EVPN to advertise the prefix. Lucy - but why to you need to do this? Jorge - no you need to advertise this as they are behind Ali - with route type 2 you have to withdraw all, when you do this you don’t have to withdraw all, you just have to pull the next hop. Giles - we need to move on to the next draft. Lucy - ok, I have a question on another use case. Giles - I’m afraid we do need to move on. Nabil - We need to move on. Lucy - but just one sentence Nabil - really we need to move on, please take it to the list 3) IM Snooping over VPLS http://draft-ietf-l2vpn-vpls-pim-snooping-05 draft-ietf-l2vpn-vpls-pim-snooping-05 Olivier Dornon (olivier.dornon@alcatel-lucent.com) Oliver presents the slides No questions 4) VPLS PE Model for E-Tree Support http://draft-ietf-l2vpn-vpls-pe-etree-02 Yuanlong Jiang (jiangyuanlong@huawei.com) Yuanlong presents the slides Giles we are going to wait until the framework is done and then we will progress this one. 5) E-TREE Support in EVPN & PBB-EVPN http://draft-sajassi-l2vpn-evpn-etree-02 Ali Sajassi (sajassi@cisco.com) Ali presents the slides Nabil mentions presentation time. Questions at the Mic Lucy - I think this is a new mechanism. Every Ali - No this is the same advertisement. In the data plane the advantage is that you don’t have to do split horizon or another label. Lucy - but for the unicast you will carry multicast Ali - let me put it this way, it makes it Lucy - I think this is complex forwarding plane 6) A Network Virtualization Overlay Solution using EVPN draft-sd-l2vpn-evpn-overlay-02 Ali Sajassi (sajassi@cisco.com) Ali presents the slides Questions at the Mic No name - I think it will help to standardise the draft and help with the deployment of EVPN Jorge - you have a question about flooding list. Ali - I did not want to spend time on a section we might revamp. To explain this would take considerable time. WE can discuss it offline if you would like Jorge - Would like to discuss it offline, there are issues around encamp. Offline would be good. Lucy - One comment to Ali, I think you are against the global label. For this draft you want to use the global label or the local label? Ali - EVPN draft talks about both of them, The global ID is in a given DC and that is still the case. If you are going from one DC to another with a different space, this is taken care off. Ali - The service layer is still MPLS, ? - You mean that is local label? Giles - Nabil and I had a discussion about the overlaps and we have reach the consensus that L2VPN is the right place for this work Nabil - the way the existing draft is being shepherding is for experimental, so what we do here has to match. Stewart - I need to revisit if it is experimental or informational. But it will be AD sponsored and not a working group doc. Nabil - if we progress then we will match it to whatever the VXLAN draft is. Stewart - lets look at this offline. 7 VPWS support in E-VPN http://draft-boutros-evpn-vpws-02 Sami Boutros (sboutros@cisco.com) Sami presents the slides. Sami - this draft has been around for more than a year now. Questions at the Mic Jorge - so the next version you are going to explain multihoming active active? Sami - the draft already talks about this Jorge - I think it is not well described are you going to elaborate, specially the active standby. One more question is even for single homed CE you are encoding the circuit into the ESI is that right? Sami - the ESIID can be a sub-interface you might want a point-to-point not for the whole port. Jorge - but you will want to group into the same route target right? Jorge - if you use a RT for AC ID you can put ESI as 0 so you don’t have the clash you mention here. Sami - this is a good point Jorge - you might want to withdraw each AC ID when you send a BGP withdraw, I think it is a better option to have it. Ali - This is a good comment and we should take it offline. 8) MAC Address Withdrawal over Static Pseudowire http://draft-boutros-l2vpn-mpls-tp-mac-wd-02 Sami Boutros (sboutros@cisco.com) Sami presents the slides. Questions at the Mic This was presenting by Himanshu in the PWE group and we are presenting it again. It has been around for more than 2 years, and we are reviving it to make the static PWE feasible. Giles - do you want to adopt it here or PWE? Sami - it does not matter which working group, the function is needed and it is holding up the deployment. Zhenbin from Huwei - in the use case for a base station to base station controller, I think that that Sami - We are talking about static PWE in an access network, in an edge scenario. We are saying you can have PWE redundancy with static PWE, but today the Mac withdraw function is missing. Zhenbin - you mention the access network use MPLS-TP Sami - no we are talking about an access PWE an edge scenario, we are not saying the core is going to use MPLS-TP, we are talking about this in access. sami - shows slides and describes where is access Nabil - In MPLs we use LDP to signal the PWE between one access network and another. here you are replacing that with static, as there is no signalling for mac withdraw, and you are replacing that. 9) Flow-Aware Transport of Pseudowires Extension for BGP http://draft-keyupate-l2vpn-fat-pw-bgp-00 Sami Boutros (sboutros@cisco.com) Sami presents slides This is a draft for Fat PWE, the same problem statement as in 6391. Questions at the Mic no name - you are quoting 6391 as the justification for doing this for BGP PWE. Today it solves the problem in a general purpose manner, so I am interested to know why you are doing this. Sami - we have already deployment in 6391, when LDP was signalled, and that happened before entropy. We are simply saying that the mechanisms already implemented we need in static. Stewart - are you running a negotiation over BGP, rather than just an advertisement. Sami - the signalling is already we are using it to negotiate the entropy label. Stewart - but if the 2 ends don’t agree then you negotiate, did I misread it? Sami - each node is advertising the support for the flow label and then you send the flow label Stewart - and we should discuss it offline. Sami - we can disuse it offline. Lucy - Several years ago I read another draft for the same purpose and I promoted that 6790, it that draft they specify BGP and LDP for doing this. So not sure why we are debating this. On the PWE you may have multiple VPN supported, so each one doing there own, you should have one tunnel doing it once. Sami - what we are saying is that this is 6391 a section for BGP that was missing. Lucy - I’m just informing you we did this for this purpose. Giles - we ned to cut the line ? - would security like MD5 be supported? giles - are you talking about Sami’s first presentation? ? - yes the mac withdraw one Sami - the static mac withdrawal as a security concern? The same apply from the ?…we can look this up Giles - lets take this offline 10) Updated processing of control flags for BGP VPLS http://draft-singh-l2vpn-bgp-vpls-control-flags-00 Ravi Singh (ravis@juniper.net) No questions 11) An Architecture of Central Controlled Layer 2 Virtual Private Network http://draft-li-l2vpn-ccvpn-arch-00 Zhenbin Li (lizhenbin@huawei.com) Zhenbin presents his slides. Questions at the Mic Ali - So this model with the control and forwarded is essentially vPE? Zhenbin - I think this is different as the the functional is moved from the PE to the controller? Ali - but that is vPE Nabil - the same questions came about in IDR when you presented . With my chair hat off this is the vPE model, and you are not changing the fundamental protocols, and is this the place to define this? Ali - vPE is not in the scope of the working group Nabil - the protocol is not in the definition of the working group. we need to define what a protocol cold carry and if it is for this working group. Ali - yes otherwise this will be a sprinkle over all working groups. Lou Anderson - I think I agree with Nabil, but I think we need to go out as WG Chairs and stop the duplicate presentations Nabil - point well taken, I was not aware of the other presentations in other working groups. Lou - we had at least 3 WG’s overlapping with each other - check audio Wim - If it is really describing the arch of how you decompose the VPE, your draft is very generic you could instead pull out the problems you are trying to solve. Nabil - Lets take info from across the working groups. Giles - you have 7 and a half mins for the next two presentations. 12) Selective Multicast in EVPN http://draft-zhang-l2vpn-evpn-multicast-00 Zhenbin Li (lizhenbin@huawei.com) Zhenbin presents his slides. Questions at the Mic Ali - just a comment we had selective multicast as the EVPN baseline. This is a good basis for a separate draft, we would need the input from the services providers on this, which would be valuable. 13) Using BGP between PE and CE in EVPN http://draft-li-l2vpn-evpn-pe-ce-00 Zhenbin Li (lizhenbin@huawei.com) Zhenbin presents his slides. Questions at the Mic Ali - Can you go back to the diagram? This multihoming, is this single active or all active? Zhenbin - all active Ali - So what is the issue ith all active? Zhenbin - oh single active Ali - Ok, but what is the issue with single? In either case, in all active if there is a failure you withdrawn the next hop. Aliasing works for single active and all active, this is why you have it. So what are we trying to solve? Zhenbin - check audio Ali - There is no issue getting the packet from CE3 to CE1 when the link fails. Zhenbin - bad audio, can’t make this out. Ali - but this is aliasing, it is for this purpose.. Zhenbin - bad audio, can’t make this out. Wim - what we do is we reuse all the IPVPN mechanism we already have. Jason Eriscon - There is a race condition I can see here. Once PE sends the withdrawal before PE2 has learnt the address, the aliasing stops, or does it? Ali - in that case it will do the forwarding, it has a back off mechanism for it. If one link fails in the bundle it goes to the next one Jason - it starts to sent it as BUM traffic yes. Another point in your draft is that if you drop and don’t forward that is when the problem happens, I have a simple solution, don’t drop it. Ali - lets make sure we are on the same page, if you have AC failure you will withdraw….the second thing is that is withdrawing the Mac , which will be when the timer expires. So for 5 mien you don’t get any traffic, and you are saying you don’t want to push traffic over the core. Jason - the 5 mins is good. Zhenbin - I know your solutions, can’t make out audio Ali - are you saying this is a faster convergence. Zhenbin - We want to unify the … Nabil - Ali you are eating your own time.. Ali - in the baseline draft it says that you can learn over AC both in data plane or the control plane. So it will distribute it. This is all in the baseline. 14) (PBB-)EVPN Seamless Integration with (PBB-)VPLS http://draft-sajassi-l2vpn-evpn-vpls-integration-00 Ali Sajassi (sajassi@cisco.com) Ali presents his slides. Questions at the Mic Jorge - there are a lot of VPLS deployments out there without BGP AD, so it would be easy to add FEC128 to this solution Ali - I agree and we should be able to static Lucy - there is no requirement to support ETREE over this requirement? Ali - We have not had a requirement on ETREE yet. 15) Proactive fault detection in E-VPN http://draft-vgovindan-l2vpn-evpn-bfd-00 Ali Sajassi (sajassi@cisco.com) Ali presents his slides. Questions at the Mic Jorge - So CFM. Only connectivity verification. You are not considering performance measure then? Ali - bad audio Jorge - so why not use Y1731 to begin with? Ali - I was talking about fault management, and not performance management Ali - does that cover your question? Jorge - for the transport you are not considering performance measurement there as well? you refer to MPLS OAM which is BFD. Ali - I’m not trying to reinvent anything, if there is something missing in transport the MPLS working group should take this. Jorge - ok so in your opinion is that sufficient? Ali - the middle layer is the even OAM layer, we need to check both fault and performance management. If we decided we need to do some which transport does not provide. Lucy - does this requirement and framework include GRE? What will be different. Ali - It is MPLS centric at the moment. Lucy - Do you have a plan to include this? Ali - Yes it should be comprehensive Nabil - we have less than 2 mins left. Lou - are you taking about MPLS OAM or MPLS TP OAM Ali - With respect to this we are talking about it is just the BDF construct for monitoring, that is not at the transport layer, it is at the EVPN layer. This is like VCCV. Back to the slides. (EVPN Network Layer OAM) Rap up Giles - Great discussion today, we would like to see more of this on the list. Done