Re: IETF 63 L3VPN Minutes

[Alex Zinin <zinin at psg.com>]

Ron,

 In many places below "Cisco" or "Alcatel" have been used instead of
 names. Can we please fix this?

-- 
Alex
http://www.psg.com/~zinin

Monday, August 8, 2005, 8:21:12 AM, Ron Bonica wrote:
> Folks,

> Thanks to Hamid, Spencer and Scott for taking minutes. If I hear no 
> objections by Friday, I will post these minutes plus all slides that 
> were presented.

>                             Ron

> ===================================================================


> Layer 3 Virtual Private Networks (Internet Area)

> Monday, August 1 at 16:30-18:00
> ==============================

> Chairs:
> Ross Callon <rcallon at juniper.net>
> Rick Wilder <rick at rhwilder.net>
> Ron Bonica  <rbonica at juniper.net>

> Agenda:

> 16:30-16:40 Document Status Review - Ross Callon

> Ross presented the status review.

> - Made good progress. Ross mentioned the new RFCs
>    published.

> - bgp/mpls vpn specification initially waiting for addressing
>    the BGP extended communities within the idr wg. Recently it
>    was approved to progress and bgp/mpls vpn can progress now.

> - iesg evaluation for pe-pe gre or ip in bgp/mpls, revised text
>    in wg last call.

> - ospf for PE-CE protocol (revised in may) waiting for iesg.
> - bgp auto-discovery revised given the feedback received, back
>      to AD.

> - Waiting for write-up:

>      - Constrained VPN route distrib updated, waiting on chairs
>      - BGP-MPLS IP VPN for IPv6 updated, recently provided to IESG.

> - Revised id needed:

>    Virtual Router Architecture and its associated applicability
>    statement.  Comments sent to authors.

> - Under AD evaluation:

>    An architecture for PP CE-based VPNs using IPsec and associated
>    AS needs review and update.

> - Architecture for PE-PE IPSec tunnel for bgp/mpls ip vpns
>    recently updated based on last call comments


> - Current work:

>     o Requirement for multicast
>     o Multicast solution in bgp ip vpn
>     o CE-CE member verification
>     o l3vpn import export verification (no activity on the above 2 docs).
>     o appl stat, framework
>     o definition textual convention mib for bgp/mpls ip vpn.


> Multicast VPN Requirements - Thomas Morin
>    draft-ietf-l3vpn-ppvpn-mcast-reqts-01
> =========================================

> Thomas Morin presented the draft-id.

> - A new update posted 2 weeks ago.
> - in this presentation will go through the changes,
>    the multicast vpn survey, and we talk about next work.
> - On the changes 2 new sections in the draft have been added.
>    o Carrier's carrier requirements and
>    o New section on QoS (ability to offer different QoS level to
> different customers.

> - Updated sections: QoS (5.1.3) maintain join and leave delay
> requirement
>   (refer to RFC2432)and minimum MTU.
> - Tunneling technologies need to mention P2MP LSP as much possible.
> - Compatibility and migration issues solution should state
>    a migration policy.
> - Trouble shooting provide the operator with
>    means LSP ping
> - inter-as section (should provide inter-as mechanism requiring
> least....
> - Big changes on section 4 (Uses cases) illustrated deployment
>    requirements.
> - describe use cases scenarios
> - provide order of magnitude for scalability requirement,
> - waiting for survey.
> - finally some edits changes...

> - For multicast vpn survey (proposed at last ietf)
> - survey overview (to be answered mainly by ISPs)
> - focus on future expected deployments.
> - typical questions Quantitative and qualitative (type of multicast
> deployed, etc).
> - Survey launched last week posted on different WG lists
>    please answer it and send completed survey to Daniel King Dan
> (dnni.com)
> - answers expected by Sept 15 05.


> Tom described the Next items for the draft such as:
>   o complete section 4 with the help of the survey and
>   o refine the requirements (PE-Ce protocols, inter-AS,
>     carrier's carrier, extranet tunneling protocols, etc)

>   o Address some open questions relevance of MTU-related sections.

> - Conclusion: requirement is mostly mature except section 4.
> Tom asked the audience to provide comments on the draft
> on the mailing list and answer and disseminate the survey.

> - Questions on the draft:  no questions.

> Multicast VPN Solution -  Rahul Aggarwal
>    draft-ietf-l3vpn-2547bis-mcast-00
> ========================================

> Rahul presented the draft:

>   This is an update on multicast draft, this is Wg document.
>   Rahul shows the co-authors/contributors,
>   fully committed to move this work forward,

>   This presentation for discussion open options.
>   Reduce options if possible, outline the issues and
>   specify required and optional procedures,
>   looking at MVPN routing information exchange
>    service provider technologies

> - need to look at scalability of entire network.
>    consider rate of churn C-joins/Prunes
>    number of protocol sessions require frequent periodic
>    changes

> - how does it fit with 2547 operational model unicast

> Rahul shows a table for MVPN routing exchange
> not exhaustive list. Periodic refresh, session per PE UI-PMSI,. etc.
> for BGP, PIM UNicast, PIM multicast...

> - Do we really need PIM-SM with GRE?  Discovery can be done using
>     BGP, shared trees can be built with PIM-SSM.

> Conclusion

> Goal to address these issues and produce 01 version.

> Ross: is you proposal is to propose this options to
> the WG list. Make Wg aware of the options:

> - Ross Callon: is your proposal to ask these questions on the WG
>    mailing list?

>    Rahul: Yes. Want to capture these questions in the
>    minutes.  Send email on the list and initiate the
>    discussion.

> -Question on BGP encoding to be published. They are already published
>   encodings for BGP to carry MVPN information

>   Rahul: Existing proposal may or may not be used.

> - Question (person from Cisco):  Wants backward compatibility with what
> has been
>      deployed for years and the Rosen draft.

>    Rahul: Certain options have been deployed, some not. Point well taken.

>           Need WG input.

> - Question: Deprecate PIM-SM?  Some providers already use
>      it.
>    Rahul: Point taken. OK.

> - Question:  need periodic refreshment. There is are WG items that try
>    to reduce periodic refreshes. Need to consider these approaches other
> that may reduce the overhead.

> - Albert, Redback: PIM needs periodic refresh?  There's some work in
>    PIM WG to reduce periodic refreshes. Need to consider these
> approaches.


>    Rahul: Yes, need to look at pragmatic options, point taken.

> - Question (from Cisco):  What is a service?  The only difference to the
>      MVPN service is whether using SSM or ASM.  The protocol you use to
>      implement it should be a separate issue.

>    Rahul: draft should talk about applicability of protocols.
>           Draft has told about tunneling technologies

>    Yakov Rekhter: I disagree. Need to specify which protocol for
>                   interoperability reasons.
>    Toerless: same with IS-IS and OSPF.
>    Yakov: yes.
>    Toerless: what does rfc2547 say about IS-IS?
>    Dan Alvarez: At least why do we think BGP is suitable.  For
>    example there is no information on dynamic building of OSPF trees
>    etc.

> Comment (Person): this is the wrong level of detail. Why this need to be
>                    specified.

> Yakov: I disagree because of interoperability reasons need to specify
> which protocol
>         to use.

> Question (Albert from Cisco): last comment: how well bgp is suitable for
> multicast
>    bgp is not used for intra multicast operation, why BGP is suitable.
>    no information on dynamic multicast tree and how it related to
> scalability.

>    why use BGP as replacement for all existing multicast protocols...

> Rahul: Not talking of building trees with BGP,
>     just transporting with BGP. one item for consideration

> Alcatel: The one thing for BGP helps is to provide a reliable transport.

> Rahul: Note that BGP does have filtering mechanism...and is applicable
>         in this case.

> Ross: some of discussion can go to the text...

> Multicast VPN MIB - Tom Nadeau
>    draft-svaidya-mcast-vpn-mib-02
> ================================

> Tom presented the draft.

> Propose this document to manage rosen-multicast doc. interacts
> well with MPLS MIBS.
> draft need to use the combined approach will be published soon
> after meeting.

> - Ross: Have you got input?
> - Tom: Yes. Will be published soon.
> - Tom: Can we publish it directly as Wg doc.
> - Ross: The MIB managed the combined draft. When the MIB will be
>       updated to reflect combined drafts then we have two choices:
>        either submit as individual contribution and request WG to
>        ask for WGs, or just make it wg doc.

> Ross: My personal option if the authors of both sides agree this is
> the agreed way forward then I don't see objections to adopt it
> as WG doc.

> Ross: Does anybody has any objection to that?

> No objection from audience...


> Virtual Router Multicast Solution - Hong-Ke Zhang
>    draft-zhang-l3vpn-vr-mcast-01.txt.
> =================================================

> Presenter: Spencer Dawkins. Spencer mentioned Zhang couldn't make the
>             meeting.


> There were some scalability questions on the mailing list such
> as the number of trees in SP core will not exceed the number of
> VRs, and does all multicast traffic in a VR share the same tree,
> answer yes.

> Does this approach require PIM-DM mode
> answer no, next version will say this more clearly.

> Questions:

> - Ross: whose read the draft:

> - Ross: few hand for those who read the document. Obvious question
> out of 10 would that be any objection for those become WGs. Not quite
> enough need to take it to the WG list and ask the question.


> - Ross: if anybody is interested in deploying multicast for
> vr please bring those to the mailing list.