The MPLS WG Archive

[Vijay Srinivasan <Vijay.Srinivasan@cosinecom.com>]

Title: draft minutes from London meeting

Here are the draft minutes from the London IETF, thanks to Eric Gray.  Comments to the list please.

The presentations are available from:
<http://www.mindspring.com/~ewgray/ietf51/aatlas.ppt>
 <http://www.mindspring.com/~ewgray/ietf51/fhellstrand.ppt>
 <http://www.mindspring.com/~ewgray/ietf51/ppan2.ppt>
<http://www.mindspring.com/~ewgray/ietf51/tnadeau.ppt>
<http://www.mindspring.com/~ewgray/ietf51/aiwata.pdf>
<http://www.mindspring.com/~ewgray/ietf51/ppan1.ppt>
<http://www.mindspring.com/~ewgray/ietf51/sganti.ppt>

MPLS Working Group Meeting - 9:00 A.M. Monday, 6 August

 Chairs       Agenda

 Fiffi Hellstrand
              "Framework for MPLS-based Recovery"
              <draft-ietf-mpls-recovery-frmwrk-03.txt>

 Ping Pan     "A Method for MPLS LSP Fast-Reroute Using RSVP Detours",
              <draft-gan-fast-reroute-00.txt>

 Atsushi Iwata
              "MPLS Signaling Extensions for Shared Fast Rerouting"
              <draft-iwata-mpls-shared-fastreroute-00.txt>

 Alia Atlas   "MPLS RSVP-TE Interoperability for Local Protection/Fast
              Reroute" <draft-atlas-rsvp-local-protect-interop-01.txt>

 Tissa Senevirathne
              "MPLS Label Stack Authentication methods and algorithms"
              <draft-tsenevir-mpls-lauth-01.txt>

 Ping Pan     "Graceful Restart Mechanism for RSVP-TE"
              <draft-pan-rsvp-te-restart-01.txt>

 Sudhakar Ganti
              "MPLS Support of Differentiated Services using E-LSP"
              <draft-ganti-mpls-diffserv-elsp-00.txt>
 Tom Nadeau   "Link Bundling Information Base Using SMIv2"
              <draft-dubuc-mpls-bundle-mib-02.txt>

 Chairs       Workgroup status

(Slides for the below presentation are located in file fhellstrand.ppt)

Presentation by Fiffi Hellstrand on "Framework for MPLS-based Recovery"
(draft-ietf-mpls-recovery-frmwrk-03.txt). She gave an update on status of
this draft and requested that the draft go to last call immediately after
meeting and move in the direction of becoming an informational RFC.

George Swallow will post the last call to the mailing list.

(Slides for the below presentation are located in file ppan1.ppt)

Presentation by Ping Pan on Juniper Fast Reroute "A Method for MPLS LSP
Fast-Reroute Using RSVP Detours" (draft-gan-fast-reroute-00.txt) - designed
by Tony Li and Der-Hwa Gan some time back.  This is a flexible approach that
protects against both link and node failure.

Flexible and extensible. Single solution for both link and node protection.
Adaptive, minimal user impact and operational.  Designed to protect LSPs
selectively.  Support LSP detour merging to reduce # of LSPs.  Helped by
refresh reduction.

Question: (David Allan) Semantics of merging, etc. seem to imply central
control and the merging seems to reduce the ability of the approach to
support against a node failure?

Question: How do you protect against node failure?  How fast does this work?
Answer: detect node failure via refresh expiry and direct link failure.

Question: (Loa Andersson) Have interoperability issues been considered (for
example with LDP Fault tolerance?  Also, there is a general issue of 
preserving state/label mappings if you have part of the LSP setup by LDP
and part of them done by RSVP-TE.
Answer: None at present.

Question: How is merging performed using signaling?
Answer: George Swallow said that a draft needs to be generated to describe
path selection for detour LSPs, but the place where that work belongs has
not been determined yet.

Question: (Loa Andersson) what is happening to this draft?
Answer: (George Swallow) We have more drafts to be heard on this topic.
Other drafts are being presented that impact on this.

(Slides for the below presentation are located in file aiwata.pdf)

Presentation by Atsushi Iwata. He presented a detailed discussion of the
content of draft "MPLS Signaling Extensions for Shared Fast Rerouting"
(draft-iwata-mpls-shared-fastreroute-00.txt).

Question: (George Swallow) is there a dependency in this approach on
ordered setup.  If something goes wrong on the backup path, do you have
to start over?
Answer: At present, yes - but he will be looking into this.

Question: (Loa Andersson) is there an assumption that resource sharing
is the only parameter needed, even if this forces setup of a non-optimal
path through paths set up for "upstream" recovery paths for the same
primary paths?  There might be other parameters that could be used, for
example, the most optimal path from the node doing the recovery to some
merge point downstream.  Otherwise you might always be forced to use
shared explicit resources for backup LSPs.
Answer: The approach does depend on use of shared resources, if you can do
this, then you can use this approach.  Optimization issues depend on the
specific network topology in use.  If an operator designs a network
considering the proposed shared fast-reroute scheme, optimization can
be done in nearly every case. However, if there are topology constraints
preventing this, the operator will want to have the option - at the ingress
node - to use the Gan approach.

(Slides for the below presentation are located in file aatlas.ppt)

Presentation by Alia Atlas. She presented a detailed discussion of the content
of draft "MPLS RSVP-TE Interoperability for Local Protection/Fast Reroute"
(draft-atlas-rsvp-local-protect-interop-01.txt).

Ping Pan said his draft does support local repair.

Comment from George Swallow (not as chair): his draft and the Ping Pan,
Der-Hwa Gan draft are not competitors and the authors from the two drafts
should form a single draft design team.

Question: There appears to be a lot of complexity in these proposals,
will they be implemented or deployed?
Answer: From George Swallow, these are already implemented and deployed
and this is an opportunity to specify how this is done.

(Slides for the below presentation are not yet available)

Presentation by Tissa Senevirathne on "MPLS Label Stack Authentication
methods and algorithms" (draft-tsenevir-mpls-lauth-01.txt).  He discussed
the reasoning for this proposal: protect against connection hijack,
bandwidth hijacking and denial of service attacks.

Question/Observation: Because the nature of this authentication mechanism
involves some computational complexity, it seems likely that this approach
may easily be subverted to introduce a denial of service attack.

Question: How is this approach better than other proposals to protect
backbone traffic such as encryption at the link layer?
Answer: This is not encryption of MPLS.

George Swallow asked if there were people in the room who felt that this
was something they needed to see in their networks.  No one answered.

(Slides for the below presentation are located in file ppan2.ppt)

Presentation by Ping Pan of "Graceful Restart Mechanism for RSVP-TE"
(draft-pan-rsvp-te-restart-01.txt). He discussed the concepts of the
draft in some detail.  Conclusions: Useful, simple and compatible.

Question: (David Allan)

George Swallow said that this work will need to be taken to CCAMP for
more general usability.

Question: (Loa Andersson) Is there any effort to make this sort of thing
work for LDP as well.
Answer: There has been work on LDP fault tolerance to handle this.

Question: (Yanguang Xu) the suggested label object is being overloaded
here and we should think about using some other object.
Answer: (Lou Berger) authors felt that the same information would then
be used to carry duplicate information since much of the same information
would have to be in the suggested label object anyways.

George Swallow suggested taking this off-line.

(Slides for the below presentation are located in file sganti.ppt)

Presentation by Sudhakar Ganti on "MPLS Support of Differentiated Services using
E-LSP" (draft-ganti-mpls-diffserv-elsp-00.txt).  He provided an overview
of the contents of the draft.  The draft essentially proposes explicit
signaling for E-LSP setup.  The draft also includes how the mechanisms may
be extended for E-LSP support in CR-LDP.  He requested that the document
become a working group draft.

George Swallow asked for a sense of how many felt that this was work we
should undertake.  Several people raised their hands (much less than 1/2
of those present).  George Swallow stated that there is a simplistic
approach described in existing specifications, and he would like to see
some stronger support before adopting this work.

Question: What is the driving application for this? What criteria are used
to determine relative survivability for large numbers of LSPs?
Answer: The driving application is mostly L2VPN deployment.  The criteria
for survivabilty is not the issue here.

Question: (Vijay Srinivasan) Given that AF and EF are not going to be on
the same LSP anyways, are we going to signal separately for each AF class?
Why would this be useful?
Answer: (Scott Bradner) there isn't a defined relationship between
AF classes and no natural presumption of a priority order between them.

Comment: (Bruce Davie) Not much added by this draft.  You have to buy into
the premise that effectively signaling L-LSPs using the EXP bits is something
we want to be able to do.  There doesn't seem to be any obvious need for this.
Answer: There is value in doing this.  A lot depends on how you want to
distinguish the different service levels in a single LSP and how we deploy
these services. End-to-end service deployment with SLA gurantees and path
protection is a good candidate for using E-LSPs

George Swallow stated that we do not want to build mechanisms for mechanism's
sake - we need to understand what the need is for the proposed mechanism.

(Slides for the below presentation are located in file tnadeau.ppt)

Presentation by Tom Nadeau on "Link Bundling Information Base Using SMIv2"
(draft-dubuc-mpls-bundle-mib-02.txt).  He first summarized the draft for
those who have not had a chance to read it, then discussed the evolution
of the draft and asked to have it accepted as a working group document.

Question: Do we have/need an overview of the various MPLS and TE MIBs to
simplify the process of determining what MIBs are used for what purposes.
George asked to see how much support there is for the link bundling MIB draft
as a working group draft.  Many hands were raised, however, George suggested
that we take the discussion of whether or not to adopt this work in the MPLS
working group to the mailing list.

George Swallow used this as an opportunity to segue into MPLS draft status,
saying that he would like to hear the status on unnumbered link, hierarchical
LSP setup and link bundling from Kireeti Kompella.  Kireeti said that he is
going to respin these drafts soon.  Kireeti talked briefly about changes that
need to be made to these drafts.

George asked if the drafts would be ready for last call by the end of
August.  Kireeti said between Yakov and himself, they should be able to get
it done.  He cautioned that he would be on vacation until the end of August
himself however.

Question for the chairs: what is the status of drafts in the IESG and
RFC editor's queue?

Scott Bradner said that he would get back on that.
###################################################################################################### This email communication may contain CONFIDENTIAL INFORMATION and is intended only for the use of the intended recipients identified above.  If you are not the intended recipient of this communication, you must not use, disclose, distribute, copy or print this email. If you have received this communication in error, please immediately notify the sender by reply email, delete the communication and destroy all copies. ######################################################################################################