LSP Stitching with Generalized MPLS TE draft-ietf-ccamp-lsp-stitching-01.txt - PowerPoint PPT Presentation

About This Presentation
Title:

LSP Stitching with Generalized MPLS TE draft-ietf-ccamp-lsp-stitching-01.txt

Description:

LSP Stitching with Generalized MPLS TE draft-ietf-ccamp-lsp-stitching-01.txt Arthi Ayyangar (arthi_at_juniper.net) Jean Philippe Vasseur (jpv_at_cisco.com) – PowerPoint PPT presentation

Number of Views:37
Avg rating:3.0/5.0
Slides: 6
Provided by: Juni5
Learn more at: https://www.ietf.org
Category:

less

Transcript and Presenter's Notes

Title: LSP Stitching with Generalized MPLS TE draft-ietf-ccamp-lsp-stitching-01.txt


1
LSP Stitching with Generalized MPLS
TEdraft-ietf-ccamp-lsp-stitching-01.txt
  • Arthi Ayyangar (arthi_at_juniper.net)
  • Jean Philippe Vasseur (jpv_at_cisco.com)

2
Changes from 00 version
  • Editorial cleanup and rewrite
  • Incorporated comments from WG
  • relevance of signaling adjacency between LSP
    segment end points
  • label allocation procedure over LSP segment for
    bidirectional LSPs
  • teardown procedures for e2e LSP over LSP segment
  • relationship between regions based on switching
    capabilities and LSP stitching
  • triggers for dynamic setup of LSP segment for LSP
    stitching

3
Discussion on list
  • Are the control plane procedures for LSP
    stitching REQUIRED for PSC and non-PSC LSPs ?
  • Yes
  • Should we allow LSP stitching procedures (control
    data planes) while traversing region boundaries
    ?
  • PSC-n to PSC-(n1) NO
  • Non-PSC NO/YES ?
  • Different opinions on this
  • Why not simply use LSP hierarchy in this case ?
  • Is there any specific benefit of using LSP
    stitching instead of LSP hierarchy for this case
    (control data plane) ?
  • Would such nodes even use LSP stitching control
    plane procedures as described in this ID or
    would they simply use LSP hierarchy procedures
    in control plane but stitch in data plane?
  • Does it affect anyone if LSP stitching between
    different SCs were disallowed ?

4
Issue of Bidirectional LSPs
  • Jean Louis brought up this point
  • GMPLS nodes use Upstream Label to detect
    bi-directionality of the LSP
  • When no Upstream Label is exchanged over the LSP
    segment hop, the information conveying
    bi-directionality is lost
  • Solution options
  • Some flag set by head end of e2e LSP to be used
    as an alternative to Upstream Label, when
    Upstream Label is missing
  • Two different signaling parameters to mean the
    same thing
  • Not compatible with existing procedures
  • Some flag/object set by the stitching node, only
    along the LSP segment hop
  • Continue to send Upstream Label in Path and Label
    in Resv over the LSP segment hop, but choose a
    unique label value to denote stitching
  • How does this affect Generalized Labels that have
    no restriction on value to be used ? E.g. Port
    Numbers
  • Should we even allow bidirectional LSP to
    traverse two different unidirectional LSP
    segments ?
  • OR
  • Any label value (no need for unique value) and
    simply place the onus on the receiver to ignore
    any Label/Upstream Label in case of stitching
  • This is already stated in the ID. Stricter rules.

5
Next steps
  • LSP stitching signaling procedures are currently
    optional for non-PSC
  • Make this a requirement, similar to PSC?
  • Resolve rules for stitching between different SCs
  • Fix the issue of absence of Upstream Label for
    bidirectional LSPs based on options listed before
  • There has been more discussion/comments on list
    after version 01
  • Adopt additional comments and publish early
    revision
Write a Comment
User Comments (0)
About PowerShow.com