Re: [mpls] draft-iwijnand-mpls-mldp-multi-topology - review

[Gyan Mishra <[e-mail]>]

Dear authors

I see the draft is in queue for WG adoption so I have provided a quick
review prior to the adoption call.

The original draft published in 2011 and the goal was to extend mLDP  P2MP
and MP2MP FEC RFC 6388  to be IGP MTID aware OSPF MT RFC 4915 and ISIS RFC
5130.

The replacement draft with same name posted in 2018 added support for IGP
Flex Algo and so now the latest draft revision 4  extends mLDP RFC 6388 to
support both IGP MTID which was the basis of the original draft as well as
now Flex Algo added to the draft.

MT IGP concept provides a common data plane with separation of control
plane functionality, thereby giving the ability for operators to create sub
topologies “network slices” of the topology as an  IGP abstraction layering
at the control plane level.

Flex Algo  draft was originally published in 2017 and now been submitted
for publication.

The goal of  Flex Algo was to also provide a simpler control plane based
“network slicing” capabilities but with simplification as well as current
applicability to only SR forwarding plane association of Algo with SR-MPLS
prefix SIDs or SRv6 locators.

One of the goals of SR-MPLS is the elimination of network state by moving
 the LDP control plane state to an SR IGP extension.

As multicast has been challenging for operators migrating to SR,
flexibility to keep LDP around for multicast only -  mLDP had been common
place and gives operators some time  to migrate to some of the up and
coming SR based multicast technologies.

This draft provides an extends mLDP FEC to be FAD aware as well as the
original goal of making MTR aware.

Throughout the document Flex Algo and MTR are mentioned and as the
extension that was originally for IGP MT, is now being extended to support
as well Flex Algo using the basis of RFC 7307 making LDP MTID aware.

My thoughts are it would be easier to keep the existing document
development in 2011 which goal was to extend mLDP to support MT and
creating a separate draft for extending mLDP to support Flex Algo.

Bundling both solutions together seems overly complicated as MT is very
different from FAD.  Also bundling together confuses the overall solution.

If you extend Flex Algo separately you don’t require any dependencies  on
RFC 7307 as now it would be a completely new solution and can focus on FAD
3 tuple and getting that encoded into a new FEC P2MP and MP2MP FEC.  I
think this drastically simplifies the solution.

IGP Flex Algo only supports SR Algo today and is being extended to support
other data planes such as now IP Flex Algo.

As IGP flex Algo draft today only supports SR data plane association with
prefix sid for SR-MPLS and SRv6 locator for SRv6, I think Flex Algo would
first require an extension to support MPLS LDP which does not exist today.
So that’s a major problem.

So if Flex Algo today does not support MPLS LDP data plane, how can you
extend mLDP (multipoint extension of LDP)  to support Flex Algo?

I think this is an important solution and would be happy to collaborate on
the draft to help progress.

Kind Regards

Gyan


-- 

<http://www.verizon.com/>

*Gyan Mishra*

*Network Solutions A**rchitect *

*Email [e-mail] <[e-mail]>*



*M 301 502-1347*