[c-nsp] Seamless MPLS interacting with flat LDP domains

[adamv0025 at netconsultings.com]

> Igor Sukhomlinov
> Sent: Friday, April 26, 2019 5:23 AM
> 
> Hi all,
> 
> I wonder if anyone has experience with integrating a UMMT/Seamless MPLS
> domain (BGP-LU running over isolated IGP regions) with an existing flat
LDP
> network.
>
Yes from well before the term was coined, but that was for the Inter-AS
setup.
They used to call it hierarchical MPLS or something along those lines. 
 
> The customer wants to make sure the existing LDP domain is still running
> while the newer BGP enabled domain is steadily coming online, but what is
> the best approach to transport the services between the networks?
> 
First of all hierarchical MPLS is not for everyone so the obvious question
is do you need to keep the domains separate due to scalability reasons?
If not then rather than investing time into "Seamless MPLS" project invest
the time into migrating onto common IGP which will result in a much simpler
network (IGPs can support huge number of prefixes nowadays and  ISIS/OSPF
can compute SPF for important prefixes first -and do by default with
possibility to customize). 

If you are doing this to overcome potential scaling issues, which should be
the only reason for doing this, then the easiest and most scalable approach
is to leave the individual IGP domains truly separate and have only BGP
carry the necessary RIDs/PE-loopbacks along with their respective labels
between PEs and ABRs/ASBRs while forming the second (inter-area/inter-as)
level of transport labels, which in combination with complete separate
process of deriving intra-area/intra-as label-stack will produce the
complete end to end label switch path.
There's no redistribution involved in this kind of setup (hence no pollution
of local intra-area/intra-as IGP), basically clear separation of concerns
and all controls or "extras" are kept to the BGP-LU layer.
  
adam