Re: [mpls] MPLS_RT review of draft-mirsky-mpls-residence-time-06.txt

Yaakov and all,

To the best of my understanding all RTM packets will cross the same link in LAG that uses any "normal" hashing mechanism for labeled packets.

This is because these packets will never be misinterpreted as IP packets (due to GACh header) and will use the same label stack.

Maybe we should add the corresoonding clarification to the draft.

Thumb typed on my LG,
Sasha

------ Original message ------
From: Yaakov Stein
Date: 28/05/2015 11:41
To: Osborne, Eric;
Cc: mpls@ietf.org;
Subject:Re: [mpls] MPLS_RT review of draft-mirsky-mpls-residence-time-06.txt

> General question: does RTM work over LAGs?  They're a flavor of ECMP in some respects.  If not, that needs to be spelled out somewhere.

> It might be worth spending a few lines in the introduction talking about who might want to use this and what sort of restrictions it places on a network design.

All the caveats that were discussed in the context of the TICTOC draft apply here as well.

In particular, all packets of timing flows must follow the same path, thus ruling out LAG, ECMP, entropy labels, etc.
Protection switching must also be avoided, since undetected delay changes are more detrimental than not receiving timing packets and going into holdover.

Y(J)S

_______________________________________________
mpls mailing list
mpls@ietf.org
https://www.ietf.org/mailman/listinfo/mpls