Re: [mpls] [EXTERNAL] Indicators in the stack and ancillary data after the BoS

[John E Drake <[e-mail]>]

Hi,

Comment inline.

Yours Irrespectively,

John



Juniper Business Use Only
From: mpls <[e-mail]> On Behalf Of Alexander Vainshtein
Sent: Thursday, June 17, 2021 5:07 AM
To: Stewart Bryant <[e-mail]>
Cc: [e-mail]
Subject: Re: [mpls] [EXTERNAL] Indicators in the stack and ancillary data after the BoS

[External Email. Be cautious of content]

Stewart,
I fully agree with your statement that "an old implementation that received a ToS GAL not at BoS would at best throw an exception or worst be unpredictable".

Regarding your statement "it is OK to have multiple GALs and GALs not at BoS IFF the creator of the LSP ensured that all LSRs on the LSP, including ECMP and FRR paths that found the GAL at ToS were known to be able to process it correctly":

  1.  I fully agree with this statement as a general restriction
  2.  Quite a lot of things have to be done in order to make this restriction work including at least:

     *   The definition of correct processing of GAL at ToS but not at BoS must be provided
     *   Advertisement of ability to process GAL not at BoS correctly in IGP and BGP must be defined
     *   Ability to set up network-wide paths that only cross nodes that process GAL correctly must be provided for different techniques (RSVP-TE, SR-TE, FlexAlgo. BGP-LU etc.)
It is still possible that, after all this work, we shall find out  that the benefits of supporting GAL at ToS but not BoS will be only available in the networks where all the nodes support the new functionality because presence of non-supporting nodes imposes too many restrictions on connectivity and/or resilience.

[JD]  As we start to add all sorts of nifty functions both within the stack and after it, we need to add these functions in such a way that a transit LSR that does not understand one or more of them still will be able to forward a packet containing them.  For example, if we were to add ancillary data in the label stack, it needs to be done in such a way as to allow a transit router to skip over the ancillary data that it doesn't understand.

I.e., incremental deployment is a fact of life and we have to live with it.

Regards,
Sasha

Office: [phone]
Cell:      [phone]
Email:   [e-mail]<mailto:[e-mail]>

From: Stewart Bryant <[e-mail]<mailto:[e-mail]>>
Sent: Thursday, June 17, 2021 10:36 AM
To: Alexander Vainshtein <[e-mail]<mailto:[e-mail]>>
Cc: Stewart Bryant <[e-mail]<mailto:[e-mail]>>; [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>
Subject: Re: [mpls] [EXTERNAL] Indicators in the stack and ancillary data after the BoS



On 17 Jun 2021, at 07:45, Alexander Vainshtein <[e-mail]<mailto:[e-mail]>> wrote:

While that might be the case, I think that the Open DT may give it a try and investigate how the existing systems will handle GAL being not the BoS label.
[[Sasha]] Great minds think alike! One useful step could be collecting the known actual behavior of popular implementations in this case, say, by running a survey among the vendors - what do you think?


That is actually a considerable amount of work that will take a while.

It seems to me that an old implementation that received a ToS GAL not at BoS would at best throw an exception or worst be unpredictable.

The original assumed processing model is to take the context of the PW label or PW+FAT label, discover the GAL and then process the GAL in the context of the PW label.

When we extended GAL to apply to LSPs we again had the model that the GAL operated in the context of the LSP label that preceded it for context. It was still BoS.

Putting the GAL further up the stack is a new behaviour.

If it arrives at an LSR that knows the new semantic all is good.

If it arrives at an LSR that does not know the new semantic then

a) An error has occurred either in setting up the LSP, or in forwarding.

b) The behaviour at the receiving node is unpredictable, but in any well written implementation should just result in the packet being dropped and counted as with any other Mal-formed packet.

So I would think that it is OK to have multiple GALs and GALs not at BoS IFF the creator of the LSP ensured that all LSRs on the LSP, including ECMP and FRR paths that found the GAL at ToS were known to be able to process it correctly.

A GAL not at BoS and not at ToS should not be inspected or processed by any LSR that did not know what it was doing, and to attempt to precess it would be a violation of the normal MPLS processing model.

- Stewart



Notice: This e-mail together with any attachments may contain information of Ribbon Communications Inc. and its Affiliates that is confidential and/or proprietary for the sole use of the intended recipient. Any review, disclosure, reliance or distribution by others or forwarding without express permission is strictly prohibited. If you are not the intended recipient, please notify the sender immediately and then delete all copies, including any attachments.