RE: Request for RTGWG Working Group adoption for draft-bashandy-rtgwg-segment-routing-ti-lfa

[Alexander Vainshtein <[e-mail]>]

Stephane,
I fully agree that doing TI-LFA with just Adj-SIDs is theoretically possible but not practical given existing HW and realistic network scales.

I also think that one big advantage of this draft is that it provides a set of implementable solutions that cover a wide spectrum of real life scenarios, and this is done by offering reasonable strategies for minimization of the label stack depth used by the repair path.

It is, of course, up to the authors to decide whether such optimization is in or out of scope of the draft.

But if it left out of scope, practical value of the draft will suffer – especially since quite a few implementations of (at least, some of) these optimization strategies are already implemented and, AFAIK, deployed.

My 2c.
Sasha

Office: [phone]
Cell:      [phone]
Email:   [e-mail]

From: [e-mail] [mailto:[e-mail]]
Sent: Thursday, July 12, 2018 4:02 PM
To: DECRAENE Bruno IMT/OLN <[e-mail]>; Alexander Vainshtein <[e-mail]>
Cc: [e-mail]; [e-mail]; [e-mail]; [e-mail]; [e-mail]; Ahmed Bashandy <[e-mail]>; Robert Raszuk <[e-mail]>; Chris Bowers <[e-mail]>; Stewart Bryant <[e-mail]>
Subject: RE: Request for RTGWG Working Group adoption for draft-bashandy-rtgwg-segment-routing-ti-lfa

Hi,

From a pure theoretical perspective, nothing prevents to do TILFA with Adj-SIDs only and this requires only the postfailure SPF to compute. From an implementation point of view, I would agree with everyone that this may cause some issues for existing HWs ☺, so optimizing the stack size is required. I do not think that the optimization algorithm is far from an SRTE algorithm which also tries to optimize the label stack size. So the optimization could be seen as outside of TILFA specification.

However, where I could agree with Sasha is that the current draft tries to propose an optimization of the label stack size as it tells about intersection of P space and Q space without giving the full details and especially the potential scaling points that implementations need to take care of.

I think what we need to agree on is if the label stack optimization is in scope of the draft or out of scope. If in scope, do we standardize one option, do we just give some guidelines/clues with pros/cons ?


Brgds,



From: [e-mail]<mailto:[e-mail]> [mailto:[e-mail]]
Sent: Thursday, July 12, 2018 14:35
To: Alexander Vainshtein
Cc: [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>; Ahmed Bashandy; Robert Raszuk; Chris Bowers; Stewart Bryant
Subject: RE: Request for RTGWG Working Group adoption for draft-bashandy-rtgwg-segment-routing-ti-lfa

Sasha,

Please see inline [Bruno]

From: Alexander Vainshtein [mailto:[e-mail]]
Sent: Thursday, July 12, 2018 1:59 PM
To: DECRAENE Bruno IMT/OLN
Cc: [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>; Ahmed Bashandy; Robert Raszuk; Chris Bowers; Stewart Bryant
Subject: RE: Request for RTGWG Working Group adoption for draft-bashandy-rtgwg-segment-routing-ti-lfa

Bruno,
The other issue I’ve raised with regard to usage of post-convergence paths is scalability.

The draft says (in section 3.2):

   We want to determine which nodes on the post-convergence path from
   the PLR to the destination D are in the Q-Space of destination D
   w.r.t. link S-F.
   This can be found by intersecting the post-convergence path to D,
   assuming the failure of S-F, with Q(D, S-F).
My reading of this text says that  Q(D, S-F) MUST be computed for each destination D that is affected by failure of link S-F (I am not aware of any other method – do I miss something?).  And this is exactly what RFC 7490 warns against in Section 5.2.1:
   Note that the Q-space calculation could be conducted for each
   individual destination and a per-destination repair tunnel end point
   determined.  However, this would, in the worst case, require an SPF
   computation per destination that is not currently considered to be
   scalable.

“Currently” in 7490 presumably refers to 2015, but I am not aware of drastic improvements in the computational power of the CPUs of modern routers that allow computation of hundreds of reverse SPF computations following every topology change.

Again, did I miss something?

[Bruno] First, as you quoted, your comment applies to RLFA. And in general, the answer is to restrict yourself to the Q space of E with respect to the link S-E.
Second,
- if you limit yourself to link protection, especially in network with symmetric link cost, some shortcut may be found. One could say it’s an implementation issue. One could argue that the draft should at least describe one algo, while stating that this choice is non-normative.
- if you want node/SRLG protection [RFC 8102] and/or LFA manageability [RFC7916] to pick an acceptable path, it’s not clear to me that the alternatives scale better. TI-LFA has the benefit of computing the alternate path in one easy step (1 SPF), leaving the computation cost to the minimization of the list of segments when needed (and this is rarely needed, especially for link protection). It’s not clear to me that computing 10s of RLFA and then trying to pick the “best” / an acceptable one (e.g. by computing a SPF rooted on the PQ of each candidate) is a better way. But we don’t even need to compare as this did not stop the publication of RFC 8102.

Regards,
--Bruno

Regards,
Sasha

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

From: Alexander Vainshtein
Sent: Thursday, July 12, 2018 2:26 PM
To: '[e-mail]' <[e-mail]<mailto:[e-mail]>>
Cc: [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>; Ahmed Bashandy <[e-mail]<mailto:[e-mail]>>; Robert Raszuk <[e-mail]<mailto:[e-mail]>>; Chris Bowers <[e-mail]<mailto:[e-mail]>>; Stewart Bryant <[e-mail]<mailto:[e-mail]>>
Subject: RE: Request for RTGWG Working Group adoption for draft-bashandy-rtgwg-segment-routing-ti-lfa

Bruno,
It seems there is some misunderstanding, and I will try to clarify it.

To the best of my understanding, the following text in Section 1 of the draft presents the benefits of using post-convergence path for FRR:

   As the capacity of the post-convergence path is typically planned by
   the operator to support the post-convergence routing of the traffic
   for any expected failure, there is much less need for the operator
   to tune the decision among which protection path to choose.  The
   protection path will automatically follow the natural backup path
   that would be used after local convergence.  This also helps to
   reduce the amount of path changes and hence service transients: one
   transition (pre-convergence to post-convergence) instead of two
   (pre-convergence to FRR and then post-convergence).

I see two different claims of benefits from using post-convergence path in this test fragment

1.       One path change and therefore one service transient instead of two

2.       Post-convergence path is taken into account in the operator’s panning (e.g., by allocating sufficient resources for traffic flows on both pre-convergence and post-convergence paths).


Speaking just for myself, I think that neither of these claims is justified for traffic flows that do not originate at the PLR.

E.g., consider Stewart’s example and the traffic flow from A to E

1.       This flow will experience two path changes (pre-convergence--> FRR and FRR --> post-convergence

2.       The network operator will not take links C-F, F-G and G-D for consideration in its planning of pre-convergence and post-convergence paths for this flow.

Did I miss something substantial?

Regards,
Sasha

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

From: [e-mail]<mailto:[e-mail]> [mailto:[e-mail]]
Sent: Thursday, July 12, 2018 12:49 PM
To: Stewart Bryant <[e-mail]<mailto:[e-mail]>>
Cc: [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>; Ahmed Bashandy <[e-mail]<mailto:[e-mail]>>; Alexander Vainshtein <[e-mail]<mailto:[e-mail]>>; Robert Raszuk <[e-mail]<mailto:[e-mail]>>; Chris Bowers <[e-mail]<mailto:[e-mail]>>
Subject: RE: Request for RTGWG Working Group adoption for draft-bashandy-rtgwg-segment-routing-ti-lfa

Stewart,

Please see 1 comment inline [Bruno]
Trimming the text to ease the focus on this point

From: Stewart Bryant [mailto:[e-mail]]
Sent: Tuesday, July 10, 2018 2:40 PM

On 09/07/2018 20:53, Ahmed Bashandy wrote:
[…]

b.       Selecting the post-convergence path (inheritance from draft-francois) does not provide for any benefits for traffic that will not pass via the PLR after convergence.

                                                               i.      The authors claim to have addressed this issue by stating that “Protection applies to traffic which traverses the Point of Local Repair (PLR). Traffic which does NOT traverse the PLR remains unaffected.”

SB> It is not as simple as that, and I think that the draft needs to provide greater clarity.

I think there will be better examples, but consider

              12
      +--------------+
      |              |
A-----B-----C---//---D----E
        10  |        |
            F--------G

Traffic injected at C will initially go C-D-E at cost 2, will be repaired C-F-G-D-E at cost 4 and will remain on that path post convergence. This congruence of path is what TI-LFA claims.

However, a long standing concern about traffic starting further back in the network needs to be more clearly addressed in the draft to clearly demonstrate the scope of applicability.

For traffic starting at A, before failure the path is A-B-C-D-E cost 13

TI-LFA will repair to make the path A-B-C-F-G-D-E cost 15 because TI-LFA optimises based on local repairs computed at C.

After repair the path will be A-B-D-E cost 14.

[Bruno] The draft is about IP Fast ReRoute (FRR).
FRR is a local reaction to failure, so by hypothesis, all nodes but the PLR are not aware about the failure. This includes all upstream nodes which do keep forwarding traffic through the same path, i.e. via the PLR.
The argument that the path would have been shorter if upstream node were aware of the failure to reroute before (or that the PLR should send the packet back in time) is not relevant.
The only question which matter is: from the PLR to the destination, which is the best path to use?
I, and the draft, argue that the best path in IP routing, is the IGP shortest path. Whichever type of metric you choose (e.g. bandwidth, latency, cost…). Do you disagree on this?


Now, eventually we can narrow down the discussion to the choice of terms. We can discuss about the term “post-convergence paths from the point of local repair », which you don’t think to like. Although, the term seems technically true to me, I would also be fine with changing from  “post-convergence path” to “optimal IGP shortest path”



So the draft needs to make it clear to the reader that TI-LFA only provides benefit to traffic which traverses the PLR before and after failure.

[Bruno] No, that is not true. cf above.
--Bruno

Traffic which does not pass through the PLR after the failure will need to be traffic engineered separately from traffic that passes though the PLR in both cases.



_________________________________________________________________________________________________________________________



Ce message et ses pieces jointes peuvent contenir des informations confidentielles ou privilegiees et ne doivent donc

pas etre diffuses, exploites ou copies sans autorisation. Si vous avez recu ce message par erreur, veuillez le signaler

a l'expediteur et le detruire ainsi que les pieces jointes. Les messages electroniques etant susceptibles d'alteration,

Orange decline toute responsabilite si ce message a ete altere, deforme ou falsifie. Merci.



This message and its attachments may contain confidential or privileged information that may be protected by law;

they should not be distributed, used or copied without authorisation.

If you have received this email in error, please notify the sender and delete this message and its attachments.

As emails may be altered, Orange is not liable for messages that have been modified, changed or falsified.

Thank you.

___________________________________________________________________________

This e-mail message is intended for the recipient only and contains information which is
CONFIDENTIAL and which may be proprietary to ECI Telecom. If you have received this
transmission in error, please inform us by e-mail, phone or fax, and then delete the original
and all copies thereof.
___________________________________________________________________________

_________________________________________________________________________________________________________________________



Ce message et ses pieces jointes peuvent contenir des informations confidentielles ou privilegiees et ne doivent donc

pas etre diffuses, exploites ou copies sans autorisation. Si vous avez recu ce message par erreur, veuillez le signaler

a l'expediteur et le detruire ainsi que les pieces jointes. Les messages electroniques etant susceptibles d'alteration,

Orange decline toute responsabilite si ce message a ete altere, deforme ou falsifie. Merci.



This message and its attachments may contain confidential or privileged information that may be protected by law;

they should not be distributed, used or copied without authorisation.

If you have received this email in error, please notify the sender and delete this message and its attachments.

As emails may be altered, Orange is not liable for messages that have been modified, changed or falsified.

Thank you.

_________________________________________________________________________________________________________________________



Ce message et ses pieces jointes peuvent contenir des informations confidentielles ou privilegiees et ne doivent donc

pas etre diffuses, exploites ou copies sans autorisation. Si vous avez recu ce message par erreur, veuillez le signaler

a l'expediteur et le detruire ainsi que les pieces jointes. Les messages electroniques etant susceptibles d'alteration,

Orange decline toute responsabilite si ce message a ete altere, deforme ou falsifie. Merci.



This message and its attachments may contain confidential or privileged information that may be protected by law;

they should not be distributed, used or copied without authorisation.

If you have received this email in error, please notify the sender and delete this message and its attachments.

As emails may be altered, Orange is not liable for messages that have been modified, changed or falsified.

Thank you.

___________________________________________________________________________

This e-mail message is intended for the recipient only and contains information which is 
CONFIDENTIAL and which may be proprietary to ECI Telecom. If you have received this 
transmission in error, please inform us by e-mail, phone or fax, and then delete the original 
and all copies thereof.
___________________________________________________________________________