[mpls] Re: Poll: IOAM and PSD

[Tony Li <[e-mail]>]

[WG chair hat: off]

Towards that end, I would like to propose:

Mutable data: data that changes while the packet is in transit.

Variable data: data that is constant while the packet is in transit, but may vary across packets in the same LSP.

T

On Aug 3, 2024, at 9:31 PM, Greg Mirsky <[e-mail]> wrote:



Hi Loa,

thank you for sharing your thoughts. Please find my notes below tagged GIM>>.

Regards,

Greg

On Sat, Aug 3, 2024 at 8:44 PM Loa Andersson <[e-mail]> wrote:

Greg,

I think Michael is right.

DEX is useful and we need it, but it does not solve all and any problem.

GIM>> Can you point me to where these "all and any problem" DEX is expected to solve? Or IOAM in general? 

I thought DEX was about transporting data/info generated by a node, and
outgoing packet.

GIM>> Perhaps some of my presentations, comments contributed to that misunderstanding. IOAM-DEX is not to transport data generated by the node but to separate the generation of the operational state and performance data from transporting them in the IOAM tagged packet that triggered the origination of data in the first place. IOAM-DEX allows operators use their preferred management plane mechanisms, e.g., IPFIX, protobuf, to collect data. Furthermore, raw data could be processed locally on the node that generated them and the results, e.g., mean, median, percentiles, exported.

This data is by defiintion immutable and is for example
never hashed for load sharing purposes

The issue with mutable data, e.g. the sequence number, is that it is in
an incoming packet and mutable.

GIM>> As defined in https://datatracker.ietf.org/doc/rfc9326/" rel="nofollow">RFC 9326, the Sequence Number is optional. And, as presented at the MPLS meeting in Vancouver, it, if needed, can use mutable 11 bits of LSE Format D. Hence, I think that the mutable concern, in my opinion, is addressed. 

If the network is doing ECMP by scanning
the label stack, then the sequence number can't in the first 20 bits of
an LSE.

GIM>> Yes, the authors of https://datatracker.ietf.org/doc/draft-mb-mpls-ioam-dex/" rel="nofollow">draft-mb-mpls-ioam-dex recognized that and propose the reasonable way to handle the situation.

Right?

/Loa

Den 04/08/2024 kl. 22:58, skrev Greg Mirsky:
> Hi Michael,
> I don't think that "IOAM requires mutable data" is an accurate
> statement. RFC 9326 IOAM Direct Export
> <https://datatracker.ietf.org/doc/rfc9326/" rel="noreferrer nofollow" target="_blank">https://datatracker.ietf.org/doc/rfc9326/> indeed defines optional
> parameters, Flow ID and Sequence Number, to ease the correlation
> process of collected telemetry data. For some applications, e.g.,
> DetNet over MPLS <https://datatracker.ietf.org/doc/rfc8964/" rel="noreferrer nofollow" target="_blank">https://datatracker.ietf.org/doc/rfc8964/>, both
> these parameters already are present as part of the DetNet MPLS
> encapsulation as S-label (Service) and Sequence Number in d-CW or
> d-ACH, respectively. Furthermore, the IPPM WG draft on collecting
> telemetry data
> <https://datatracker.ietf.org/doc/draft-ietf-ippm-hybrid-two-step/" rel="noreferrer nofollow" target="_blank">https://datatracker.ietf.org/doc/draft-ietf-ippm-hybrid-two-step/> can
> be used without the use of the optional IOAM-DEX fields, i.e., Flow-ID
> and Sequence Number.
>
> Regards,
> Greg
>
> On Sat, Aug 3, 2024 at 6:43 AM Michael Menth <[e-mail]>
> wrote:
>
>     Hi all,
>
>     we're not operators but researchers who implemented some IOAM
>     flavors in combination with SR using P4 on Tofino. We've presented
>     some results at an interim meeting before:
>     https://datatracker.ietf.org/meeting/interim-2024-mpls-01/session/mpls" rel="noreferrer nofollow" target="_blank">https://datatracker.ietf.org/meeting/interim-2024-mpls-01/session/mpls
>
>     IOAM requires mutable data. In combination with multiple labels,
>     e.g., for SR, these mutable data need to be near the bottom of the
>     stack, otherwise they get popped along the way. Moving them to PSD
>     makes encoding more efficient without increasing the effective
>     header length that needs to be parsed. Therefore, we think PSD is
>     the better option for applications like IOAM.
>
>     Regards
>
>     Michael
>
>     Am 30.07.2024 um 17:25 schrieb Tony Li:
>>
>>     [WG chair hat: on]
>>
>>     Hi all,
>>
>>     We’ve had many discussions about IOAM and PSD over the last few
>>     years. We need to reach consensus on the problems that need to be
>>     addressed in these areas. Therefore, we would like to hear from
>>     everyone, especially independent operators:
>>
>>     1.
>>
>>         There are many flavors of IOAM.  Which ones would you like to
>>         deploy/implement with MNA?
>>
>>     2.
>>
>>         Do you have other applications of MNA that have not been
>>         proposed yet?
>>
>>      This poll will close in two weeks, at 9am PDT, Aug 13.
>>
>>     Regards,
>>
>>     MPLS chairs
>>
>>
>>     _______________________________________________
>>     mpls mailing list [e-mail]
>>     To unsubscribe send an email [e-mail]
>
>     --
>     Prof. Dr. habil. Michael Menth
>     University of Tuebingen
>     Faculty of Science
>     Department of Computer Science
>     Chair of Communication Networks
>     Sand 13, 72076 Tuebingen, Germany
>     phone: [phone]
>     mailto:[e-mail]  <mailto:[e-mail]>
>     http://kn.cs.uni-tuebingen.de" rel="noreferrer nofollow" target="_blank">http://kn.cs.uni-tuebingen.de
>
>     _______________________________________________
>     mpls mailing list -- [e-mail]
>     To unsubscribe send an email to [e-mail]
>
>
> _______________________________________________
> mpls mailing list -- [e-mail]
> To unsubscribe send an email to [e-mail]

--
Loa Andersson
Senior MPLS Expert
Bronze Dragon Consulting
[e-mail]
[e-mail]

_______________________________________________
mpls mailing list -- [e-mail]
To unsubscribe send an email to [e-mail]