Re: [Detnet] Magnus Westerlund's Discuss on draft-ietf-detnet-mpls-11: (with DISCUSS)

[Stewart Bryant <[e-mail]>]

I think it has to be dropped, because as the work stands I cannot see that we have a way of bounding the latency.

I know that we are talking about MPLS, but of course we need to look at all of the data plane drafts in this respect.

- Stewart



> On 9 Sep 2020, at 16:39, Lou Berger <[e-mail]> wrote:
> 
> The  doc currently reads (asterisks indicate the sentence under discussion):
> 
>    1.  Introduction
> 
>    Deterministic Networking (DetNet) is a service that can be offered by
>    a network to DetNet flows.  *DetNet provides these flows with
>    extremely low packet loss rates and assured maximum end-to-end
>    delivery latency.*  General background and concepts of DetNet can be
>    found in [RFC8655].
> 
> The sentence in question was copied from the draft version of RFC8655, which now reads slightly differently:
> 
>    ... which provides a capability for the delivery of
>    data flows with extremely low packet loss rates and bounded end-to-
>    end delivery latency.
> 
> I suggest either (a) updating the draft to match the RFC text or (b) dropping it altogether and let the reference to RFC8655 stand alone.
> 
> Lou
> 
> On 9/9/2020 11:20 AM, Toerless Eckert wrote:
>> On Wed, Sep 09, 2020 at 01:50:34PM +0000, Bal?zs Varga A wrote:
>>> Hi Toerless,
>>> 
>>> Many thanks for the comments. One remark:
>>> - I disagree with your statement "DetNet like any other IP/MPLS network with per-flow forwarding provides"
>>> Just as an example, PREOF functions are not available in current MPLS networks.
>> PREOF is not subject of the sentence part in question. My concern is only about:
>> 
>> ... DetNet provides zero congestion loss and bounded latency and jitter
>> 
>> Of course, now you mention it: The MPLS forwarding plane of this spec does
>> support PEROF, but the sentence only talks about "DetNet", for which at
>> large in my assesment this is not true (no current PREOF for IPv4/IPv6 AFAIK).
>> 
>> Aka: also for the part of PREOF its better to re-scope the sentence to talk only
>> the MPLS forwarding plane of this document instead of (unnecessarily?) make
>> claims about DetNet at large.
>> 
>> Cheers
>>     Toerless
>> 
>>> Thanks
>>> Bala'zs
>>> 
>>> -----Original Message-----
>>> From: Toerless Eckert <[e-mail]>
>>> Sent: Tuesday, September 8, 2020 9:13 PM
>>> To: Magnus Westerlund <[e-mail]>
>>> Cc: The IESG <[e-mail]>; [e-mail]; [e-mail]; [e-mail]; [e-mail]
>>> Subject: Re: [Detnet] Magnus Westerlund's Discuss on draft-ietf-detnet-mpls-11: (with DISCUSS)
>>> 
>>> Thanks Magnus, *:
>>> 
>>> Related to your comments, i would like to raise a concern about the initial sentence in the spec:
>>> 
>>> ...DetNet provides zero congestion loss and bounded latency and jitter.
>>> 
>>> To me, this is overselling what DetNet actually "provides" or that uniquely distinguishes DetNet from other solutions. It sounds as if DetNet provides a novel solution whereas in reality it just allows to adopt existing or new solutions.
>>> 
>>> With the definitions DetNet has done today, any IP or MPLS network where end-to-end flows can be identified as e.g.: an IP 5-tuple or an LSP identifier and that manages to figure out how to implement or operationalize one of the solutions for bounded latency such as a PHB in support of rfc2212.
>>> 
>>> Aka: one could equally write:
>>> 
>>> ...DetNet like any other IP/MPLS network with per-flow forwarding provides zero congestion loss and bounded latency and jitter.
>>> 
>>> Which would be equally true and equally misleading.
>>> 
>>> So, here is proposed IMHO more technically correct text to replace the IMHO misleading "marketing" sentence segment:
>>> 
>>> ...DetNet MPLS sets up point-to-point LSPs end-to-end across DetNet domains.
>>> 
>>> Because of this, DetNet MPLS can integrate with pre-existing and/or future Per-Hop-Behavior
>>> (PHB) (such one derived from RFC2212) that can provide per-flow (e.g.: LSP) bounded latency, bounded jitter and no congestion loss, as long as such a PHB does not require additional network packet header information beside the flow/LSP identification.
>>> 
>>> Cheers
>>>     Toerless
>>> 
>>> On Tue, Sep 08, 2020 at 08:09:21AM -0700, Magnus Westerlund via Datatracker wrote:
>>>> Magnus Westerlund has entered the following ballot position for
>>>> draft-ietf-detnet-mpls-11: Discuss
>>>> 
>>>> When responding, please keep the subject line intact and reply to all
>>>> email addresses included in the To and CC lines. (Feel free to cut
>>>> this introductory paragraph, however.)
>>>> 
>>>> 
>>>> Please refer to
>>>> https://www.ietf.org/iesg/statement/discuss-criteria.html
>>>> for more information about IESG DISCUSS and COMMENT positions.
>>>> 
>>>> 
>>>> The document, along with other ballot positions, can be found here:
>>>> https://datatracker.ietf.org/doc/draft-ietf-detnet-mpls/
>>>> 
>>>> 
>>>> 
>>>> ----------------------------------------------------------------------
>>>> DISCUSS:
>>>> ----------------------------------------------------------------------
>>>> 
>>>> I like to thank the TSV-ART reviewer for helping me consider one
>>>> aspect of the issue I see needing some discussion for this document.
>>>> 
>>>> This relates to Section 4.2.2.2. and 4.2.2.3.
>>>> 
>>>> So both of these section discuss the use of the sequence number for
>>>> removing packet duplicates and handling reorder. As the text discusses
>>>> there can be a configured limit for how deep the buffer and state are
>>>> for performing these operations. We all know that the implementation
>>>> of this will have a practical limit in both buffer space for
>>>> reordering as well as state for tracking which sequence numbers that
>>>> have been forwarded. I think that should be more clearly expressed in
>>>> the document that these practical limits exists. Thus, the
>>>> implementations will have tracking and determination of what are new packets (increasing sequence number within a window higher than previous largest seen.
>>>> And consider sequence number form currently highest seen and a bit
>>>> backwards as older packets. Thus how this is implemented will impact
>>>> how this acts in cases of disruptions of the packet flow. Thus, I
>>>> wonder if there is actually need to be  a bit more specific in how
>>>> classification should be done. Especially if the wrap-around of the
>>>> sequence number space approaches a small multiple of round trip times for the path which is likely for the 16-bit space.
>>>> 
>>>> Then  sections fails to discuss how the duplication removal, the
>>>> reordering buffering and bound latency interacts and affet each other.
>>>> So if the latency is bounded then the reordering has an hard time
>>>> limit for the maximum delay. If there is a boundary for reordering
>>>> then there are no point in de-duplicating packets that will not be
>>>> forwarded due to the reordering. And even if there are no bounded
>>>> latency the reordering buffer size will still impact the depth of
>>>> de-duplication. These practical limits will also be limitations on the guarantees that can be provided.
>>>> 
>>>> Thus, from my perspective there is need for more text on the
>>>> requirements of the implementation of these functions and their
>>>> interactions of creating limitations.
>>>> 
>>>> Another point on 4.2.2.2:
>>>> 
>>>> When configured, the
>>>>    implementation MUST track the sequence number contained in received
>>>>    d-CWs and MUST ensure that duplicate (replicated) instances of a
>>>>    particular sequence number are discarded.
>>>> 
>>>> That second MUST I think is possible to meet given that one discard
>>>> all packets outside of the current window where one have information
>>>> if a packet sequence number have been forwarded or not. Given that a
>>>> very late packet beyond the amount of state for the flow likely anyway
>>>> have little utility that is likely the right choice. However, I think
>>>> it needs to be made explicit that this is okay.
>>>> 
>>>> In Section 4.2.2.3:
>>>> 
>>>>  When configured, the
>>>>    implementation MUST track the sequence number contained in received
>>>>    d-CWs and MUST ensure that packets are processed in the order
>>>>    indicated in the received d-CW sequence number field, which may not
>>>>    be in the order the packets are received.
>>>> 
>>>> I think this part needs to be explicit that packets that are to fare
>>>> out of order for the implementation to handle will/shall be dropped.
>>>> 
>>>>    Note that an implementation MAY wish to constrain the maximum number
>>>>    of out of order packets that can be processed, on platform-wide or
>>>>    per flow basis.  Some implementations MAY support the provisioning of
>>>>    this number on either a platform-wide or per flow basis.  The number
>>>>    of out of order packets that can be processed also impacts the
>>>>    latency of a flow.
>>>> 
>>>> If there exists a latency requirement then that will interact with
>>>> this when it comes to reordering. In fact a significant issue here is
>>>> that if the packet flow is not periodic at a steady pace the maximum
>>>> latency in the reordering buffers based on packet sequence numbers can
>>>> not be ensured. Instead some form of time limit needs to exist also.
>>>> If that time limit is only local then there exists a risk that over
>>>> multiple reordering buffers if multiple independent service labels are
>>>> used the jitter and latency becomes cumulative. If the goal is to
>>>> avoid this then the individual packets would need to carry a time
>>>> stamp to ensure that from ingress of the service label path until the egress a maximum latency is added.
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> _______________________________________________
>>>> detnet mailing list
>>>> [e-mail]
>>>> https://www.ietf.org/mailman/listinfo/detnet
>>> --
>>> ---
>>> [e-mail]