Re: [aqm] Benoit Claise's Discuss on draft-ietf-aqm-eval-guidelines-11: (with DISCUSS and COMMENT)

[Mirja Kühlewind <[e-mail]>]

Hi Benoit,

I would not expect to see a bis doc here. As I said I would rather expect a 
new, separate document to specify and register the metrics (and providing 
more details). Given the very different scope and use case, I would not even 
see it as a problem if those two documents don't align fully.

I can go back to the authors and ask if they would be interested in a 
directorate review. But I guess that would also mean that they first have to 
change the text in section 2.X to the format/structure guidelines in RFC6390 
(sec 5.4.2 only in this case, I'd say), correct?

I still have to say that looking at RFC6390 again I find it not fully 
applicable for this case; e.g. Measurement Timing is not a great fit when you 
talk about a one time measurement (for one test run) of the flow completion 
time. I'd fear that using the template provided in RFC6390 would make it 
rather more confusing than clear.

Mirja

On 10.06.2016 09:36, Benoit Claise wrote:
> Thanks Al,
> This is exactly the type of feedback requested from the Performance Metric
> Directorate.
>
> Mirja,
> I buy into your argument:
>
>     I guess as soon as we have a registry, maybe there is someone interest in IPPM to catch up these metrics again and provide a RFC6390 definition but I would rather not like this document doing it.
>
> As such, we probably don't need the RFC6390 template in this document. Fair
> enough.
> However, what would be a pity is that if/when someone would like to register
> this metric, we end with questions such Al's one below, because the metric
> definitions in the published RFC are not precise. That could result in a RFCbis.
>
> Therefore, going through the Performance Metric Directorate feedback now is
> important IMO.
>
> Regards, Benoit
>> Because we are using Goodput, G, I take as given that there
>> must be a protocol with retransmission capability.
>> Otherwise, further simplification is possible (with dummy traffic).
>>
>> But yes, Fs and G need to be reported on payload
>> at the same layer, so the protocol layer chosen is
>> an input parameter for this metric.
>>
>> Al
>>
>>> -----Original Message-----
>>> From: Mirja Kuehlewind (IETF) [mailto:[e-mail]]
>>> Sent: Wednesday, June 08, 2016 7:21 AM
>>> To: MORTON, ALFRED C (AL)
>>> Cc:[e-mail];[e-mail]; The IESG; draft-ietf-aqm-
>>> [e-mail]; Benoit Claise;[e-mail]
>>> Subject: Re: [aqm] Benoit Claise's Discuss on draft-ietf-aqm-eval-
>>> guidelines-11: (with DISCUSS and COMMENT)
>>>
>>> Actually, it really doesn't matter that much in this case, I’d say. As
>>> we are talking about a lab environment, you might use dummy traffic that
>>> has some headers or not, that you might take into account of not, mostly
>>> depending on which information can be more easily accessed. What is
>>> important is that you do the same thing for all schemes that you
>>> compare.
>>>
>>> I guess one could add a note that there are different ways to measure
>>> this and that it is important to measure G at the same layer. Does that
>>> make sense?
>>>
>>> Mirja
>>>
>>>
>>>> Am 08.06.2016 um 13:03 schrieb MORTON, ALFRED C (AL)
>>> <[e-mail]>:
>>>> Here's one area which could use more detail:
>>>>
>>>>    ...The Flow Completion Time (FCT) is
>>>>    related to the flow size (Fs) and the goodput for the flow (G) as
>>>>    follows:
>>>>
>>>>    FCT [s] = Fs [Byte] / ( G [Bit/s] / 8 [Bit/Byte] )
>>>>
>>>> What protocol layers are included and excluded from Fs?
>>>>
>>>> Also, G needs to be measured at the same layer, and the
>>>> definition in RFC 2647 is a bit vague about layers, too.
>>>> It would be good to clarify which bytes to count here.
>>>>
>>>> Al
>>>>
>>>>> -----Original Message-----
>>>>> From: Mirja Kuehlewind (IETF) [mailto:[e-mail]]
>>>>> Sent: Wednesday, June 08, 2016 5:40 AM
>>>>> To: MORTON, ALFRED C (AL)
>>>>> Cc: Benoit Claise;[e-mail];[e-mail];
>>>>> [e-mail];[e-mail]; The IESG
>>>>> Subject: Re: [aqm] Benoit Claise's Discuss on draft-ietf-aqm-eval-
>>>>> guidelines-11: (with DISCUSS and COMMENT)
>>>>>
>>>>> Hi Al,
>>>>>
>>>>> what kind of detail are you looking for? Because I thought with the
>>>>> given equation this one was pretty clear.
>>>>>
>>>>> Do you have a reference to the benchmarking work?
>>>>>
>>>>> Mirja
>>>>>
>>>>>
>>>>>> Am 08.06.2016 um 11:18 schrieb MORTON, ALFRED C (AL)
>>>>> <[e-mail]>:
>>>>>> Hi Mirja,
>>>>>>
>>>>>> That sounds fairly reasonable to me.
>>>>>> Would it be possible to ask the authors provide a bit more
>>>>>> detail on Flow Completion Time?
>>>>>>
>>>>>>>>> Flow Completion Time is close to a definition for a new metric,
>>>>>>>>> and could benefit from more attention, perhaps a few more
>>> details.
>>>>>>>>> RFC6390 will provide some areas for improvement.
>>>>>> I imagine that related benchmarking efforts may wish to measure this
>>>>>> metric, and there would be independent implementations based on
>>>>>> the description provided here.
>>>>>>
>>>>>> regards from Geneve'
>>>>>> Al
>>>>>>
>>>>>>> -----Original Message-----
>>>>>>> From: Mirja Kuehlewind (IETF) [mailto:[e-mail]]
>>>>>>> Sent: Wednesday, June 08, 2016 4:46 AM
>>>>>>> To: MORTON, ALFRED C (AL); Benoit Claise
>>>>>>> Cc:[e-mail];[e-mail]; The IESG; draft-ietf-aqm-
>>> eval-
>>>>>>> [e-mail];[e-mail]
>>>>>>> Subject: Re: [aqm] Benoit Claise's Discuss on draft-ietf-aqm-eval-
>>>>>>> guidelines-11: (with DISCUSS and COMMENT)
>>>>>>>
>>>>>>> Hi Benoit,
>>>>>>>
>>>>>>> I finally had another look at the document as well as at RFC6390. I
>>>>>>> guess the metrics in question are Flow completion time (sec 2.1.),
>>>>> Flow
>>>>>>> start up time (2.2.) and Packet loss synchronization (2.4.). And
>>> you
>>>>> are
>>>>>>> right that these metric could be see as Application-Specific
>>>>> Performance
>>>>>>> Metric as defined in RFC6390. However, I agree with Al that given
>>> the
>>>>>>> scope of this document is providing
>>>>>>> "a generic list of scenarios against which an
>>>>>>>   AQM proposal should be evaluated, considering both potential
>>>>>>>   performance gain and safety of deployment.“,
>>>>>>> I don’t think these metric need to be defined and registered this
>>>>> way.
>>>>>>> I guess as soon as we have a registry, maybe there is someone
>>>>> interest
>>>>>>> in IPPM to catch up these metrics again and provide a RFC6390
>>>>> definition
>>>>>>> but I would rather not like this document doing it.
>>>>>>>
>>>>>>> Is that acceptable for you?
>>>>>>>
>>>>>>> We could add one more sentence in the abstract to make the scope on
>>>>> lab
>>>>>>> testing during development and before deployment clear. Would that
>>>>> help?
>>>>>>> Mirja
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>> Am 25.05.2016 um 14:22 schrieb Mirja Kuehlewind (IETF)
>>>>>>> <[e-mail]>:
>>>>>>>> Hi Al, Benoit, hi all,
>>>>>>>>
>>>>>>>> thanks for the feedback. Sorry for me delaying this maybe a little
>>>>> but
>>>>>>> I need t have another look at the document which will be next week
>>> at
>>>>>>> this point. In general I agree that this does not need to only rely
>>>>> on
>>>>>>> registered metrics because is mostly for lab tests; further this
>>>>> might
>>>>>>> probably not the right doc to register new metrics. However, I
>>> would
>>>>>>> still like to have another look at the doc and see if we can
>>> improve
>>>>>>> anything or figure out if any of the ’new’/non-registed metrics
>>>>>>> should/could be taken up by e.g. ippm.
>>>>>>>> Mirja
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>> Am 20.05.2016 um 14:53 schrieb MORTON, ALFRED C (AL)
>>>>>>> <[e-mail]>:
>>>>>>>>> All,
>>>>>>>>> a few replies in-line below,
>>>>>>>>> Al
>>>>>>>>>
>>>>>>>>>> -----Original Message-----
>>>>>>>>>> From: Benoit Claise [mailto:[e-mail]]
>>>>>>>>>> Sent: Thursday, May 19, 2016 5:38 AM
>>>>>>>>>> To: The IESG
>>>>>>>>>> Cc:[e-mail]; wes@mti-
>>> systems.com;
>>>>>>> aqm-
>>>>>>>>>> [e-mail];[e-mail];[e-mail]; linda
>>> Dunbar;
>>>>>>>>>> MORTON, ALFRED C (AL)
>>>>>>>>>> Subject: Benoit Claise's Discuss on draft-ietf-aqm-eval-
>>>>> guidelines-
>>>>>>> 11:
>>>>>>>>>> (with DISCUSS and COMMENT)
>>>>>>>>>>
>>>>>>>>>> Benoit Claise has entered the following ballot position for
>>>>>>>>>> draft-ietf-aqm-eval-guidelines-11: Discuss
>>>>>>>>>>
>>>>>>>>> ...
>>>>>>>>>> ----------------------------------------------------------------
>>> --
>>>>> --
>>>>>>> --
>>>>>>>>>> DISCUSS:
>>>>>>>>>> ----------------------------------------------------------------
>>> --
>>>>> --
>>>>>>> --
>>>>>>>>>> Has a RFC6390 performance directorate review done for the 2.X
>>>>>>> metrics?
>>>>>>>>>> It
>>>>>>>>>> should.
>>>>>>>>> [ACM]
>>>>>>>>> I reviewed this draft about 18 months ago.
>>>>>>>>> Mostly, it points to existing RFCs for fundamental metrics,
>>>>>>>>> and discusses others.  I read this:
>>>>>>>>> ...This document provides characterization guidelines that
>>>>>>>>> can be used to assess the deployability of an AQM, whether it is
>>>>>>>>> candidate for standardization at IETF or not.
>>>>>>>>> as restricted to lab testing.
>>>>>>>>>
>>>>>>>>>> Seehttp://www.ietf.org/iesg/directorate/performance-
>>> metrics.html
>>>>>>>>>> I guess that the metrics will be recorded in the future (See
>>>>>>>>>> draft-ietf-ippm-metric-registry-06
>>>>>>>>>> ), right?
>>>>>>>>> [ACM]
>>>>>>>>> That's up to the authors, they might simply point to
>>>>>>>>> metrics in the registry contributed by others
>>>>>>>>> (when following these guidelines at a future time).
>>>>>>>>>
>>>>>>>>>> For example, Flow Completion Time and Packet Loss
>>> Synchronization
>>>>>>> are
>>>>>>>>>> new, I believe.
>>>>>>>>> [ACM]
>>>>>>>>> Flow Completion Time is close to a definition for a new metric,
>>>>>>>>> and could benefit from more attention, perhaps a few more
>>> details.
>>>>>>>>> RFC6390 will provide some areas for improvement.
>>>>>>>>>
>>>>>>>>> Packet loss sync full methodology is described in [JAY006],
>>>>>>>>> according to the text.
>>>>>>>>>
>>>>>>>>>> And some other metrics are already documented in RFC6390
>>> compliant
>>>>>>>>>> documents. Pointers should be provided.
>>>>>>>>> [ACM]
>>>>>>>>> Most others are discussion sections and provide references.
>>>>>>>>>
>>>>>>>>>> See
>>>>>>>>>> https://tools.ietf.org/html/draft-ietf-xrblock-independent-
>>> burst-
>>>>>>> gap-
>>>>>>>>>> discard-01#appendix-A
>>>>>>>>>> for an example
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> ----------------------------------------------------------------
>>> --
>>>>> --
>>>>>>> --
>>>>>>>>>> COMMENT:
>>>>>>>>>> ----------------------------------------------------------------
>>> --
>>>>> --
>>>>>>> --
>>>>>>>>>> - Random Early Detection (RED), BLUE, and Proportional Integral
>>>>>>>>>> controller (PI)
>>>>>>>>>> Would you have references?
>>>>>>>>>>
>>>>>>>>>> - BDP is mentioned a few times. Please expand.
>>>>>>>>>>
>>>>>>>>>> - Glossary section = terminology section, right? If we want to
>>> be
>>>>>>>>>> consistent across documents
>>>>>>>>>>
>>>>>>>>>> - section 12.2. Why not a MUST below?
>>>>>>>>>> In order to understand an AQM's deployment considerations and
>>>>>>>>>> performance under a specific environment, AQM proposals SHOULD
>>>>>>>>>> describe the parameters that control the macroscopic AQM
>>> behavior,
>>>>>>>>>> and identify any parameters that require tuning to operational
>>>>>>>>>> conditions.
>>>>>>>>>>
>>>>>>>> _______________________________________________
>>>>>>>> aqm mailing list
>>>>>>>> [e-mail]
>>>>>>>> https://www.ietf.org/mailman/listinfo/aqm
>