Minutes IETF121: bmwg: Thu 09:30
minutes-121-bmwg-202411070930-00

Meeting Minutes Benchmarking Methodology (bmwg) WG
Date and time 2024-11-07 09:30
Title Minutes IETF121: bmwg: Thu 09:30
State Active
Other versions markdown
Last updated 2024-11-13

minutes-121-bmwg-202411070930-00

Benchmarking Methodology Working Group (bmwg) - IETF 121 Dublin

Date and Time: 07/11/2024 Thursday 09:30-11:30 UTC
Location: Liffey MR 3, Convention Centre Dublin
Chairs: Sarah Banks, Giuseppe Fioccola
AD: Warren Kumari
Zulip Stream: https://zulip.ietf.org/login/#narrow/stream/bmwg

Note taker(s)

Paolo volunteers to take the minutes.

Administrative and WG Status (Chairs)

Giuseppe chairs the session.

WG Drafts

Multiple Loss Ratio Search, Maciek Konstantynowicz and Vratko Polak

draft-ietf-bmwg-mlrsearch

Gabor: volunteering to review draft but need more time.
Guseppe: we'd like to start WG LC in the next weeks.

A YANG Data Model for Network Tester Management, Vladimir Vassilev

draft-ietf-bmwg-network-tester-cfg

Giuseppe: valuable work, is the model flexible to accodomate the coming
RFCs (MLRSearch, multiple IP addresses, etc.)?
Vladimir: yes, followed the MLRSearch work. It should be no problem to
accomodate, as it's a simple python script. We haven't tried yet. For
other things, the model can be augmented based on the specifics of a use
case. E.g. incremental changes to some fields to add a protocol specific
context. The goal is to keep the model simple.
Giuseppe: You have introduced new paramenters in the last version. Do
you need to add more?
Vladimir: the draft is becoming stable, we need users (and reviewers),
but best is users to get feedback.
Realistically, we can afford to wait one more year to have meaningful
feedback from real experience.
Vratko: useful front-end for traffic generators/analyzers. Possible to
pair with backend generators, e.g. T-REX, but integration requires
additional work. We see it in MLSSearch, we would need single front-end
to define traffic profile independent from back-end.
Vladimir: we have support for 1 commercial tester. We can do it for
other network testers like TREX which is a good option.

Considerations for Benchmarking Network Performance in Containerized Infrastructures, Minh-Ngoc Tran

draft-ietf-bmwg-containerized-infra

Giuseppe: thanks for the detailed explanation of the exchange with
Maryam. For WGLC we'll discuss it later on.

Benchmarking methodology for Segment Routing, Paolo Volpato

draft-ietf-bmwg-sr-bench-meth

Gabor: just add a sentence on security.
Carsten: we are late with providing some comments, but will do soon to
state the use of the methodology in public tests.

Proposals

Recommendations for using Multiple IP addresses in Benchmarking Tests, Gabor Lencse

draft-lencse-bmwg-multiple-ip-addresses

Giuseppe: anyone interested to provide reviews? Anyway will push it to
the list. Let's wait for more feedback.
Boris will review te draft (from the chat).

SRv6 Service Benchmarking Guideline, Xuesong Geng

draft-geng-bmwg-srv6-service-guideline

Gabor: do you put the network in lab or it's live?
Xuesong: in a lab. We'll give some explanations in the draft that it's
for a lab. We can also contribute to the SRv6 benchmarking draft.

Benchmarking Methodology for Source Address Validation, Libin Liu

draft-chen-bmwg-savnet-sav-benchmarking

Giuseppe: are you preparing the tool to do some hackathon for the next
IETF?
Libin: yes, at the next IETF.
Giuseppe: are you presenting in SAVNET?
Libin: yes, but we want to direct this draft to BMWG.
Giuseppe: please ask the people there to join the discussion on the BMWG
mailing list.

Calibration of Measured Time Values between Network Elements, Luis M. Contreras

draft-contreras-bmwg-calibration

Xuesong: do you think there is relationship to our SR E2E work?
Luis: this can be used by Flexalgo. The scope is to see if this can be
used for these applications without bias. Take decisions for TE, not
based on latency.
Xuesong: maybe latency is not so important for BE, but it can be
considered.
Luis: more valuable for very precise low-latency path.
Xuesong: understand, it's for low-latency cases. The scope is to
calculate the processing delay time.
Giuseppe: are you planning to do with hybrid methods, such as IOAM or
AltMark?
Luis: to be considered.
Paolo: do you plan to baseline against for example PTP?
Luis: no, but we can discuss.
Carsten: doubts on calibrating on fiber spool, it may be influenced by
the interface speed. Also, we use emulators at EANTC. if you have a
generic calibration methods then for fixed calibration I am not sure it
is going to work.
Luis: here the scope is to allow the control plane to take the best
decision for TE. The scope is not calibrating the hardware.
Carsten: ok, we used a calibrating generator and it was not so precise.

Luis: ok, we can take it into considertion.
Vladimir: I'm interested to contribute and review. The YANG model should
support implementing this measurement (from the chat)

Characterization and Benchmarking Methodology for Power in Networking Devices, Romain Jacob

draft-cprjgf-bmwg-powerbench

Giuseppe: suggest to update the draft for the next meeting.
Carsten: excellent, work much needed. I have comments that will be taken
offline.
Warren: fascinating, should see more of it the next time.
Giuseppe: continue the discussion on the list as people are interested.

Romain: should do loose benchmarking to easily run tests and aggregate
the results.

Presentations

AOB