Skip to main content

Minutes IETF111: bmwg

Meeting Minutes Benchmarking Methodology (bmwg) WG
Date and time 2021-07-26 19:00
Title Minutes IETF111: bmwg
State Active
Other versions plain text
Last updated 2021-09-08

BMWG Minutes
IETF 111

 - Over 27 people attended the session at the start.
 - Bill Cerveny took the minutes, a contribution for which the chairs are
 eternally grateful.

WG Status
 - RFC 9004, Updates for the Back-to-Back Frame Benchmark (Update to RFC2544)
   Published in May 2021.
 - Benchmarking Methodology for EVPN and PBB-EVPN
   status: IESG Review on July 1, 2021; Resolving DISCUSS and Comment ballots

   *Al: Milestones need to be updated.*

WG Drafts:

 - Multiple Loss Ratio Search
   - This version describes the new logic and examples on throughput in
   transactions per second, not just packets.  Some more to do in next ver.
   - Reviewers Requested!
    *Maciek Konstantynowicz presenting.*
    *Gábor Lencse and Vladimir Vassilev confirmed they will review mlrseach

- Benchmarking Methodology for Network Security Device Performance
   - Draft was updated several times since March 2021 meeting
   - WGLC on -08 ended in May
   - several extensive reviews and supportive comments
   - Draft was updated on May 21, with Sarah's LC comments yet to address
   - WG decision: "...and this document obsoletes [RFC3511]."

       *Al says there must be standalone line in abstract, "this document
       obsoletes RFC3511..."* *Brian Monkman presenting initial slides and is
       taking written comments* *Carsten Rossenhoevel reviews open comments and
       later slides.* *Comment 1/11: Sarah Banks responds that proposal makes
       sense to her. *Comment 2, Sarah proposes calling devices as
       active/inline versus "next generation". Alex Samonte said that he didn't
       know if NG implied inline or other state. Sarah likes description of
       firewall being a little broader. Alex: This doesn't cover a passive
       device. Sarah: There is a bit of confusion about use of term, "next
       generation" firewall. Comment 3/11:Sarah agrees in regards to "fail
       close" text. Comment 4/11: Alex: Two scenarios. In general, they are
       behind a router or switch. If test equipment, lots of MAC addresses, not
       like real environment. 2)  I like virtual test tools in the test bed.
       Jurrie van del Breekal: Impact of latency is a typically rounding error
       in testing. Sarah: Goal is to keep things simple as possible in the test
           Alex: SEC devices are normally behind a router or switch, reduces L2
           traffic. Also, need to do a B2B test to eliminate issues. Jurrie:
           Most customers who test use the switches in their testing. Sarah
           agrees. Al: Add some notes into the draft to cover this discussion.
       Comment 5/11: Sarah Offered to provide text to authors.
       Comment 6/11: Jurrie: Virtual routers are used to avoid overrunning MAC
       tables and for a realistic perspective -- it's more realistic to use
       virtual routers.
           Comment 7/11: to be resolved with comment Comment 5/11
       Comment 8/11: Agreement to remove NG IDS from draft. Sarah agrees this
       is resolved. Comment 9/11: Sarah to review and propose
       something. Amritan Putatunda commented  Working to remove test
       vendor-specific details now. Carsten: We need to include the TCP stack
       attributes Window sizes, etc.  -- Sarah: Need vendors tell you what
       parameters are. What happens to open source tools that don't have that
       info. Jurrie: Test defines key stack parameters that defines how TCP
       behavior. It's what the draft calls out and configure tests to
       parameters. Al: Need to define key parameter values for TCP in the draft
       (as agreed).
           Carsten: Even the need to use the same way to close TCP connections.
       Comment 10/11: Appeared to be agreement with author's proposal.
           Sarah: cover the switch config with generic configuration.
           Carsten: this was our goal - we believe we have already done this.
           Alex: yes, we have details, we don't ask for all details; the B2B
           test without SUT is the baseline for SUT evaluation.
       Comment 10/11: Sarah: If you can ask them to document details, then you
       can test apples to apples. Carsten: These details might be dangerous.
       Specifying details may force tester to use specific router/model. Sarah:
       As long as details are documented without specifying vendor/model, it
       shouldn't matter. Sarah: Has to go back and look at draft to clarify,
       but is interested in being able to replicate test. Alex: We don't go
       into that level of detail. We don't ask for that to be documented.
       Sarah: BMWG is big on repeatability. Erring on having details in regards
       to how that's the case. Alex: We want to get to that same point as much
       as possible, but want to avoid locking people out. Maybe we can reach a
       balance. Comment 11/11: Sarah concurs on proposed text. Al asks for any
       comments on this version of the draft. None noted. Al said he wanted to
       move this forward as fast as he can. There are a few more details and
       we'll go from there. Brian: Should we work with Sarah and go from there?
       Al: We should get the next steps done. Version 10 will be posted, Doc
       Shepherd will review, there may be some nits to fix.  Warren will review
       as AD, resolve his comments.  Then Area Review teams prepare comments
       during Last Call. Try to put on IESG telechat agenda in October. Authors
       need to reply to IESG comments - ASAP! Al describes remaining steps to
       publication. Warren Kumari suggests that authors send reminders to
       people in publication process.*
           Sarah: RFC Editor editing and AUTH48 takes about 12 weeks.
           Bill: Make the draft as clean grammar-wise, spelling-wise,
           Capitalization consistency-wise as you can. The fewer Questions from
           RFC Editor is best - leave them with less opportunities to
           misinterpret things. Everyone is happy.


 + A YANG Data Model for Network Interconnect Tester Management
   - This version adds capture "start-trigger" and "stop-trigger" configuration
   to the YANG model (a common feature for many traffic analyzer
   implementations) - Draft updated 11 July, 2021 - Vladimir has shared "Yet
   Another RFC2544 Testimplementation" open-source/hardware implementation with
   the BMWG. *Vladimir Vassilev presenting* *No feedback from attendees. Al:
   Your (Vladimir Vassilev) progress has been continuous and prolific. We might
   want to take it as a working group draft. Sarah agrees with making it a
   working group draft. Al: Will have a WG adoption call on the list, in
   August. *

NEW !!
 + Benchmarking Methodology for Stateful NATxy Gateways using RFC 4814
   Pseudorandom Port Numbers
   - ~5 people have discussed the draft on the list in May-July
   - Reference to procedures in RFC 8219
   - TCP-based testing = compliments the NG security device testing?
    *Gábor Lencse Presenting*
    *No feedback from attendees. Al: Will be productive to talk about this on
    the mailing list and we'll take it from there. Gábor: will attempt to
    upload new version of draft in August.*

 + AOB