Skip to main content

Minutes for LMAP at interim-2016-lmap-3
minutes-interim-2016-lmap-3-1

Meeting Minutes Large-Scale Measurement of Broadband Performance (lmap) WG
Date and time 2016-06-13 07:00
Title Minutes for LMAP at interim-2016-lmap-3
State Active
Other versions plain text
Last updated 2016-06-20

minutes-interim-2016-lmap-3-1
LMAP WG Meeting
Date and Time: MONDAY 6/13/16, 1PM EDT

Meeting information:
http://www.ietf.org/mail-archive/web/lmap/current/msg02560.html

1. Note Well, Note Takers, Agenda Bashing - (Chairs, 5 min)

2. WG Status - (Chairs, 5 min)

3. IPPM relevant work Status - (Al, 5 min)

4. Open Issues: schedule events (Tim, 30 min)

5. YANG and in the information model (Juergen, 45 min)

6. Next steps, advancing documents to WGLC - 30 min

* Attendees

  - Al Morton
  - Alissa Cooper
  - Barbara Stark
  - Dan Romascanu
  - Godfred
  - Greg Mirsky
  - Jason Weil
  - Juergen Schoenwaelder
  - Tim Carey

* IPPM Update (Al Morton)

  - There is a new I-D draft-ietf-ippm-initial-registry-00.txt in IPPM.
  - The I-D draft-morton-lmap-examples was just for internal use.
  - The Performance Metrics Registry has addressed a key open issue, the Metric
  Naming Format, based on previous experience with projects like m-Plane and
  other measurement projects.
This format will be tested using the metrics in the initial registry contents
draft.

- There is new work to expand the IPPM Framework to include
IPv6 measurement details, which are more complicated than originally thought. 
There is also an effort to define the traceroute measurement as a registered
metric, since there are many parameters that affect the measurement outcome, in
addition to different procedural methods.

- Jason mentioned that IPPM co-chair Brian Trammell presented a summary of the
WG status at the European Internet Mapping project. Al added that Hiroshi Ota
presented Al's slides describing ITU-T SG 12 work on IP performance there as
part of the same agenda item.. There was an extended discussion and the
presentations were much appreciated and may influence the future of the project.

* Cycle-ID Discussion

  Cycle-ID = Cycle-Context-ID + Cycle-Number.

  - Al shows a slide from IETF 50 (IPPM)
  - What is the scope of the Cycle-Number? Al wants to have the
    Cycle-Number aligned across systems.
  - Al wants a simple global index into the results.
  - Al wants a Cycle-Number that includes the date and a bin number.
  - Alissa remarks that this can be calculated on both ends, the
    discussion seems to be whether there is an advantage to do it on the MA.
  - 6 digit of date YYYYMMDD + number (in GMT?)
  - Cycle number is a bin of N minutes starting from midnight.
  - Tim says one would have to fix the overall length and the bin length.
  - Al says that he wants the cycle ID calculated from the start time of the
  schedule (the time the event fired), not the start time of the actions. -
  Alissa suggest that it might help to have concrete text that says how this
  feature is configured, what the format of the Cycle-Number is, and how the MA
  calculates the Cycle-Number (from the event fire time). - Al to provide some
  concrete proposal.

  - Greg asks how are the Cycle-Context-IDs assigned?
  - Al responds that these names are assigned by the controller and the
    controller guarantees the uniqueness.

  - Tim: Should the Cycle-Context-ID be a tag or an option?
  - Al says that the tag should not be modified by the MA and simply be echoed
  back. - Both can be made to work. This was on the mailing list, I will try to
  bring it up again.

* LMAP Information Model Issues (Tim Carey)

  - Addition of ma-schedule-start and ma-schedule-end.
  - Tim wants to remove ma-schedule-start and ma-schedule-end.
  - Al asks if we run a one-off test, do we give up the capability to
    define the end time?
  - Tim says if we keep the duration, we should put it on the actions and not
  on the schedule. And it should be part of the options anyway. - Consider a
  passive iperf server that I start at a specific point in time and which I
  want to turn off at either another point in time or after a certain duration.
  Note that iperf does not have options to control this. - Al says this is like
  pressing crtl-c in an automated fashion.

  - Tim will bring this up on the list again.

* YANG Model and Information Model (Juergen Schoenwaelder)

  - skipped

* Next Steps (Chairs)

  - The plan is to have the issues (above) resolved quickly and to have I-Ds
  ready for WG last call beginning of July (at the submission deadline) so that
  we can issue the last call and discuss last call comments at IETF 96