Minutes for LMAP at IETF-91
minutes-91-lmap-1

LMAP WG meeting, Thursday, 9:00-11:00
Chairs: Dan Romascanu, Jason Weil
Notes: Barbara Stark, Marius Georgescu

Agenda:
http://tools.ietf.org/wg/lmap/agenda?item=agenda-91-lmap.html

Slides at http://tools.ietf.org/wg/lmap/agenda?item=agenda-91-lmap.html

No agenda bashing took place

Jason went on to WG status.

--- 11 minutes into meeting

Framework

Al Morton: [http://tools.ietf.org/agenda/91/slides/slides-91-lmap-4.pdf]

Benoit Claise: the passive and active, both active and passive are in scope,
recalling our discussion, by passive we want to observe flows and packets ...
Let's have a traffic filter monitoring.  I asked you two questions, do you want
to have passive in there? and do you understand the consequences of that... Al:
IPPM will expand the passive Mikael Abrahamson: Would like to see inclusion of
mention that passive measurements can also be done by the IP stack. Al: We
dont want to go into the details in framework. Mikael: Im a fan of not
injecting traffic but having the stack report performance. Mention it could be
in the IP stack. Jared March: Many already collect TCP performance, etc. Dont
want injecting data on my network just because someone wants to measure
something. Andrew MacGregor: The approach to do passive measurement would be ok
but it does have privacy issues. Alan Whinery: There is TCP instrumentation.
Al: So we can achieve better balance in discussing active and passive.
[proceeds presenting slides] Dan Romascanu: So we need to see new text for
passive before proceeding to next revision? Authors should make a proposal and
see if its ok. Benoit: I dont think this can be fixed by editorial changes.
Dan: Do we stop this doc and address or do things elsewhere (e.g., IPPM
charter)? Al: I think our figures are showing that were operating in a
different scope in the framework. Were happy to mention them but I dont think
we want to go into the details of sampling, etc. Barbara Stark: I do not
believe thorough discussion of active/passive belongs in framework. Benoit: OK.
Propose changes and we will discuss. Reading document with fresh mind perceives
bias towards active measurements. Al: OK

--- about 30 minutes into the meeting

Information Model

Trevor Burbridge: [presenting remotely]
[http://tools.ietf.org/agenda/91/slides/slides-91-lmap-1.pdf] [presented all
slides without interruption]

Tim Carey: Can you explain the reasons for moving channel configuration into
task configuration? what does that mean? Trevor: [Trevor presented various
examples of how this would work] The task configuration we had a set of
parameters. At that point we'll add the information about what channel to use.
Let's not make an exception for the channel configuration, and integrate them
in the task configuration. Tim: We had some relationships on how to configure
channels related to tasks. Trevor: Channels were initially only for getting
info to the collector. Tim: I just want to make sure I can hook it up to BBF
bulk data collector. Channels are obscured into options of task configuration
now. This may make things difficult. Phil Eardley: [remotely] Do you only
configure channel for task once or change it frequently? Trevor: The controller
task will take a channel. We can call that one task every hour, and it isnt in
individual measurement tasks. Dan: No more line at mic. Next steps? Trevor: I
will write these proposals to the list and see if anyone disagrees. Take a few
weeks for discussion. Dan: OK. Get feedback on proposals in slides and let
discussion happen over a couple of weeks. We will also look at calendar for
good date for another interim. This concludes discussion on information-model.

--- 1:04 hour into meeting

REST Style Large Measurement Platform Protocol

Lingli Deng:  [presented slides
http://tools.ietf.org/agenda/91/slides/slides-91-lmap-6.pdf] [showed demo on
laptop] Dan: Questions or comments? None. OK. Comments to proposals are
invited. We are getting close to the time when we need to make a recommendation
for what to propose as a WG solution. The editors of the drafts should help
each other, by critically reading the drafts and asking questions.

--- 1:18 hours into the meeting

YANG module for LMAP Controller

Arne Oslebo: [presented
http://tools.ietf.org/agenda/91/slides/slides-91-lmap-7.pdf] [explained
RESTCONF will be added to netconf-call-home] Dan: One of the next steps would
be a draft? Arne: Yes Tim: Question with RESTCONF. Are you suggesting this
would be a protocol between MA and Controller only, or also with Collector?
Arne: It could be with Collector too. Tim: I want consistency. Im concerned
with how RESTCONF would be used to report data. With Controller, would you use
the YANG interface model? There are many interfaces that arent standardized. I
need to be able to have all interfaces defined. Arne: Yes. Interfaces can be
defined. Mikael: Modeling in YANG is important. You can initiate work [to
create new interface models]. No need to sit around and wait. Tim: I just want
to make sure I can plug and play at the interface level. Arne: Yes.

--- 1:33 hours into meeting

Large MeAsurement Platform Protocol

Vaibhav Bajpai: [presenting remotely] [presented
http://www.ietf.org/proceedings/91/slides/slides-91-lmap-8.pdf]

Dan: [on Open Issues slide] Just introduce issues. We will continue discussion
on mail list. Vaibhav: [continues] Dan: People should now express opinions on
open issues. (1) Data modeling language. (2) Control Protocol. (3) Report
protocol. Will start separate thread for each question. Discuss now path
forward of WG. We need to organize an interim where we can focus more on
details of issues remaining for WG to determine. We should try to have 3 such
meetings: December, January, February. Questions? Comments on proposed path
forward? Tim: Youve accurately reflected work still to do. Information-model
still needs some work but attention is now turning to protocols. I don?t want
to work on 4 at once. The sooner we can pick one, the better. Agree that part
of work is to determine how we get to point of converging on a protocol. Dan:
We need criteria for protocol selection. It would be useful if someone who is
not an author of protocols could write criteria. This could be done just in
email. [Barbara and Tim volunteer]. Please contact Barbara and Tim if you also
want to participate. Jason: Note that report and control protocols don?t have
to be the same. If this is useful, it can be said. Dan: Timing. I suggest 3rd
week of Dec. We will set time and exact date by Doodle poll. Try for 3 hours
virtual meeting. Dan: Now for items not in charter. Its good for people to
bring in experiences and consideration for possible items for a next charter.

--- 2:00 hours into meeting

LMAP Internet Measurement

Motoyuki Ooki: [presented
http://tools.ietf.org/agenda/91/slides/slides-91-lmap-5.pdf]

Dan: Questions?
Michael Ackerman: do you get any resistance from trying to increase the number
of MAs. How many do you have? Motoyuki: 150 MAs and planning to enlarge MAs all
over Japan, distributed by our company. The plan is to add 50 MAs each year.
Michael: are the other ISPs supportive of this or pushing back ? Motoyuki:
pushing back Michael: Are you looking at passive data, or using traffic
injection as well? Motoyuki: All measurements are active measurement. Dan: are
the ISPs collaborating ?

"Did not get the name": As I understood the collectors are in the master
servers. Is that ok? Motoyuki: we are improving our system "Did not get the
name": very heavy duty on that servers isn't it. What is the main advantage of
increasing the number of MAs? Motoyuki: Main advantage is the lower workload.
In that case we can store many types of that. We would like to stoe many types
of data. It is good for us

Dan: Move discussion to the list. We have 2 more presentations.

--- 2:17 into meeting

Use-cases for Collaborative LMAP
Lingli Deng: [presented
http://tools.ietf.org/agenda/91/slides/slides-91-lmap-3.pdf]

-- 2:30 minutes into meeting

Kamala Subramaniam: [presented
http://tools.ietf.org/agenda/91/slides/slides-91-lmap-2.pdf]

Nalini Elkins: We're working on some things in IPPM. We should talk more about
this offline. Dan: this is a real problem. I'm not sure if it fits in the lmap
WG Kamala: it doesn't belong in IPPM, as we are not developing metrics. I made
some points about why it should belong in lmap. Dan: Maybe you should write an
information model or a data model in YANG for your next version.