Minutes for DETNET at IETF-93
minutes-93-detnet-1

DetNet BoF Minutes For IETF 93
>
>  Assigned minute taker: Diego Dujovne
>
> MONDAY, July 20, 2015
> 1300-1500  Afternoon Session I
> Location:        Grand Hilton Ballroom
>
> Presentation        Start Time        Information
>      0              13:00              Title:           Administrivia
& Intro
>                 Presenter:             Chairs
>                 Slides:
https://www.ietf.org/proceedings/93/slides/slides-93-detnet-0.pptx
>                 Slides:
https://www.ietf.org/proceedings/93/slides/slides-93-detnet-0.pdf
(pdf links are broken, use .pptx)

Lou Berger and Pat Thaler chair the BoF
Pat presents opening slides to set the context for the BOF

>      1              13:20              Title:        User's view
>                 Presenter:        Jouni Korhonen, Pascal Thubert,
Craig Gunther, Patrick Wetterwald, Subir Das
>                 Slides:
https://www.ietf.org/proceedings/93/slides/slides-93-detnet-1.pptx

>                 Slides:
https://www.ietf.org/proceedings/93/slides/slides-93-detnet-1.pdf

Jouni Presentes slides.
Mary Barnes asks for the pdf of the slides.
Craig continues: (A/V applications) Example: Live show audio
reproduction requirements (strict requirement)
recording: Soft requirements: Sync, but not realtime.
Example: Key adjustment for singers
Multiple streams of audio on the same cable.
Interconnection between TV studios; e.g. 46Tbps of data on fiber.
Intserv (local) and Diffserv (wide) in
Ralph Droms: Question: Configuration Timeframe for this application?
Craig Gunther: Base configuration and dynamic behavior on top

Patrick continues with the slides: Electric Grid reliability
 Symmetric latency on the line
 Optimize telecom network: OT/IT integrated network, delay restrictions
giventhe distances
 Current solutions: TDM, no mixing of IT/OT
 New requirements from: IEC61850
 The utility controls everything (can live with distributed but they
like centralized)

 Lou Berger: these slides only describe the presenters point of view,
Not chartered work items

 Subir Das continues with the Building Automation Systems
 Different protocols on field networks and local controllers, lack of
security on protocols

 Pascal continues with the slides: Industrial applications.
 Points out the relevance of wireless on deterministic networks.
 Applications for large scale monitoring
 Mix deterministic and stochastic traffic; manage them as independent flows.
 Virtualized operational technology to guarantee transmission from
sensor through backbone to
 the actuator somewhere else.

Use case is Industrial Internet. Industrial Networks have been deterministic
for 40 years and deterministic Wireless appeared 10 years ago.
But still silo'ed and cannot benefit from IT technology. Goal is now to lower
cost and improve ROI of networks.
Cost can be lowered by converging networks with IT to reduce OPEX. ROI
can be improved by augmenting efficiency of processes using large scale
(stochastic) monitoring and big data.


 Jouni continues with the presentation.
 Application: Connectivity between Remote radios and baseband processing
units. Also connectivity between base stations, and between core networks
and the base stations.
Next generation systems have 10-times more strict end to end delay
requirements than before/current systems
Summarizes the most relevant points:
    - Time sensitiveness,
    - Open standards
    - Mix of different types of traffic

Lou Berger: asks for comments/questions on the user presentations

John Dowdel: deterministic ethernet applications. Some IPR already
present in the aeronautic field
Lou Berger: The IETF has rules related to IPR disclosure, all
contributors to IETF process are expected to follow these rules.
Adrian Farrel: It sounds like there is some disagreement between the
presented requirements. Connection Running IP over determinstic layer 2,
or interconnect L2 islands
Lou Berger: You pick up on there being two different scenarios being
requested.  It is for the charter/WG do discuss which or both are
covered and what order to address these.
...? :Which of these different use cases / scenarios are in scope?
Lou Berger: Please re ask question as part of the scope discussion.

>      2              14:00              Title:        DetNet Problem
Statement (including example possible/plausible solutions)
>                 Presenter:        Norman Finn
>                 Slides:
https://www.ietf.org/proceedings/93/slides/slides-93-detnet-2.pptx
>                 Slides:
https://www.ietf.org/proceedings/93/slides/slides-93-detnet-2.pdf

Norm starts presentation.
Main tasks:
- Eliminate congestion generated packet loss (reserve resources) using
max buffer allocation
- Seamless redundancy, eliminate duplicates in intermediate routers
Questions:
...?: Once organized suggest collaborate with research two
functionalities information delivery aggregation... other application
with dedicated BW underneath. it may not achieve both
together. Mixture of the traffic(?) already done?
That is a question for the group
David Black: Why a single encapsulation?
Lou Berger: Scope discussion
Pascal: there are different type of applications that may or may not
need all the features. Some of the requirements (10 to the minus 12) may
not fit to the wireless, but it may improve wireless reliability.

>      3              14:20              Title:        Working group
scope and deliverables discussion
>                 Presenter:        Chair moderated
>                 Slides:
https://www.ietf.org/proceedings/93/slides/slides-93-detnet-3.pptx
>                 Slides:
https://www.ietf.org/proceedings/93/slides/slides-93-detnet-3.pdf

Lou Berger:presents the scope.
Erik Nordmark: Control plane?
Lou: The control plane is intended to be a superset of distributed
control plane and centralized controllers

 >               14:50              Title:        Concluding discussion
& polling
>                 Presenter:        Chairs

James Loewe: Environment from use cases: solutions to be provided for
the mentioned aspects. Will applied to support these use vertical cases
or the solutions as systems. Will be cross domain (different operators
managing the networks for a single domain? requirements single domain?
Lou Berger: Use cases to support: Breakdown: interconnected islands and
end-to-end. Users say they like this in scope, they may be building a
product on these.
Erik Nordmark: Sequency? Deliver something useful to start from a single
application, and add another use case...?
Lou Berger:That may fall on the second case
Ralph Droms: Sense of the amount of work
Lou Berger: 18-24 months. Reasonable time?
Ralph Droms: Optimistic. Base this on milestones.
Norman Finn: Techonologies for all the pieces. It won' t be exactly for
the applications, but it would be for the main scope
Pat: Multiple domains might not be achievable in 24 months. Single domain
should be.
Lou: Depending on the techonology we piggyback on, we will need
cross-domain or not.
Subir Das: Which is the best bet, specifically on packet loss. Tested on
wireless or wireline?
Pat: Should be wire/wireless agnostic. Address both, even though the differences
We define what is going on the router level: interfaces.
Subir Das: We need mapping information about the low-level hardware used.
Meeting reasonable timelines should be part of the initial scope
independently of the
Control plane no specific delivery but data flow information model. can
it be included there?
Lou: on the architecture deliverable
Ralph Droms: Solutions space: considerations to interact with the IEEE,
no references on the description.
Pat: queueing, forwarding and policy managing, coordination with IEEE?
if there is something not specificed, we must collaborate with IEEE
Norman: In an application we don't care wether is using a specific cable
or going through a bridge or a satellite link. it is an IP application.
My PC should not care, only the service oriented, request the service
from the router.

Lou Berger: asks for the feeling on the room
Is this a problem that should be solved (TSN over routers)? (say about 50%)
Should we, the IETF work on this problem (about the same)
Who thinks the stated deliverables can be completed in a reasonable
time*?  (say about 25%)

Andrew Sullivan? ok problem, not reasonable time?

Acee Lindem: that would be a surprise to have this on 24 months

Deborah's point: only the deliverables on the slide

David Black: there is no nail on the floor saying we are going to do THIS.
Not boil the ocean

Subir Das: Reasonable time: problem-solution
Lou: data plane, architecture documents close to the IESG
Kireeti Kompella: whatever you do, dont break my hardware.
Alia Atlas: On reasonable time, focus on interoperable standards and
getting the solvable pieces done

Who is willing to author, review and otherwise contribute to a DetNet WG?
(About the same number, say 20-30)

Deborah Brungard: Asks for feedback, even privately. New to determine
level of interest and if doable in reasonable time.

Lou Berger: Comment on the ML for the group proposed charter.
> Adjourn              15:00