Minutes IETF107: raw
minutes-107-raw-02

**********************************************************************
          Agenda & Minutes
**********************************************************************

  20:00  Intro                Chairs                                           
              [10mn] 20:10  Use cases      Carlos Bernardos                    
                    [20mn]
                                        draft-bernardos-raw-use-cases-03
  20:30  LDACS           Nils Maeurer                                          
      [20mn]
                                        draft-maeurer-raw-ldacs-01
  20:50  5G                  Janos Farkas                                      
         [20mn]
                                        draft-farkas-5g-raw-00
  21:10                        Georgios Papadopoulos
             Pareo                 draft-papadopoulos-raw-pareo-reqs-01  [15mn]
             OAM                  draft-theoleyre-raw-oam-support-01      
             [10mn]
  21:35  Discussion    Chairs                                                  
         [25mn]

Additional RAW drafts:
    draft-pthubert-raw-problem-statement-04
    draft-thubert-raw-technologies-04

* Introduction
(Chairs)

Eve introduces the first non F2F meeting with history of past meetings. Eve
presents herself, was cochair of mmusic. Eve is also part of the IOT
directorate, interested in QOS

Motivation for RAW: companion WG to DetNet. Going from probabilistic to
Deterministic DetNet was missing use cases related to wireless. Also gaps in
MANET.

Eve lists typical DetNet properties, including bounded latency; RAW examinizes
wireless technologies that can be scheduled. Eve introduces the charter, going
through it. Short term 12 to 18 months, informational RFCs for the first run

Use cases articulated around wireless; by Year End have an evaluation of
existing IETF technologies and a gap analysis

Rick discusses the document workflow at the IETF. RAW is to be short lived, we
need to make sure we get the docs finished off promptly.

RAW use case
-------------------
Carlos Bernardos presents v -03
The named evolved with the tentative WG name; documents 7 use cases.
Presentation focusing on industrial application and gaming taken as example due
to limited time. Identifying the distinguished properties for RAW.

Carlos introduces the plant connectivity with sensors, actuators and PLCs.
Specifics:
    * heterogeneous technologies
    * multiple simultaneous links for reliability
    * variable link conditions, not always due to mobility
    * different traffic types, e.g., control loops with reliability and latency
    constraints, co-existing with monitoring.
Requirements:
* coexistence
* multiple access technologies simultaneously

Carlos introduces wireless gaming
Specifics
* shared real time information between players for real time gaming
* also console gaming, low latency and jitter => a  focus for RAW
* and cloud gaming
Requirements
* Carlos provides numbers
* for RAW:  time sensitive, priority, time-aware shaping, see slides

Summary
* input from multiple people
* inviting others like smart grid

Toerless:
* slide 8 seems not requirements but technology list. Need analysing methods
that can be used, should be more. * in multicast there's a document for
multicast in wireless where solutions are compared Rick: * plerase find that
doc for the WG Carsten: * There's req that this need to work over paths with
multiple wireless links. Does that mean it works over links that thenconstitute
paths or is it working across multiple links spanning. Carlos: * some use cases
have multiple links of multiple technologies. Can be multihop networks, or
single hop with multiple physical links. Carsten:
    * the latter is answering my question: Yes, work across multiple links
Bob Hinden:
    * what's the need for gaming? Seems to be going on already.
Carlos Bernardos:
* I did not propose that use case, need exists, offline to ML
Rick Taylor:
    * not just about wireless gaming. This is afrom left field use case, a lot
    of others exist too
Eve Schooler:
    * multiple streams from multiple cameras, precision stitching and better
    visulalization
Kuhnn:
    * muted

Eve Schooler introduces LDCAS as the next presentation
Rick Taylor:
    There will be no humming for adoption in this sessioN
Nils Mäurer present LDACS
* second version of the draft
* comes from aeronautics communication, foreseen trabsition to IPv6 and several
links, one of them LDACS (and AeroMACs for airports communication) * does not
matter which link it takes, what's important is in-time delivery * draft 01
introduces the technology and IPv6 over it. Going through the TOC. * no
standard at ICAO, opportunity to define the rules for an IPv6 link that covers
multiple datalinks * details the protocol stack and services, see slide 7 *
OFDMA hardened for aeronautical interference, FDD 2*500KHz, superframes *
ground station GS controller handles 500 planes * How do we transport IPv6
safely to the plane? * < 2Mbps over <400Km in regulated 1GHz spectrum * RAW:
multihop multi link system, LDACS is an open standard in development at ICAO
Stuart Card: * how this related to ICAO trust framework (GRAIN), will post
slide from X Da Silva on that Thomas Gräupl :
    * this is an FAA project not what we are doing at ICAO; GRAIN uses
    commercial off the shelf. The use of COTS for air traffic control is
    debated, no standard at ICAO can do that legally as of now. ICAO dedicates
    links. Rest is far in the future

Janos Farkas on 5G:
    * provide some details on 5G, indicates it is a reliable and available
    tech, good match for RAW;
Torsten Dudda:
    * outline, industrial use cases a focus
    * 5G include URLLC, all about providing bounded low latency, include time
    synchronization * 5G NR is scheduled by gNB, QoS framework, include packet
    error rate * low latency features, see slides
Janos Farkas
* 5G has options for multi connectivity. e.g; DUal UEs, see slides
* TSN integration with 5G added in release 16
* virtual TSN bridges incorporate the ethernet PDU sessions, enables end-to-end
deterministic ethernet * done to cover Carlos' requirements earlier * TSN
translator the key component * DetNet had TSN in mind as subnet technology.
Rechartering DetNet to enable others. * How do we combine 5G for DetNet.
Various combonations possible, see slides * Summary, see slides Rick Taylor:
    * are the drafts presented in scope of the charter? I believe 5G is
    interesting, does it apply to RAW? RAW looks at heterogeneous. * Are the
    authors calling for WG adoption?
Tom Kapela:
How to avoid deadlocks? Solves similar problems push flow control to the edges,
convergence at odds with congension mgmt activities. How does that apply here?

Janos:
    PRE offers 1+1 protection with duplicate copies with seq nb on different
    path

802.1CB replication and elimination, disjoint paths, sequence numbers. No
failover time. Job of central controller in SDN approach. Scheduling and the
central controller avoids congestion loss Tom: What about single UE contacted
by multiple incoming sources - how do we deal with that, potential deadlock.
Janos: Agree, need to deal with that. Abdussalam Baryun: * I'm interested in
use cases to know challenges, so what are the main challenges for this
technology use case. Janos Farkas: * high availability and reliability critical
for industrial. Downtime is expensive. * typical closed control loops with
periodic CBR, time requirements must be met. Latency also a factor.

Defering next questions in interest of time

Georgios Papadopoulos presents PAREO requirements
* slides 6-12 detail the PAREO operation over  a scheduled multihop mesh
* Alternative parent selection not written in the draft yet, to be discussed in
ML whether we add it or not

Lou Berger:
    * What about this is specific to wireless (vs. any DetNet).
Georgios Papadopoulos:
    * Overhearing, constructive interference
Lou Berger:
    * interaction between subnet and IP and how this is exposed
    *  such as one subnet path taking away resources from another - how is this
    exposed to IP layer.
Greg Mirsky:
    Where RAW reliability is built
Georgios Papadopoulos:
* can do tricks at MAC layer for single or multiple acks in case of
overhearing. Need to avoid ack collision, to be discussed at layer 2 Rick
Taylor:
    * Discussing different things: probability of delivery vs guarantee of
    delivery. * I think this is interesting work not in the immediate charter
    but could be adopted and worked on, whether it ends up here or not is
    secondary

Georgios Papadopoulos presents OAM:
Could be topic for RAW or DetNet, TBD.
* specific challenges, how much resource to allocate, energy efficiency
Rick Taylor:
    * Req for adoption - address on the ML
    * for Nils, unclear if you are requesting adoption, will send email
Nils Mäurer
    * We are looking for reviews and are aiming for working group adoption at
    IETF 108 :)
Toerless Eckert:
    * Interested in PREOF, most interesting part of DetNet.
    * scared if tried to limit to known building blocks for PREOF, there's a
    lot of research out there
Florin Baboescu on 5G:
Lou Berger:
* Fundamental question for the group is what is the model of integration for
the DetNet layer and wireless subnet layer? Is it integrated or layered? * WRTo
Tom's question earlier re flow control - there are two control loops, a longer
control loop that requires subnet information to be exposed to the control(ler)
-- for it to pick the right (IP) paths and a shorter control loop that allows
the IP sub-net entry point's queuing to adjust to congestion at the wireless
sub-net level.  While we're not at solutions yet DLEP from the MANET WG can be
used to support both Stewart Bryant: Reconciling guaranteed delivery vs
guaranteed latency - can't solve for both. Can just increase probabiity of
losing a packet by some amount. Would have to wait an infinite amount of time
for guarantee. Pascal Thubert: Yes this is correct, working for bounded
latency, best delivery ration. Even wired has limited number of nines, wireless
will have fewer. Looking at all the ways to increase the number of nines for
wireless. Stewart: But then how do you set a number around that? Puzzled since
worked on PREOF solution, still trying to define this e.g. for safety of life
system. Toerless: In DetNet zero percent loss, or 100% bounded latency -
depends on which use case for what the requirements are for this. Less strict
requirements can be acceptable given the use case. So incremental improvement
options would help even if not purely deterministic a la DetNet.

Stewart:
Benchmark is to just multiply the bandwidth and see if that is good enough.
Janos:
    Are you considering this as an issue just with wireless or with wired also?
Eric:
    Please take to the list.
Eve:
So there are good questions here about mission and gaps, specifying QoS,
determinism or broader, reconciling competing requirements. Good for future
discussions. Any comments from chat window - make sure they were captured in
notes. Maybe there were 73 participants - not all in blue sheet? Will make
request to ML to verify. Meeting adjourned.