Minutes interim-2019-coinrg-02: Tue 09:00
minutes-interim-2019-coinrg-02-201910080900-00

Minutes

Meeting Recording at:
https://ietf.webex.com/recordingservice/sites/ietf/recording/playback/895d01c2a33b4d35b7ede03a898ce5bb

Intro and Administrivia

Marie-Jose: Thank you to those getting up early and those staying up late to
join us today

This is an interim meeting of the Proposed RG COIN  - Computing in the Network

With a focus on the Edge-Cloud continuum

We feel lucky that the topic is cropping up at many conferences and in many
academic venues.

The COINRG is moving from using Github as our supplemental meeting space to
using the Wiki

One of the reasons for GitHub is the ability to store large files, such as this
recorded meeting

PRG Status

Charter – pretty stable and no change is planned before next meeting IETF 106

Chartering IAB breakfast meeting Mon Nov 18th to discuss COIN RG

Will need to be careful defining milestones and achieving them

At present, most milestones proposed have related Internet Drafts that have
been submitted – see meeting slide 6 for details

Dec 2019

-          New topics to be included

-          Articulate COIN challenges – I-D on this topic likely to be updated

-          Capture the SoTA of the COIN landscape – to include H2020

April 2020

-          Target COIN case studies, from architecture, implementation and use
case standpoints

-          Identify CION network-related eco-system dependencies

-          Discuss/catalog COIN requirements and implications for network
elements (including network services, network SW stacks, network HW design, etc)

Nov 2020

-          Work toward defining a COIN scope appropriate for the IRTF, within
which new architectures, mechanisms and protocols can be proposed

To advance to an official RG, we want to show that it is not just the co-chairs
doing the work and that there is genuine interest on the eco-system and
architecture sides as well

-          Industrial use case is thriving

-          For the XR draft, UCLA are interested to help MIT to evolve this
draft

-          Would like to help Noa (and other academics) get more involved in
the RG, despite the lack of cycles to write drafts

There is another dimension to this work we are trying to achieve in the
Hackathon

-          How to define the programmable data plane

-          P4 is one way to go, but it is certainly not the only way to do
programmability in the network

-          Let’s also revisit the possibility to have a common project for the
group and there are a bunch of sub-projects or sub-pieces that could fold into
it

-          Would like feedback today and ongoing on the list

Meeting at IETF 106

-          Requested a time slot

-          May be first meeting as an official RG

-          Deadlines

o   New I-Ds: 2019-11-04

o   Agendas: 2019-11-06 – aiming to get in early due

-          Agenda items welcome and please review/update your drafts

-          Beware that the I-D submission tool is not as friendly as one would
hope – get a good html editor

Hackathon@106

-          Sat and Sun 11/16-11/17

-          Develop further the data filtering work started in Montreal that has
value in industrial networks and video distribution

o   Basically, based on the identity/attributes of the packet, “do something”

o   In industrial nets, the identification of packets is similar to stuff done
in autonomous vehicles

o   Be able to tell which packets represent an alarm, or the status of a
machine, etc, then want to be able to filter the packet early in the chain of
processing

o   In video distribution, the filter can actually be based on the kind of
video, destination, type of content the video carries, as described in the XR
I-D

o   Not just video filtering but anomaly- or event-driven data filtering, as
presented at the COIN BoF re the Ubiquitous Witness use case, which focuses on
video analytics for anomalous events (e.g., an accident at an intersection)

o   Classification of packets – accumulation of packets in the right places,
and actions for training data

-          At the moment, the language used so far for the 1st COIN hackathon
is P4

o   Thus, it would be good to continue the follow-up and develop data filtering
with P4, perhaps to seed a larger project

o   Once we have basic P4 libraries we can ask what else we want to do

o   Post Singapore – let’s discuss an ongoing dev effort, where we continue to
build out libraries

§  NoviFlow, Kaloom, Barefoot acquisition, other devs beyond those who
participated

§  Open source set of libraries

-          Helpers welcome/needed!

o   Unlike Montreal, there are currently no plans for developer support during
the Hackathon

o   Marie-Jose’s students will come

o   Helpers are welcome, especially those working on P4, even if not P4 expert,
even if just doing this kind of programming in another language

o   Not going to do P4 necessarily

o   There are benefits to hanging out with other COIN RG folks at the
hackathons and in letting the IETF feed you J

Documents Update

-          Expect Jorg and Dirk to update their Challenges I-D

-          Dirk Trossen and team still working on and intend to update their
draft

-          Jeffrey expects to update draft and slides on DC – likely some new
use cases

-          Eve expects to update Edge Data Discovery, particularly lead a
discussion on the mailing list re expanding the I-D to cover Data lifecycle

-          New Virtualization draft on the horizon

-          Will talk to the team in Chicago to talk about video

-          Also Noa Zilberman – hard for people at universities to write I-D,
not just due to lack of cycles, but also lack of academic credit for I-Ds

-          Marie Jose has the intention to see the RG to help folks get work
written up

Colin – IETF is very focused on drafts, but the IRTF doesn’t have to be

IRTF could give air time to any doc, which could be a latest publication

There are a few additional ways RG can have influence

-          Writing down architectures and requirements and drafts with
particular use cases and experimental protocols

-          Enabling a forum for people to get together and talk about things

o   Papers presented elsewhere

o   Early draft papers

o   Discussions across pockets of people (industry and academics)

o   Facilitating other work (beyond I-Ds) also useful

-          Can upload all sorts of documents, and as long as a pdf it will be
fine

-          Hold an interim when there is a major conference in the area, as a
debrief and/or schedule at the same time, e.g., holding ICN RG at ICN ACM

A lot of conferences to happen at the beginning early 2020 that might influence
this group

-          Plan to provide a list of related conferences or projects in
Singapore

o   Would allow us to promote them in this group but also this group in those
conferences

-          E.g., NETPROC@ICIN2020

o   1st WS on Flexible Net Data Plane Processing

o   https://www.icin-conference.org/netproc-2020/

-          Be aware of what else is going on and cross pollinate with our ideas

New topic

-          Latency by Virtualization – Deferring to Singapore

Open Discussion

Marie-Jose et al

-          COIN RG would really like to help researchers get involved who don’t
typically engage with the IETF

-          The mailing list activity would seem to suggest that impact is thru
I-D creation, yet that mechanism is not always a good fit for academic
researchers

-          Rather getting papers into INFOCOM, SDNI, SIGCOMM are more
compelling for academics

-          Yet it is really important to have discussions between IETFers and
researchers – use cases, data stewardship, making research real

-          Would really help to come up with other ways to engage

o   e.g., Dagstuhl minutes shared on the mailing list

o   e.g., help researchers disseminate info on the mailing list who would never
post directly

Daniel

-          Agree that not all presentations need to be I-Ds, but rather could
give visibility to academic research, which may not even be written up as
academic papers yet

-          there is certainly value in drafts that exist and how standard
protocols will need to shift

-          it is helpful to see and share pseudo code

-          as it is impossible to follow everything going on, having linkages
to other conferences and research papers will be helpful because not even
SIGCOMM will have everything – could use the Wiki to help make them visible

Jeffrey

-          Let’s move and maintain the References list from the Wiki to GitHub

-          Track Major papers

-          Host Invited talks

Colin

-          Can view the RG as more of a discussion place for a start

-          Enable getting in as many viewpoints as possible at first

-          Naturally, coalesce around a group of docs worthy to publish

Eve

-          Aspiration to make COINRG the meeting place for the broader community

-          It will take work

Ike Kunze

-          We may have different definitions of what we mean by Industrial IoT

-          Working on 2 different topics re industrial networks

o   Really big machines, production plants, building cars

o   Whereas we think others might mean large data centers? [Absolutely not says
Marie-Jose]

§  Devices are in the plant vs a DC

§  Data filtering

-          Not sure then if there is a common understanding yet

-          A common problem with the Industrial draft so far: we don’t have the
implementation or we don’t talk yet about what is really done in industry,
because we often find closed source solutions – biggest challenge

o   Cannot present what’s been done already

o   Related to what Noa raised

o   What can we do to help here?

o   How to support/collaborate with real industry innovators

o   Could be a topic for ISRG dinner – because all groups need to face this
divide, e.g., between industry and academia

o   Let’s start a trend[à thread?] in COINRG to capture – alongside the
Academia challenge stated by Noa - the problem statement, i.e., that it is hard
to capture real experiences, problem statements, feedback and best practices
from real companies

Threaf.

Hannu

-          Q: Hannu was asked, from your industrial research perspective, how
do you feel about the direction of COIN RG?

-          Closest standardization paths seem to be 3GPP/SA6 (app level) and
SA2 (infrastructure level) and ETSI/MEC and ETSI/NFV, they are going deeply
into architecture starting from the mobile network

-          Though we are not looking at standardization just yet

-          Q: Plans to implement any of this? Absolutely. Operators/industries
interested in investigating the benefits of processing close to where things
happen.

TUM – [listen to audio]

He just said he was from Jorg’s team.

Jeffrey

-          Interested in Edge computing and DC

-          I-D mainly focused on DC use case

-          How can COIN can help build database inside DC?

-          Secondly, also have some implementation of a PoC on COIN for AI
chaining

-          improvements to transport protocols

-          Some of the experiences in the DC (re DB management) can be applied
to computing use case in context of Edge computing – expect in the near future
to explore that topic and those who are interested please contact Jeffrey

Eve update

-          Eve gave an update on the genesis of the Edge Data Discovery I-D and
how it relates to COIN

-          Computation needs to marshal input data from somewhere and to place
computational output somewhere

-          Had great feedback after -01 and , A lot of drafts that cover more t
han one research group, so it would be fine

-          ICN and Compute First Networking – we overlap

-          Could overlap with DIN and other WGs in the IETF

-          Autonomous migration of data

Xavier de Foy

-          What is your interest? Involved with Dirk Trossen (InterDigital) and
interested in Edge computing in 5G Networks, related to 3GPP and ETSI/MEC
activities

-          IoT and Edge

Dong Lin

-          Actually working with Jeffrey (Huawei)

-          Hoping to contribute to this working[àresearch] group

CoRE WG

-          Carsten Bormann – mostly lurking because in the end, constrained
networks particularly constrained node networks are looking for friends to
offload to

-          Lots of IoT related issues, could pick up in CoRE and T2T

-          COINRG branched off from T2T and ICN, and a few other things

Action Items

-          Start a thread on the mailing list on how we can help academia to
participate, shepherd docs, allow more discussions, supporting smaller group
discussions as necessary

-          Start a parallel thread on how we can help industry to participate

-          Look at progressing the GitHub presence, i.e., moving the references
there

-          Load references that are part of the Wiki into the Data Tracker

-          Catalog the list of conferences/workshops that are happening in this
area

-          If you have a draft, make sure it is updated on time

-          Please send us Agenda items for Singapore – ok to participate
remotely and offer ideas for presentations