►
From YouTube: Kubernetes WG IoT Edge 20220601
Description
June 1, 2022 meeting of the Kubernetes IoT Edge Working Group. Discussion about this Group moving to CNCF (is/was under Kubernetes) Why. Some impacts of the move - new links. Charter working document. Rishit researching federated learning at the edge
A
Good
yeah
we
just
had
a
long
weekend,
so
this
Tuesday
has
been
quite
eventful.
Getting
back
what.
B
A
Yeah
wow,
that
is
quite
the
other
time,
I
had
a
a
7
A.M
one
this
morning.
Actually
it
was
just
kind
of
a
horrible
bookmarks.
A
I
went
to
the
container
orchestrated
device
working
group
meeting
this
morning,
and
so
it's
funny
just
having
bookend
Community
meetings
on
either
side.
But
that's
the
fun
part
about
being
the
space
is
people
are
all
over
the
world
so.
B
A
Awesome
yeah,
a
lot
of
us
are
coming
back,
so
is
it
10
a.m?
For
you,
then,.
A
B
B
D
A
Yeah,
do
you
want
to
kick
us
off
with
the
Spiel
Steve
sure.
D
So
welcome
to
the
June
1st
meeting
of
the
kubernetes
iot
edge
working
group.
This
group,
as
usual,
abides
by
the
kubernetes
code
of
conduct,
which
means
all
meetings
are
public
and
recorded
So.
If
you
don't
want
to
be
recorded
and
have
that
recording,
put
up
to
YouTube
Drop
now
or
turn
off
your
camera
and
keep
yourself
muted
with
that
said,
unless
somebody
added
something
the
agenda
tonight
had
an
item
discussing
about.
D
Why
and
how
this
group
is
moving
over
from
this
kubernetes
project
over
to
the
cncf,
so
I'm
going
to
let
Tiana
Kate
takeover
that
subject,
because
they've
been
the
ones
doing
the
work
on
doing
the
Arrangements
on
this
and
I.
Don't
know
if
you've
got
anything
to
prepare
Kate
but
I'm
going
to
make
you
a
co-host
too.
A
Yeah
I
think
just
a
general
discussion,
so
we
talked
so
we
joined
the
runtime
tags
session
at
kubecon
Europe
and
in
that
session
kind
of
talked
about
some
of
the
motivation
there.
So
that's
also
a
good
resource
to
kind
of
look
at
some
of
the
statements
we
made
around
this.
But
previously
this
working
group
has
lived
in
the
kubernetes
working
groups
face
and,
as
we
went
to
kind
of
renew
for
the
next
year,
we
kind
of
thought
about.
Okay.
A
Maybe
there
would
be
a
better
place
for
this
and
oftentimes
kubernetes
working
groups
are
working
towards
specific
objectives
and
some
of
those
objectives
are
sometimes
adding
things
to
kubernetes
Upstream
and
that
hasn't
necessarily
been
the
scope
of
this
working
group.
It's
more
been
around
bringing
people
from
across
different
projects,
whether
cncf
or
external,
to
kind
of
discuss,
Edge
and
iot
solutions,
and
a
lot
of
them
are
kubernetes
based,
but
not
all
of
them.
A
So
one
of
the
topics
that
Steve
talked
about
and
the
working
group
session
at
kubecon
EU,
was
about
secure
device
on
board,
which
is
a
tool
for
provisioning
servers
on
the
edge
kind
of
in
a
zero
touch
way,
and
that's
not
kubernetes
specific.
A
It
can
be
used
to
provision
kubernetes
clusters,
but
it's
not
kubernetes,
but
it's
specific,
and
so
we
were
looking
at
transitioning
to
the
cncf
to
kind
of
expand
the
scope
of
the
working
group
Beyond
kubernetes,
and
it
also
is
a
good
space
for
for
not
necessarily
having
to
have
the
objective
of
contributing
to
Upstream
kubernetes.
A
So
then
I
think
or
I
guess
any
questions
on
that.
D
Well,
I
guess
one
thing
that
I
think
is
a
natural
to
account
for
is
just
when
it's
likely
to
take
place.
In
other
words,
when
are
we
officially
gone
from
kubernetes
and
officially
in
cncf,
I
have
a
feeling
that
one
of
the
repercussions
of
that
is
that
our
Zoom
links
will
change,
for
example,
and
there's
there's
probably
some
other
administrative
like
the
kubernetes
project,
won't
keep
us
in
their
calendar
anymore.
D
They
probably
won't
keep
us
under
the
kubernetes
project
for
the
video
recordings
on
YouTube,
but
I
assume
the
cncf
has
an
alternate
place.
So
I
think
what
we'll
do
is
probably
just
leave
some
pointers
behind
in
all
those
places.
So
people
can
discover
where
we
move
to,
but
we'll
probably
need
a
little
prep
time
to
lay
those
things
down
and
begin
operation
at
the
other.
One.
A
Yeah
I
do
think
the
first
steps
I
was
talking
to
Ricardo
from
the
runtime
tag,
is
really
to
create
our
pull
request
into
runtime,
and
so
the
bash
working
group
The
like
bash
system
initiative
working
group,
has
their
PR
up
because
they
were
the
other
working
group
that
presented
at
the
runtime
tag.
That's
also
joining
and
I
have
a
message,
Ricardo
kind
of
asking,
because
another
thing
was
as
a
part
of
a
kubernetes
working
group.
We've
been
able
to
present
at
kubecon's
kind
of
the
work
that
our
group
has
brought
together.
A
Currently
with
the
runtime
tag,
they
kind
of
have
one
talk
for
their
working
group,
and
so
he
mentioned
having
the
pull
request
up
would
be
a
good
spot
to
like
to
have
that
conversation
around
of
creating
a
session
for
us
within
the
cncf,
but
having
that
Charter
in
progress
shows
that
that
is
a
legitimate
stepper
taking
and
so
I
can
I'll
drop
a
link
to
that,
and
also
we
can
put
this
in
the
notes.
But
this
is
the.
A
This
is
the
one
that
the
bash
system
initiative
working
group
has
up
and
so
I
think
we
would
want
to
dejan
started
a
charter
for
us
and
so
I
think.
Maybe
we
can
switch
that
to
being
kind
of
this
markdown
format,
where
we
give
an
introduction,
a
background,
some
goals,
some
in-scope
and
out
of
scope
operations,
details
so
like
who
are
our
Toc
liaison
and
chairs
for
the
from
the
Sig
and
then
who
are
the
tech
leads
and
then
Community
Information.
A
So
the
things
you
were
talking
about,
Steve,
where
it's
I
think
for
now
we
can
probably
point
to
our
existing
YouTube
channel,
but
we're
going
to
want
to
make
a
new
slack
Channel
and
maybe
a
new
mailing
list
and
I
think
it's
fine
to
use
our
existing
YouTube
channel
but
and
existing
Zoom
links
for
the
time
being.
A
Kubernetes
is
a
part
of
a
cntf
like
I
I,
think
that
is
probably
fine
to
start
with,
but
certainly
something
we
can
explore
in
the
full
request.
Yeah.
D
A
Yeah
we
can
have
GitHub
page
redirects
too,
so
the
redirect
can
happen
from
I,
don't
know.
Actually,
if
you
can
do
that
with
individual
pages
and
then
you
can
do
with
orgs,
but
something
to.
A
B
B
A
D
A
D
Opry
live,
does
akri
live
in
the
kubernetes
slacker
over
in
cnco
we.
D
So
maybe
we
don't
even
have
to
move
it.
I
don't
know.
A
I
mean
I
think
that's
reasonable,
like,
for
example,
the
container
orchestrated
device
working
group.
That's
under
runtime
doesn't
have
its
own
stock
Channel.
They
just
use
the
runtime
tags
back
Channel,
which
I
would
prefer
not
to
have
the
case.
I
like
that.
We
have
our
own
Channel
people.
Ask
good
questions
about
scenarios,
so
I
think
yeah.
Keeping
it
in
kubernetes
would
probably
be
fine.
B
D
A
I
think
every
two
weeks
is
probably
good
to
start
out,
especially
because
we
had
some
people
coming
out
of
kubecon
who
wanted
to
present.
Like
I,
know
the
container
orchestrated
device
working
group.
Those
folks
wanted
to
come
to
this
I
went
to
their
meeting
this
morning.
A
I
was
telling
daehan
that
and
they
are
basically
making
it
so
that
you
can
request
devices
with
parameters
and
adding
it
extending
the
kubernetes
API
to
do
that,
and
so
they
have
some
interesting
things
that
they
could
present
and
share,
as
they
do
pertain
to
the
iot
space.
They
actually
pertain
very
heavily
to
awkward,
which
is
why
I
went
to
listen
into
a
bit
but
yeah
so
they're
an
example
and
then
I
know.
Steve
talked
to
some
other
people
who
are
interested
and
yeah.
So.
D
Yeah
I
think
poor
tainer
wanted
to
come
back,
it's
been
about
a
year
since
they
talked
to
us,
so
they
were
interested
in
doing
an
update
and
Intel
was
aware
of
the
secure
device
on
board,
but
they
have.
They
have
some
more
tools,
sort
of
related
to
that,
but
maybe
extend
the
mission
Beyond
just
the
initial
onboard
and
they
said
that
they'd
be
interested
in
giving
a
presentation.
A
B
No
yeah,
it's
okay,
I
mean
we
hear
today,
Rashid
here
for
for
for,
for
whom
we
created
these
kind
of
third
times
for
to
be
like
a
you
know,
good
fit
for
for
India
and
China
and
like
that.
So
if,
if
we
get
direction,
it's
it's
even
better
so
and
yes,
they
will
know
I
mean
the
biggest
hurdle
we
we
ever
had
with.
This
working
group
is
changing
the
the
meeting
time
times
because,
like
six
months
later,
people
are
still
showing
up
on
Fridays,
even
if
it's
more
to
Wednesday.
A
D
Yeah,
if
we
want
to
go
two
weeks
instead
of
monthly,
you
know,
one
of
these
two
times
is
a
little
difficult,
but
I've
managed
to
live
with
it.
So
I
can
continue
to
live
with
it
and
it's
probably
convenient
for
some.
So.
A
Yeah
I
think
it's
I
think
it
would
also
be
a
fair
point,
though
we
do
have
a
pretty
big
Community.
It
might
be
worth
just
putting
out
a
poll.
We
have
533
people
in
that
slack
like,
and
maybe
we
should
pick
what
the
majority
of
people
pick,
especially
for
reevaluating
everything
if
we
have
to
create
a
whole
new
document,
people
are
going
to
expect.
Some
things
have
changed
so
I
think
now
is
the
time
to
reevaluate
our
times.
If
we're
going
to.
A
B
So
what
worked
for
us
in
the
past
is
that,
even
when
we
have
a
much
bigger
attendance,
it
was
that,
like
a
a
regular
meeting
cycle,
was
much.
D
B
Visited
so
so
so
that
was
the
regular
session
and
we
used
this
kind
of
Apec
sessions
as
a
as
a
as
a
working
sessions,
because
if
you
just
us,
we
can
work
on
a
quick
comp,
presentation
or
or
whatever
it's
coming.
So
that
was
that
Cadence
was
also
good
for
that
right.
So
one
more
public,
one
more
private,
but
not
by
Design
right
yeah,.
A
Well,
I
think
it
depends
how
much
we
want
to
dive
into
this
being
a
working
meeting.
I
think
we
could
start
looking
at
making
that
PR
and
go
back
to
that
Charter
that
you
started
on
I.
Think
that
would
be
a
reasonable
thing
to
do.
A
C
Patrick
over
there-
and
he
motivated
me
to
some
of
the
working
group
activities,
so
this
is
actually
my
first
meeting.
So
let's
see
how
the
working
group
functions.
D
Now
often
we
have
speakers
and
sometimes
No
Agenda
at
all,
in
which
case
just
whoever
shows
up
Kelsey.
We
tell
each
other
what
we're
interested
in
at
the
moment,
or
you
know
interesting
things
we've
seen
recently,
so
maybe
we
can
take.
You
know,
take
a
few
minutes
to
do
that
right
now
and
get
back
to
working
on
the
transition
after
we
have
a
little
bit
of
discussion.
Maybe
you
can
tell
us
what
kinds
of
things
are
you
interested
in
related
to
Edge?
C
Function
so
I
usually
work
a
lot
with
machine
learning,
so
it's
office,
research,
I
wanna
I,
have
also
been
interested
in
it.
One
of
my
recent
papers
was
actually
about
a
algorithm,
so
I'm,
quite
interested
in
running
machine
learning
on
the
edge
and
recently
I've
been
exploring
a
lot
of
Federated
learning
on
the
edge.
How
you
can
do
is
kubernetes,
and
actually
so
I
actually
worked
with
that
and
also
have
a
top
coming
of
contact,
so
hopefully
I'd
keep
going
or
some
other
place.
C
Probably
but
yeah
I
have
to
stop
Count
that
one
Federated
machine
learning.
D
As
far
as
I'm
concerned,
you're
welcome
to
get
that
talk
here,
even
if,
even
if
you
want
to
do
a
practice
session,
go
ahead
and
do
it
for
a
future
meeting.
That's
that
topic
certainly
interests
me.
D
What
what
would
it
be
that
you'd
cover
related
to
Federated
learning
at
the
edge?
You
know
right
now,
in
my
experience,
the
more
common
scenario
is
people
collect,
training,
sets
at
the
edge
and
then
do
the
actual
learning
up
at
a
higher
level
than
at
the
edge
itself,
but
I'm
kind
of
curious
as
to
what
the
advantage
is
of
doing
it
at
the
edge
and
what's
different
about
it
versus
you
know
doing
it
in
a
public
cloud.
C
So
the
main
difference
is
oh,
so
one
of
the
things
we
were
excited
so
my
friend
and
I
are
actually
working
on
this
effort.
So
one
of
the
things
we
are
we
were
interested
in
is
not
having
the
data
leave
the
device
itself.
Oh.
C
Yeah,
so
the
data
does
not
leave
the
device
but
to
improve
the
model
somehow
and
also
improve
it
for
other
users.
So
tattoo
is
one
of
our
main
motivations
while
exploring
Federated
machine
learning
and
oh,
and
so
the
talk
which
we
were
planning
to
do
was
how
to
run
computations
in
in
this
panel.
Also,
we
start
off
with
something
like
how
do
you
run
computations
in
this
sense,
and
it
could
just
be
generating
conversations
as
well
and
then
move
on
to
a
machine
learning.
C
D
D
Gpu-Like
accelerator
chips
that
are
appropriate
for
Edge
devices
coming
out,
you
know,
and
in
the
early
days
those
things
used
to
be
such
power
hungry
and
usually
demanding
PCI,
Express
buses
and
things
to
host
them.
But
it
seems
like
more
and
more
of
these
microcontrollers
now
are
coming
out
with
chips
that
assist
the
machine.
Learning
process
out
of
edge
locations
that
have
reduced
cost
reduced
power
and
maybe
are
even
bundled
into
a
lot
of
these
Edge
devices,
are,
are
those
things
supported
by
the
tensorflow
libraries
and
things
typically.
C
All
right,
yeah
yeah,
so
so
the
idea
was
that
you
take
a
tensorflow
right
model
and
tensorflow
Lite.
It's
a
project
which
I
contribute
a
lot
to
and
what.
D
Do
you
know
if
Google
has
started
selling
the
tpus
that
they're
using
in
like
pixel
6
for
use
in
Edge
devices,
or
is
that
on
the
roadmap?
Or
do
you
even
know.
C
D
You
know
this.
The
same
processing
units
that
are
in
cell
phones
should
be
fairly
inexpensive
and
fairly
conservative
about
power
consumption
and
likely
have
big
production
volumes
which
tends
to
align
with
low
cost,
so
it
it
wouldn't
surprise
me
to
find
out
that
those
find
their
way
out
at
Edge.
Unless
you
know
the
current
chip
manufacturing
constraints
are
maybe
causing
them
to
be
in
short,
supply.
B
A
In
all
these
scenarios
are
the
models
running
on
the
devices
that
are
capturing
the
data
as
well,
or
is
there
kind
of
a
mixed
mix
of
like
running
on
separate
compute
that
has
access
to
those
devices.
A
C
So
often
each
devices,
but
not
including
microcontrollers,
the
AVR
I
I,
have
explored
this
with
this.
Just
mobile
phones,
because
we
have
something
called
tensorflow
Federated,
which
very
well
supports,
are
simulating
such
kind
of
devices.
So
right
now,
I
want
to
explore
this
in
a
little
capacity
with
no
business.
C
Oh
yeah,
so
yeah
I'm
actually
submitting
it
for
keep
one
North
America.
All
right
I
was
wondering
how
it
might
work,
because
I
already
had
a
dog
and
a
Cubed
on
Europe
but
I,
but
on
the
machine
learning
and
data
track,
and
they
told
me
about
the
time
constraints.
It
was
an
accepted
the
they
shifted
it
to
you've
won
any
so
I,
don't
know
how
it
would
work
out,
but
I
I,
probably
anyways,.
D
C
D
A
I
think,
as
long
as
they
won't
with
the
reviewers
I
mean
Steve
actually
knows
this
way
better.
But
from
my
understanding,
the
reviewers
don't
know
that
you
had
another
one.
D
The
only
maintainer
track
so
I
believe
and
actually
I
think
there
is
another
exception
that
if
you're
on
a
panel,
then
you
can
have
a
second
non-panel,
but
otherwise
I.
Don't
think
that
two
will
ever
be
accepted.
You
are
allowed
to
submit
to,
but
they're
likely
to
not
ex
I,
don't
believe
the
rules
would
let
them
take
to
conventional
presentations.
D
A
D
C
D
I'm,
not
Stephen,
I
probably
am
not,
but
you
know
you
never
know,
but
I
think
at
this
time
I'm
not
I
I'm
not
booked
for
travel.
There,
normally
I
only
get
sent
to
them
if
I'm
speaking,
but
you
never
know
anyway
later
in
the
year,
there's
Europe
and
there's
also
Japan
coming
up
in
December
I.
Believe.
A
D
Other
place
it
might
fit
is
if
they're
having
the
kubernetes
on
edge,
Day
event
again,
which
I
think
they
all.
They
probably
are
that's
a
separate
submission
and
it
typically
comes
later.
D
There
were
a
few
machine
learning
talks
at
kubernetes
on
ancient
Europe,
so
clearly
they
are
looking
for
talks
on
that
subject.
So
I
think
you
have
a
lot
of
venues.
You
can
go
for
with
with
a
talk
like
that.
D
C
Good
question:
how
do
I,
what
What's
the
best
way
to
let's
say
a
part
of
these
greetings,
which
I
love
first
video,
but
what's
another?
What's
the
other
best
ways
to
get
involved
with
what
they
do.
D
Well,
the
slack
Channel
you
can
kind
of
just
asynchronously
ask
questions
or
say
whatever
you
want
to
encourage
discussions,
so
that
works
pretty
well
there.
There
in
theory,
is
a
mailing
list
that
when
you
join
the
group,
there's
an
email
list
and
I
think
any
member
who
wants
could
put
something
on
there,
but
nobody
really
ever.
Does
it
so
I,
don't
think
it's
commonly
used
as
a
mechanism
for
interactive
exchange.
I
think
that
pretty
much
occurs
on
Slack
foreign.
D
And
then
we
could
get
back
to
a
working
group
meeting,
but
one
other
thing:
I'll
leave
it
to
the
audience
to
see
if
you're
interested
so
I
did
that
secure
device
on
board
demo
at
kubecon
and
one
thing
you
know
because
of
the
limited
time
The
Limited
bandwidth
to
keep
it
from
going
live.
It
kind
of
got
cut
off.
So
I've
still
got
that
note
that
I
kicked
off
running
right
now
in
my
home
lab.
D
So
if
anybody
was
curious
and
wanted
to
see
the
Remote
Management
of
that
under
portainer,
which
is
the
part
that
got
left
on
The
Cutting
Room
floor
I
could
just
briefly,
if
nobody's
ever
seen,
portaine
or
remote
manage
a
kubernetes
installation
before
I
could
spend
a
few
minutes
just
sharing
a
screen
and
showing
that,
but
up
up
to
you
and
we
might
have
the
pertainer
person
coming
to
do
a
presentation
in
the
near
future
anyway.
So
maybe
that's
not
needed
at
this
point.
D
Okay,
can
you
see
my
browser
yeah?
Okay,
so
this
is
just
a
portaner
is
running
on
some
node
it
doesn't
have.
It
typically
would
not
be
the
one
that
your
remote
managing,
but
it
could
be
I
suppose
so.
I've
just
got
it
running
on
another
Edge
Linux
box,
it
it
it's
a
low
powered,
Intel,
atom,
eight
gig
of
RAM
physical
machine
that
is
hosting
this
running
in
Docker,
and
it
just
presents
a
web
interface
I've
already
logged
in-
and
this
is
the
dashboard
for
this
kubernetes
node-
that
I
onboarded
using
secure
device
onboard.
D
So
like
I,
said
in
the
kubecon
presentation,
it
took
like
25
minutes
going
from
kind
of
nothing
just
a
bearer
OS
that
got
upgraded.
It
pulled
down,
OS
upgrades
docker
and
installed
the
kubernetes
distro
and
deployed
a
single
node
cluster,
and
this
cluster
is
just
a
kind
cluster.
Basically,
so
it's
pretty
simple,
but
this
is
the
dashboard
I
in
the
dashboard.
What
it
says
our
applications
are
actually
the
pods
that
are
running
and
I'm
clicking
on
it
here.
D
It's
it
uses
an
agent
model,
so
the
kubernetes
that
it
is
managing
has
an
agent
that
only
calls
out
at
no
point
does
that
kubernetes
need
to
open
a
port
for
pertainer
to
reach
back
in
So
security
wise.
It
should
be
relatively
safe,
and
this
is
showing
all
the
pods
that
are
running
I,
guess,
I'll,
just
arbitrarily
pick
the
kubernetes
cluster
API
server,
you
can
drill
down
into
here.
D
I
can
get
logs
for
this.
For
this
pod.
D
Some
of
this
can
be
made
more
responsive
because
when
you
set
this
thing
up,
you
define
a
ping
rate
for
how
often
this
managed
remote
node
hangs
into
the
portainer
I
believe
I've
got
it
set
for
five
seconds,
so
anytime
I
would
do
something
like
try
to
access
these
logs.
I
might
have
to
wait
up
to
five
seconds
before
that.
Node
is
even
aware.
I've
got
a
request
in
so
kind
of
that
Square
equivalent
of
hourglass.
D
There
is
expected,
even
if
you're,
on
a
fast
Network-
and
this
is
all
on
gigabit
e
inside
my
home
lab.
So
it's
running
pretty
fast
compared
to
perhaps
what
you
might
actually
find
in
an
edge
network,
but
you
can
see
I
can.
Even
if
this
kubernetes
node
was
out
there
somewhere
at
an
edge,
I
can
drill
down
and
even
look
at
individual
pods
drill
down
into
logs.
D
D
So
cubecon
will
get
pods.
So
all
in
all,
it's
pretty
plausible
in
terms
of
what
do
you
do
after
you
remotely
deploy
the
kubernetes
on
the
Headless
node
now,
I'm,
not
sure
what
the
experience
would
be
at
scale
right
I
mean
you.
If
I've
got
10
000
of
them
out
there
I
don't
necessarily
want
to
be
using
a
GUI
for
this,
but
you
know
it
it.
D
It
at
least
is
I,
don't
know,
I
I
think
it's
more
than
better
than
nothing.
It
actually
looks
like
a
pretty
useful
tool.
So
it's
interesting.
A
How
they
are
when
you
opened
up
the
shell,
it
looks
like
the
fourteener
agent
deployed
a
pod
to
enable
this
shell
execution.
I,
don't
know
if
you
noticed
that,
but
there
was
a
pod
that
had
only
been
running
for
29
seconds,
that.
D
D
So,
even
though
that
says,
Cube
cuddle,
shell
I
believe
you
can
actually
run
Helm
at
a
command
line
in
it
and
then
I
inquired
the
particular
kubernetes
distro
I
used,
which
was
the
tons
of
one,
has
its
own
CLI
and
I,
asked
about
what
it
would
take
to
support
that,
and
they
said
that
this
thing
actually
runs
as
a
sidecar
and
that
what
you
could
do
is
put
together
a
custom
version
of
this
container
to
enhance.
What's
in
this
Cube
cuddle
shell,
they
don't
give
you
raw,
complete
access,
probably
for
security,
but
yeah.
D
That
is
that,
in
fact,
is
how
you
install
the
portaner
agent
that
use
Cube
cuddle
to
lay
it
down
on
a
kubernetes
with
some
creds
and
IDs,
so
that
it
knows
which,
where
to
find
the
portaner,
that's
going
to
manage
it
so
anyway,
that
was
my
goal.
I
just
you
know.
If
time
had
allowed,
I
would
have
shown
this
in
that
presentation,
but
it
didn't
but
I
figured.
It
was
worth
five
minutes
to
drive
it
around
the
block
for
an
audience
so
yeah.
It
is
what
it
isn't.
D
I
found
it
pretty
easy
to
install
and
use.
So
unless
somebody
wants
to
see
anything
else,
I'm
gonna
stop
the
share.
D
B
On
a
chapters
so
do
we
have
any
yeah,
as
I
said
in
it's
like
I
dropped
the
ball
a
little
bit
on
that
in
the
last
few
weeks.
But
now
it's
a
it's
a
it's
a
good
time
to
pick
it
up.
Yeah.
D
Having
done
Charters
over
in
kubernetes,
my
suggested
technique
is
to
find
somebody
else
who
recently
did
one
on
another
project
and
just
go
use
the
use
them
as
a
boilerplate
that
you
would
go
at
it
from
because
it
and
you
know,
don't
use
one.
That's
still
in
progress,
use,
one
that
got
accepted
and
there's
sort
of
an
implication
that
it
complies
with
whatever
the
expectation
is.
D
Because
otherwise,
I
find
that
on
kubernetes
you're
looking
at
these
Community
documents
that
describe
what
goes
in
a
charter,
but
they
often
didn't
have
an
actual
example.
So.
A
So
I
mean
in
that
case
we
could
look.
If
we
want
to
focus
on
ones
within
runtime,
we
could
look
at
the
container
orchestrated
device,
one
I
will
say
it
seems
pretty
specific
too
it's
because
it
is
definitely
a
working
group
that
had
a
deliverable
to
so
we
can
look
at
that.
A
The
Bash
one
is
in
progress
like
you
were
saying
so
yeah.
We
could
also
look
at
I,
don't
know
if
y'all
know
of
other
working
groups
and
other
tags
that
maybe
we
can
look
at.
D
Yeah
I
found
I
haven't
gone
and
investigated
it,
but
if
you
know
where
those
live,
maybe
you
could
put
a
link
to
one
there
or
open
one
up
and
share.
A
Them
yeah
I
haven't
been
pulled
up
right
now.
Cool
I'll
just
share
my
screen
can.
B
D
A
Okay,
so
this
one,
the
container
orchestrated
device
working
group-
this
is
the
one
that
I
actually
went
to
this
morning,
so
they
put
together
CDI
or
the
container
device
interface,
which
is
supported
by
container
D.
Cryo
I,
don't
know,
oh
I,
don't
know
exactly
the
list,
but
a
bunch
of
run
times
and
what
it
basically
does
is.
A
It
allows
you
to
more
declaratively,
request,
access
to
resources,
access
to
device
or
heart
physical
hardware,
for
your
containers
and
it's
not
kubernetes
specific,
and
then
this
working
group
has
kind
of
gone
beyond
that
now
and
it's
creating.
It
has
proposals
for
creating
Dynamic
resource
allocations,
so
they're
essentially
going
to
enhance
the
kubernetes
Pod
spec
so
that
you
can,
just
as
you
would
add,
storage
volumes.
You
could
add
devices
in
that
way,
and
so
it's
basically
providing
an
alternative
to
the
device.
A
Plugin
interface,
where,
instead
of
requesting
extended
resources
created
by
device
plugins,
you
would
request
these
Dynamic
resources
and
it
would
look
different,
and
so
they
have
a
whole
kep
and
it's
been
a
process
they've
been
undergoing
for
a
while
I
think
it's
led
by
a
few
folks
from
Intel
and
so
there's
definitely
resources
on
those
that
I
can
drop
in
the
notes,
but
really
interesting
work.
There
that's
very
much
in
progress
and
they
presented
their
work
was
referenced
at
kubecon.
A
In
a
couple
talks,
so
anyways
you
can
see
that
they
kind
of
explain
their
problem,
say:
space.
D
A
Not
really
knowing
how
to
handle
like
Hardware
like
gpus
and
fpgas,
across
different
orchestrators,
so
across
kubernetes
and
Nomad
and
docker,
and
wanting
to
create
a
consistent
experience
kind
of
as
their
statement.
D
Yeah
one
thing
I
found
that
I
like
in
certain
other
groups,
are
sigs,
is
both
a
statement?
I
don't
see
it
here
so
far,
but
maybe
it's
in
here
somewhere
of
even
a
firm
definition
of
what
the
scope
of
their
coverage
is
and
also
as
important
what
it
isn't.
You
know
the
things
that
they
don't
want
to
cover.
I,
don't
see
it
in
this,
but
I
I
think
it
might
not
be
a
bad
idea.
Yeah.
A
It's
not
in
that,
but
that's
a
good
transition
to
the
one.
That's
in
progress
for
the
runtime
tag,
which
I
know
it's
not
completed,
but
for
they
do
exactly
that
they
have
their
goal
in
scope,
operations,
communities,
it's
very
brief,
and
it's
it's
in
progress.
So
maybe
that
brevity
is
why
it's
still
in
progress
or
maybe
that's
fine-
and
we
could
do
this.
We
could
create
something
quite
brief
and
then
yeah
build
from
there
in.
A
A
Yeah
also
look
at
our
current
one,
like
what
does
it
say
after
kubernetes
currently
yeah,
it.
D
A
Yeah
and
and
I
guess,
I
mean
that
makes
sense.
Like
hashicorp
has
a
lot
of
stuff
they've
contributed
to
kubernetes
and
as
well
like
their
vault
plugins.
D
Yeah
in
the
early
days,
I,
don't
think
the
cncf
tried
to
broaden
the
orchestrator
spec
to
something
like
Apache
mesos,
but
it
was
maybe
because
the
Apache
Foundation
already
owned
it.
So
I,
don't
remember
seeing
things.
The
only
thing
like
that
that
went
cross
orchestrator
was
the
CSI
project,
but
it
sort
of
lives
off
on
its
own.
D
But
it
would
actually
be
an
interesting
topic
if
we
get
scope
broader
than
kubernetes
over
our
new
CNC
at
foam
to
even
have
a
presentation
on
using
Nomad
at
Edge
mm-hmm.
If
that
clear,
that
almost
would
have
been
like
Heretics
in
the
in
the
church.
Trying
to
do
that,
while
we
were
under
the
kubernetes
project,
but.
A
Yeah
no
I
think
that's
an
option,
I
think
with
a
web
assembly,
some
of
those
Solutions
they're
using
Nomad,
because
nobody
has
better
support
for
webassembly.
So
there's
interesting
stuff.
There
too.
A
Okay,
yeah,
that's
a
good
note.
So
I'll
just
put
discussion.
D
B
A
Well,
okay,
so
Let's
do
we
want
to
just
do
a
quick
brainstorm
and
continue
with
the
discussion
of
what's
in
scope
and
what's
out
of
scope
for
us,
so
that
was
one
thing
that
you
mentioned,
and
it's
also
something
that
the
fascist
and
initiative
Marketing
Group,
so
I
think
we
can
come
up
with
an
introduction
and
background
background
can
can
include
how
we
used
to
live
in
kubernetes,
I
think
maybe
goal
it's
interesting.
A
How
to
distinguish
introduction
background
and
goal
I,
don't
know
if
y'all
have
immediate
thoughts
on
that,
but
those
are
definitely
could
be
some
overlap.
There.
D
B
D
Think
it
would
be
to
you
know,
Assist
users
and
act
as
a
a
forum
for
users
to
exchange
ideas
and
best
practices,
but
then,
in
addition,
bring
up
the
fact
that
we
May
intend
to
go
off
and
do
white
papers
reference
architectures,
those
kinds
of
things
as
kind
of
models
for
how
someone
might
do
things.
D
B
D
At
some
point
we
could
even
potentially
I
don't
I,
don't
know
how
often
we're
going
to
come
into
venues
where
there's
critical
mass.
But
you
know
over
in
the
kubernetes
project
they're,
perhaps
trying
to
investigate
bringing
back
the
whole
Meetup
thing
of
local
meetings
in
big
metropolitan
areas
to
cover
areas.
I
don't
actually
know
if
Edge
would
get
enough
critical
mass
that
you
could
get
a
physical
meet
up
going,
but
maybe
in
certain
cities
you
could.
A
Cool
yeah,
so
I
think
introduction
and
background
we
can
come
up
with
and
then
goals
it
sounds
like
we
have
multi-goals
I
think
we
can
almost
bullet
it
like.
You
were
saying
users
for
them
to
exchange
and
issues
and
best
practices.
Then
it's
also
to
put
content
out
there
in
the
form
of
white
paper
as
a
reference,
Archer,
there's
and
then
third
of
all,
it's
to
organize
community
in
the
form
of
conferences
and
meetups,
and
this
working
group
in
general.
So
it
seems
like
we
have
a
three
faceted
yeah.
A
D
Yeah,
the
other
one
that
has
a
pretty
good
background,
actually
was
kind
of
the
I
I
believe
that
the
description
for
the
kubernetes
on
edge
day
it
maybe
is
pretty
kubernetes
specific,
but
it
has
some
nice
text
on
kind
of
the
scope
of
edge.
That
might
be
a
decent
starting
point
for
putting
you
know
for
making
part
of
the
background.
A
Uh-Huh
yeah
I
think
I,
don't
know
where
the
the
charter
is
not
linked
to
this
mark
down
here.
D
A
I
mean
we
have
our
white
paper.
If
that's
what
you're
thinking
about
too,
like
the
original
white
paper
and
then
there's
also
the
security
one.
D
A
B
D
A
I
was
just
gonna
say
anything
on
you
created
one,
so
I'm
taking
notes
here,
and
we
can
Port
this
information
over
to
there.
Foreign.
A
And
then
we
also
have
like
five
more
minutes
too.
Is
there
any
top
of
Mind
in
scope
and
out
of
scope
that
folks
have.
D
Well,
I
think
kubecon
and
it's
related
to
inscope
I
think
the
idea
of
covering
device
management.
You
know
both
the
onboarding
and
the
compute
device,
as
well
as
the
I
o
devices.
The
Kate
covered
really
seemed
to
draw
quite
an
audience
and
quite
a
level
of
Interest,
so
that
clearly
should
be
called
out
as
in
scope.
D
D
Well,
I
think
we
should
be
careful
to
not
overlap
other
groups,
so
you
know
we
don't
cover
generic
web
assembly,
for
example.
So
maybe
if
we
can
come
up
with
what
the
divining
line
is
and
it
could
even
be
kind
of
a
broad
Gray
colored
dividing
line,
but
we
still
should
have
some
working
idea
of
you
know
how
suppose
somebody
wants
to
come
in
and
just
talk
about
webassembly
in
general.
Is
that
appropriate
for
this
group
or
not?
Maybe
it
isn't,
but
we
should.
We
should
not
have
to
make
that
call
in
advance.
A
Yeah
I
think
the
line
is
literally.
Where
is
it
running
and
where
are
you
focusing
your
efforts?
Kind
of
on
that,
like
Linux
Foundation,
like
Edge
zones,
if
you're
putting
web
assembly
on
the
user,
Edge
or
and
then
also
this
is
a
Steve,
something
that
we
came
across
as
people
asking
about
kind
of
that
Network
networking
like
Network
message
is
that
edge?
A
If
that
isn't
it's
not
in
the
cloud
necessarily
it's
more
of
like
kind
of
in
the
fog
area,
but
they
are
huge
servers
and
network
specific,
usually
network
provider,
specific
I
mean
so
we've.
D
Had
this
come
up,
even
in
the
kubernetes
on
edge
day
on
the
review
of
cfps,
where
people
will
come
in
with
like
a
totally
generic
service
mesh
talk
but
put
in
a
one-liner
saying.
Yes,
this
is
service
mesh,
but
you
could
run
it
at
Edge,
but
really
the
whole
talk
is
so
generic
that
there's
no
aspect
of
it.
That's
unique
to
Edge
so
I
think
within
the
networking
Realm.
D
We
need
a
way
of
calling
out
the
you
know.
Edge
has
some
unique
demands
related
to
connectivity
and
things
that
deal
with
that
directly
clearly
are
in
scope,
but
just
generic
Network
topics.
Probably
not
you
know.
Let's
suppose
somebody
came
in
and
wanted
to
give
a
talk
on
IPv6
with
the
angle,
the
GE
Edge
is
going
to
need
a
whole
bunch
of
new
addresses.
Therefore
this
is
related.
Edge
I.
Don't
think
that
cuts
it
myself,
you
know
I,
don't
I,
don't
think
it
would
belong.
B
D
B
D
Was
just
brought
up
today
of
the
the
the
accelerators
for
machine
learning?
You
know
a
talk
that
covered
the
kinds
of
devices
that
are
made
for
Edge
low
power,
low
cost.
You
know
things
that
might
run
on
arms
at
Edge
nodes
versus
things
that
consume
6,
000
watts
and
need
16
Lanes
of
PCI
Express.
You
know,
I
I
think
we
can
maybe
stake
a
claim
on
I.
Don't
know
almost
generically
call
out
that
these
broad
areas,
like
networking,
machine
learning,
acceleration,
Hardware
Etc
they're
in
scope.
D
Only
if
what's
being
discussed,
is
specifically
Edge
related.
A
Yeah
I
think
a
good
way
we
could
do.
That
is
for
our
in-scope
out
of
scope.
Maybe
we
have
one
in
scope
line.
That's
like
machine
learning
with
regards
to
devices
with
constrained
compute
and
air
gaps,
connectivity,
blah
blah
like
give
a
line
yeah
and
then
the
out
of
scope
would
say
like
machine
learning
in
general,
like
just
kind
of
give
that
example
uh-huh.
D
A
D
D
You
know,
I
think
it's
been
my
experience
that
the
bulk
of
our
Edge
discussions
have
dealt
with
fixed
location
Edge
as
opposed
to
mobile,
like
things
running
in
the
car
or
a
truck,
nothing
against
covering
them,
and
there
is
some
overlap,
but
there's
also
a
lot
of
unique
aspects
to
that
come
into
play.
When
the
you
know
your
Edge
is
actually
moving
around.
B
I
I
think
the
main
like
a
glue
to
all
this.
Is
that,
like
we
can
talk
about
machine
learning,
we
can
talk
about
embedded
devices
mobile
devices,
but
it
has
to
be
in
the
context
of
of
integrating
that
with
some
kind
of
a
Cloud
solution,
because
to
me
that
that
makes
sense
like
that
defines
that
this
is
the
edge
of
the
cloud
system.
So
it's
not
the
is
Computing
on
its
own.
It's
not
embedded
development
on
its
own.
It's
Building
A
system
that
that
tied
those.
D
B
Yeah
but
but
it
still
has
a
cloud,
it's
just
just
a
yeah
local
Cloud
right.
B
It
has
to
have
some
kind
of
like
a
cloud
native
experience
for,
for,
for
yeah,
admit
users
and
administrators
of
those
kind
of
systems.
Okay,.
D
D
A
Yeah
I
think
that's
a
good
point,
though
this
is
within
the
cncf.
It
must
have
some
sort
of
cloud
native
Computing
experience,
whether
that
is
an
orchestration
experience
due
to
the
size
of
the
number
of
servers
that
you
have
or
whether
it's
literally
integrating
your
Edge
system
with
your
cloud
system,
these
two
separate
environments,
I
think
that's
a
good.
That
was
a
good
ad.
D
D
A
A
And
it's
in
markdown,
even
though
this
is
a
Google
doc.
So,
and
also
this
is,
is
this
public
dehon
can
I?
Add
it
to
the
agenda,
notes.
A
Okay,
you're.
A
A
I
added
it
there
great
yeah,
maybe
we
I
think
by
the
next
meeting.
It
would
be
good
to
have
that
PR
up
and
I
think
that's
feasible.
I,
don't
know
what
are
y'all's
thoughts.
A
Great,
we
could
even
aim
for
earlier
yeah
yeah.
D
B
A
Yeah,
maybe
we
go
ahead
and
create
like
a
kind
of
have
slots
at
the
top
of
our
document,
where
people
can
kind
of
sign
up
and
choose
one,
because
I
think
the
the
container
orchestrated
device
folks
were
interested
in
presenting
that
kind
of
is
the
opposite
of
what
we
were
saying
about
scoping
away
from
other
working
groups.
So
I
don't
know
if
we
want
to
do
that,
but
I
think
it'd
be
interesting.
Yeah.
A
Yeah
I'll
yeah
I'll
reach
out
to
them
then
and
say
that
they
have
the
next
one
and
then,
if
they
say
no,
we
can
then
they
want
another
couple
weeks.
We
can
pass
it
off.
Okay,.