►
From YouTube: Kubernetes WG IoT Edge 20230125
Description
January 25, 2023 meeting of the CNCF IoT Edge Working group. Discussion of group activity and general bords of a feather discussion including discussion of the choice of a Kubernetes distribution for a robot application
A
A
Edinative
white
paper
status,
review
I-
think
Brandon
may
have
put
that
on,
but
I
don't
think
he's
here
yet,
but
I
expect
he
will
be
so.
Maybe
we'll
shift
this
to
a
later
part
of
this
unless
somebody
else
was
responsible
for
that.
A
Just
a
quick
reminder:
I'll
cover
right
now
in
line,
but
there
will
be
one
of
the
kubernetes
on
edge
days
at
kubecon,
Europe
in
April
and
because
the
cncf
is
cutting
back
the
scale
of
the
event
it's
even
though
it's
called
day
it's
going
to
be
a
half
day
as
I
understand
it.
So
we
won't
have
as
much
time
for
presentations
but
I
think
they've
already
started
to
come
in.
A
So
if
you're
interested
in
speaking
at
kubecon
and
the
talk
is
related
to
Edge,
it's
too
late
to
get
it
into
the
main
kubecon,
except
for
the
maintainer
track
session
sessions,
but
The
Edge
day
is
still
open
until
I
think
February,
the
5th
I.
A
Right:
okay:
well,
then,
you've
got
even
more
time
so,
but
keep
that
in
mind
if
you've
got
anything
you
wanted
to
talk
about.
Some
of
us
on
this
call
have
been
on
the
Review
Committee,
so
whether
they
call
it
out
in
the
cfp
or
not,
you
can
feel
free
to
submit
either
what
are
called
full
length
sessions,
which
I
think
are
expected
to
be
35
minutes,
plus
q
a
or
if
you've
got
ideas
for
lightning
talks.
Often
one
of
the
segments
is
put
together
with
those,
so
either
one
should
be
workable.
A
A
But
we're
still
going
on
this
and
I
added
a
whole
bunch
of
things
that
I
found
through
Google
search
or
just
personal
awareness,
got
it
up
to
101
rows
and
quit
I
think
there
are
actually
more
of
them
out
there,
but
I
didn't
have
time
to
fill
in
all
of
these
links
to
the
GitHub
repo
or
a
concise
phrase
of
what
the
project
is
about.
A
But,
strangely
enough,
they
don't
seem
to
have
comprehensive
portals
of
what
these
are
I
think
part
of
it
is
they
come
out
of
different
business
units
within
these
organizations
and
nobody
at
the
top.
You
know
at
the
top
level
they
promote
their
commercial
projects
or
products,
but
not
so
much
the
open
source
ones
so
they're
a
little
harder
to
find.
But
if
some
of
you
who
are
affiliated
with
these
organizations
are
used
to
be
from,
there
are
aware
of
those
lists.
A
Maybe
you
could
just
take
a
quick
list
and
a
quick
review,
and
even
if
you
don't
have
time
to
fill
it
out,
just
put
the
name
there,
because
once
the
name
is
there,
somebody
can
come
along
behind
and
fill
out.
The
details
I
realize
having
pursued
it,
that
it
sounds
easy
to
go,
collect
all
of
these
things
like
the
dock
like
the
source
link,
but
the
you
start
doing
it
and
the
next
thing
you
know
an
hour
has
been
burned
and
you
filled
out
about
10
rows
worth
so
it
does
take
some
time.
A
Okay
back
to
the
agenda,
maybe
I'll
turn
this
over
to
to
you
down
for
the
maintainer
track
proposal
for
kubecon
Europe.
C
Sure
so
I
think
the
deadline
is
on
Friday
and
we
still
didn't
come
up
with
with
with
a
topic
that
we
would
like
to
propose
if
any
I
I
previously
offered
to
to
do
a
session
on
on
applying
as
native
principles
in
practice
and
and
do
a
demo
about
it.
But
I
need
to
back
off
from
that,
because
I
got
some
a
lot
of
things
on
the
plate
at
the
moment
and
I.
Don't
think
I
would
be
able
to
prepare
it
properly
for
for
April.
C
Yeah
just
saying
we
didn't
have
too
much
other
proposals,
ideas,
I,
think
Kate.
You
had
idea
about
the
panel
which
which
ill
can
do
if,
if
he
can
find
enough
people
to
do
so
and
I
I
would
just
like
to
open
it
up
for
discussion.
What
everybody
thinks
we
could
have
I'm.
A
I'm
still
pot
working
on
the
possibility
of
panelists
but
I
think
it's
a
stretch
that
I'm
going
to
get
them
by
Friday.
The
you
know.
The
tough
thing
with
a
panel
is
that
once
you
get
over
two
people,
there's
diversity
requirements
too,
and
I
actually
track
down
some
people
who
are
interested
but
I,
don't
it
it
doesn't
meet
the
diversity
requirements.
A
So
if
we
don't
get
it
in,
it's
not
the
end
of
the
world.
To
be
honest
under
the
new
reduced
length,
kubecon
they've
already
telegraphed,
they
cut
it
down
from
five
days
to
three
and
then
formerly
the
maintainer
track.
Proposals
were
fairly
High
likelihood
of
acceptance,
but
they've
telegraphed
that
it's
going
to
be
much
tougher
now,
because
they've
cut
two
days
out
of
the
conference,
so
I
think
it
would
be
a
long
shot
to
get
it
in
unless
it's
excellent
and
if
people
have
ideas
and
want
to
play
a
part
in
it.
A
A
We
can
still
get
a
presence
in
that
kubernetes
on
edge
day,
because
I
think
you
know
we
can
communicate
the
existence
of
this
group
and
I
think
some
of
us
are
likely
to
be
attending
that
event
regardless,
so
we'll
see
how
it
goes.
But
once
again,
if
anybody's
got
any
ideas
for
talks
to
submit
to
kubecon
under
the
mahainer
track
in
order
to
fit
the
rules,
it
can't
just
be.
You
know
any
old
Edge
related
talk.
A
It
really
should
be
somewhat
related
to
this
group's
activity,
and
also
I
can
tell
you
that,
having
been
on
program
committee,
they
don't
like
drop-in
speakers
that
haven't
been
showing
up
regularly
to
the
working
group
or
Sig
meetings
under
the
auspices
of
maintainer
track.
So
if
you
know
somebody
who's
got
a
great
talk,
but
they've
never
showed
up
for
a
meeting
in
this
group
before
it
likely
would
not
fly
as
a
talk
submitted
as
maintainer
track
under
this
group.
B
I
think
before
we
push
this
to
slack,
I'd
be
curious
for
folks
who
are
on
this
call,
even
if
we
don't
make
this
happen
before
Friday
I
think
this
is
an
important
time
to
discuss
what
we
envision
for
these
talks
in
the
future,
namely
maybe
North
America
so
other
folks
on
the
call
I'm
curious.
If
you
were
to
attend
a
talk
that
is
kind
that
is
backed
by
this
working
group.
What
would
you
anticipate
it
being
about?
B
Would
you
want
it
to
be
about,
or
is
there
just
an
edge
talk
in
general
that
you
think
is
Broad
in
scope
that
you'd
be
interested
in
hearing
about
foreign.
D
So
I
guess
one
of
the
things
I'd
be
interested
in
seeing
would
be
an
application
of
of
the
technology,
so
something
that
either
showed
an
end
use
case
and
what
it
was
doing,
or
some
sort
of
application
running
on
the
device
or
in
the
cloud.
D
A
Yeah
I'll
concur
I'm,
not
exactly
in
the
shoes
of
an
end
user
being
with
a
vendor,
but
I
have
seen
a
lot
of
these
enough
to
gauge
attendance
and
I.
Think
that
what
we
historically
it's
been
a
little
difficult,
but
I
think
that
the
best
talks
for
drawing
a
crowd
are
end
user
case
studies
where
and
and
things
where
they're
not
using
just
one
thing,
you
know
it
a
lot
of
times.
A
Vendors
will
line
up
somebody
who
uses
their
product
and
trot
them
out
there
really
as
kind
of
a
hidden
sales
pitch
or
a
not
so
hidden
sales
pitch.
But
someone
who
is
walking
in
the
shoes
of
a
user
who's
combining
five
or
six
different
things
coming
out
of
Open,
Source
and
practically
applying.
It
are
the
kinds
of
things
that
I
think
people
want
to
hear.
It
has
credibility.
A
They
often
tend
to
be
presentations
where
they're,
not
sales
pitches
and
they
go
into
both
the
good
and
bad
experiences
encountered
and
I
would
love
to
do
that.
But,
sadly,
a
lot
of
those
user
Things
fall
into
these
categories.
A
If
somebody
has
contacts
for
a
user
who
would
like
to
talk
and
by
the
way,
kubecon
North
America
I
think
is
in
the
third
quarter,
like
October
November,
if
I'm
not
mistaken
in
Chicago,
so
off
times,
you
know
if
you
can
get
somebody
who's
talking
who's
kind
of
local
to
the
area.
At
least
the
travel
costs.
A
Aren't
that
great
and
with
signs
that
perhaps
the
world's
headed
into
a
recession,
I
know
that
being
an
organizer
of
the
scale
conference,
we're
having
trouble
with
a
few
of
these
speakers
who
got
accepted
coming
in
late,
saying
that
the
company
has
imposed
travel
freezes,
and
you
know
it
wouldn't
be
outlandish-
I
think
to
expect
some
of
those
difficulties
both
with
kubecon
Europe
and
North
America
later
this
year.
Unless
some
things
turn
around
foreign.
A
Let's,
unless
there's
any
final
comments
or
thoughts,
let's
close
out
the
talk
on
the
maintainer
track
proposal,
we
still
had
someone
put
in
the
white
paper
status
review
if
someone
has
shown
up
who
wants
to
talk
about
that,
I'm,
not
sure
who
put
it
on
the
agenda.
B
I
didn't
add
it,
but
I'm
happy
to
add
some
words
to
that
I
think
a
week
ago,
or
maybe
two
weeks
ago,
Brandon
mentioned
via
email
having
a
connection
with
the
kubecon
kind
of
media
and
marketing
staff
about
promoting
the
paper,
but
that
we
just
needed
to
get
final
revisions
in,
and
so
he
put
together
the
final
revisions
and
I
pushed
him
up
to
GitHub
and
they
got
merged
and
I.
That
was
the
end
of
the
action
item.
I
thought
at
that
point.
B
I
think
he
was
going
to
take
it
to
the
cncf
marketing
folks,
so
I
I
followed
up
with
him
about.
If
that
happened,
but
haven't
heard
back
so
I
think
it's
on
its
way
to
potentially
being
pushed
out
into
a
cncf
Blog,
but
I
think
we'll
need
to
hear
more
about
that
from
Brandon
foreign.
A
If
we
run
out
of
items,
we
usually
turn
this
into
a
birds
of
a
feather
open
discussion,
but
one
thing:
I
just
wanted
to
I'll
prime
the
pump
for
that
I
guess,
because
tamoya
I
noticed
that
you
put
a
comment
on
the
slack
recently
about,
but
working
on
a
proof
of
concept
with
open
year
and
I'm
kind
of
curious
to
hear
about
that,
and
also,
if
you
want
to
tell
us
any
updates
on
what's
been
going
on
lately
with
the
the
Robot
Operating
System
and
the
things
you've
been
working
on.
D
E
Mentioned
like
the
elephants
products,
so
you
want
us
to
add
the
encryption
there
will
like
in
this
photo
shoot.
Yes,.
A
I,
would
you
can
add
descriptions
to
the
projects
that
are
already
there
and
if
there's
any
projects
missing,
go
ahead
and
add
new
rows
at
the
end,
we're
going
to
organize
them
by
category
later,
but
we
thought
we'd
just
get
the
list
first
and
then,
once
we
get
the
list,
we
can
see
what
categories
these
things
are
coming
into.
A
You
know
we
we
already
know
that
some
categories
are
things
like
operating
systems
and
orchestrators,
but
we've
got
categories
more
related
to
connectivity
and
data
Communications
device
management,
but
I
think
it
rather
than
first
come
up
with
categories
and
then
fit
them
in
we'll
just
get
the
list
of
everything
out
there
and
then
see
what
they
break
into.
If
that
makes
sense,.
A
And
it
is
moving
forward,
I
think
that
thing
started
out
a
couple
months
ago
and
we're
up
to
100.,
my
gut
feel
is
there's,
maybe
150
of
them
out
there.
Although
some
of
these
projects,
some
of
them
I
left
out
because
as
I
was
looking
up,
the
data
I
discovered
that
you
know
you
go
to
their
get
repository
and
nobody's
touched
the
thing
for
six
years
and
I
think
anything.
That's
sort
of
in
that
category
of
archived
dead,
no
longer
maintained
they're.
A
It's
just
pointless
to
add
those
kinds
of
things
on
the
list,
because
you
waste
people's
time
because
ultimately
they're
like
even
if
they
discover
it
they're,
never
going
to
use
it.
One.
One
thing
that
we
might
add
that
is
more
work
is
even
like
a
health
assessment
of
I
was
thinking
after
I
got
into
it
that
if
there
was
a
column
for
when
the
last
update
occurred
or
a
metric
like
how
many
PRS
are
being
urged
per
year,
that
would
be
a
quick
pass
to
a
limit.
You
know
to
kind
of.
D
A
B
E
Okay
thanks
so
the
that
is
like
a
white
paper,
the
status
review.
So
what's
the
major
difference
or
out
there
for
our
last
time,
I
think
I
need
time
to
take
a
look
at
it.
But
you
describe
like
a
major
update
in
a
row
that
would
be
helpful.
A
E
E
Okay,
that's
the
VPN,
so
yeah
that
question
was
H
native
white
paper,
so
I
was
asking
what
is
like
a
major
update
for
last
time,
so
I
mean
I.
I
need
I
need
time
to
take
a
look
at
it.
But
if
you
describe
like
an
overview
like
it
would
be
helpful
to
understand,
what's
the
difference,
if
anybody
has
any
idea.
B
From
last
time,
there's
not
a
big
difference.
It's
what
we
did
as
a
syntactical
read
through
so
make
sure
commas
are
where
they
should
be.
Grammar
is
appropriate
if
you've
read
the
content
before
or
been
a
part
of
the
writing
of
it
there's
not
much
new
there.
The
update
wasn't
of
the
content,
it's
a
of
the
what
we're
doing
with
it
is
more
of
the
update.
So
the
update
is
that
we're
undergoing
the
process
of
putting
it
out
into
more
of
a
public
sphere
via
the
cncf
blog.
E
A
So
we
left
it
out
there
for
a
comment
period
just
to
see
if
there
were
any
comments,
indicating
serious
omissions
or
errors.
But
since
we
didn't
get
any
of
those
change,
requests
that
got
promoted
from
the
draft
stage
to
the
I
don't
know,
I'd
hesitate
to
call
it
final,
because
you
know
in
a
rapidly
changing
field
of
Technology.
What's
ever
final,
but
it's
at
least
not
a
draft.
It's.
A
B
We
changed
the
wording
to
first
edition
with
the
idea
that
we
might
have
revisions.
So
this
is
the
first
edition
of
the
paper,
but
it's
no
longer
a
draft.
That
was
that
may
be
arbitrary,
but
that's
kind
of
what
we
went
with.
E
E
E
Yeah,
so
I
was
invited
by
Alibaba,
so
I
think
the
Alibaba
cloud
and
I
I
had
a
I
had
a
presentation
at
Alibaba
cloud:
developer,
Summit
2023
at
Jakarta,
Indonesia
and
I.
Think
the
my
talk
is
kind
of
like
an
exception,
because
everybody
else
was
talking
about
Alibaba
Cloud,
because
you
know
it's
the
hosted
value
of
us,
so
it
should
be.
The
talk.
E
Talk
should
be
I
thought,
algebra
crowd,
but
I
just
talked
about
all
priorities
because
we
have
been
working
on
the
POC
and
evaluation
that
how
it
works
using
like
AWS
cloud
and
Edge
device
is
connected
to
like
internet
and
using
the
open
yard
and
I
was
working
on.
The
contribution
to
you
know,
put
back
everything
into
the
main
line.
So
there's
a
like
a
offer
from
opening
of
the
team
that
hey.
Why
don't
you
talk
about
open
Earth?
That's
the
history
and
are
you
sharing
screen.
A
A
E
So
yeah
so
the
background
that
we
have
in
like
it's
so
HD
device
getting
matured
and
it's
going
to
be
more
distributed
than
Collective
system.
The
we
have
use
cases
like
it's
device
like
a
live
streaming:
entertainment
like
a
TV
station
and
automatic
autonomous
driving
this
kind
of
stuff.
So
that's
the
use
case
so
and
the
the
key
part
that
I
think
from
the
system
perspective
is
like
hdbls
system.
E
Integration
is
really
important
for
us,
because,
even
if
the
faction
is
almost
same,
we
have
to
integrate
the
application
to
specific
platform
every
single
time.
So
we
don't
want
to
do
that
so
taking
advantage
of
the
container
or
whatever
you
know
like
to
provide
the
capability
for
the
application
modularity.
E
So
that
that's
that's
the
background
and
the
goal
is
like
that
something
we
expect
is
like
in
the
factory.
A
We
can
hear
you,
but
your
your
audio
is
cutting
in
and
out.
We
can
see
the
presentation,
but
the
audio
has
had
periods
where
we
think
I
think
we
lost
chunks
of
about
10
seconds.
So.
E
C
C
E
So
the
the
something
that
we
want
to
do
in
one
world
is
like
this
is
like
a
typical
like
application,
that
deployed
to
like
a
robotics
or
robots.
E
So
the
the
there's,
a
hardware
and
operating
system
getting
a
sensor
from
sensor
devices
and
we're
going
to
deal
with
the
sensing
data
operation
recognition
and
we're
going
to
do
the
instant
process
based
on
the
user
preference-
or
you
know
like
so
like
that
and
to
send
the
data
back
to
like
Edge
device
or
something
like
that
to
do
navigation
of
path
planning
and
we
actually
control
the
actuators
to
give
the
feedback
for
the
end
user.
So
this
is.
E
This
is
like
a
typical
like
a
robotics
application,
so
something
we
want
to
do
is
like
we
want
to
change
dynamically
use
case,
A
and
B.
So
the
use
case
a
that
describes
that
the
only
thing
that
processing
is
happening
is,
and
everything
else
will
be
processed
in
the
cloud
infrastructure.
But,
on
the
other
hand,
use
case
b
is
the
mode.
Let's
say
we
have
more
already
Associates
for
the
edge
and
maybe
we're
just
accessing
to
the
cloud
to
the
user
preference
or
something.
E
So
we
want
to
change
this
like
configuration
dynamically,
so
that
that's
the
the
basic
like
motivation
that
we
have
so
we
we
actually
I
didn't
mention
that
we
tried
kubets,
because
that's
that's.
That's
not
good
for
the
presentation
so,
but
we
have
tried
the
two
batch,
but
it
doesn't
work
because
the
kubets,
the
bridges,
the
kubernetes
API
server.
There
is
like
a
Cloud
Core
running
in
the
cloud
and
that
does
not
support
all
of
the
API
to
the
kubernetes
API.
E
So
in
that
case,
for
the
user,
sometimes
the
specific
cni,
for
example,
like
a
weave,
does
not
work
because
the
when
the
Pod
initializing
for
weave
it
it
will
access
the
kubernetes
to
get
the
network.
Topology
for,
for
example,
know
the
information.
So
if
that
happens,
if
there
is
no
consistency
between
kubernetes
API
to
bridge,
for
example,
like
a
Cloud
Core
from
kubets,
it
doesn't
work.
So
we
just
ended
up
not
using
kubets
and
fall
back
to
opioid.
E
The
the
reason
that
we
use
open
yard
is
it:
it
supports
kubernetes
API
as
it
is
so
we
can
actually
use
the
mainline
API
without
any
change.
So
that's
the
that's.
That's
the
reason.
Number
one
and
number
two
is
open
your
to
provide
the
thumbnailing
for
the
control
play.
So
in
that
case
you
can
connect
4-H
devices,
for
example
over
the
Internet
to
the
cloud,
for
example
like
kubernetes
API.
So
that's
that's
the
reason
that
we
choose
the
open
your
phone.
E
E
Yeah,
so
our
POC
will
be
something
like
this,
so
there
are
many
like
its
devices.
Actually,
this
POC
demonstration
is
like
a
live
streaming,
so
there
is
a
many
camera
devices
and
view
devices
and
connecting
to
the
single
cluster
and
some
application
running
in
the
cloud
infrastructure,
and
we
can,
after
this
cluster
is
established.
We
can
you
know
we
can
change
the
deployment
as
we
like
so,
for
example,
like
from
this
picture.
That
is
a
c
and
instance.
E
B
has
some
application
container
running
in
the
cluster,
but
if
we
want
to,
we
can
move
this
application
containers
to
the
Edge
Edge
servers.
So
in
that
case,
if
the
network
topology
is
shared
and
the
mesh
network
is
supported,
even
if
we
disconnect
the
the
internet
between
cloud
and
Edge,
but
we
can
robot
live
stream,
we
go
on
edge
environment.
So
that's
something
we
want
to.
You
know
like
achieve
and
see.
If
our
idea
can
work-
foreign,
yes,
so
I
think
so
far
so
good,
but
we
tried
cni
something
like
a
frontal
psyllium.
E
A
wave
just
works
out
of
the
box
and
the
open
year
provides
interoperability
with
kubernetes.
That
is
good
and
we
can
also
deploy
the
service
mesh
to
create
because
the
final
goal
is
to
you
know
like
to
create
the
services
for
the
business
Logics.
So
we
already
tried
to
use
ecl
sidecar
with
this
environment,
and
that
works
fine
and
we
have
some.
We
had
some
problems,
but
most
of
them
are
solved.
A
Oh
I
was
going
to
ask
a
question
on
your
previous
slide
on
those
other
cnis
that
you
called
out
you.
You
know.
You
said
that
we've
worked,
but
there
were
a
few
others
listed
like
flannel.
Did
those
not
work
or
you
just
didn't
get
around
to
using
them.
E
Oh
yeah,
so
the
application,
if
you
it
depends
on,
like
your
physical
network
configuration
so
the
weave
is
the
layer
2
emulation.
So
that
means
that
it
can
go
like
go
beyond
the
network,
but
I
don't
think
the
serial
on
the
funnel
is
not
a
layer
to
emulation
more
like
layer
three.
So
in
that
case
the
if
your
physical
network
connection
is
constructed
with
different
network
I,
don't
think
for
the
application
container
cannot
go
beyond
the
network.
E
E
So
the
problem
we
see
right
now
is
like
a
too
much
tax
for
the
system.
For
example,
like
kubes
open
your
Tonia
link
that
we
are
using
actually
like
a
Raspberry
Pi
4,
this
kind
of
like
armor
64
devices.
It's
not
really
a
resource
Trend
pretty
much,
but
we
try
to
use
raspberry
pi4,
but
constantly
it
consumes
like
a
10,
CPU
and
probably
15
megabyte
memory
allocation
or
something
so
thinking
about,
like
the
you
know,
like
a
business
use,
cases
that
we
have
right
now
is
too
much
tax.
E
We
have
to
pay
for
the
system
and
it
doesn't
do
any
good
for
the
application,
so
the
what
I
think?
What
do
we
think
right
now
is?
We
need
to
reduce
the
memory
footprint
and
more
performance
Improvement
for
the
CPU
consumption.
That
is
one
of
the
requirements
for
us
now
we
don't
know
if
we
will
reduce,
we
will
try
to
contribute
revision
to
the
open
yards
or
find
something
else,
but
at
least
using
opener
to
provides
that's
something
we
like
to
have
and
on
the
other
hand,
we
have.
A
Do
you
have
any
idea
of
where
that
CPU
was
burned
up?
For
example,
if
your
connectivity
is
using
TLS
and
you're
engaging
in
the
crypto
functions
on
these
low
powered
devices,
sometimes
that
burns
a
lot
of
resource,
but
I
don't
know
if
you've
got
that
going
on
or
not.
E
E
Yes,
so
the
other
thing
I
want
to
point
out
here
is
it's
not
only
for
open
yard
but
kubernetes.
So
once
it
comes
to
like
Roblox
and
the
robots
application,
the
the
platform
could
be
more
complicated,
so
the
device
obstruction
more
to
support
a
more
complicated
devices
to
bind
to
The
Container
runtime
would
be
required.
So
I
we
already
have
like
a
PR
for
the
device
plugin
that
provided
provided
by
kubernetes
and
the
device.
Plugin
works
okay,
but
it's
not
really
well.
E
You
know
well
this
designed
like
interfaces.
So
if
the
device
is
really
that's
simple
enough,
I
think
the
device
plugin
can
work
for
us,
but
we
we
have
like
a
proprietary,
much
more
complicated
devices.
That
requires
like
a
setting,
parameter,
pre-configuration
and
initialization,
and
when
it
shuts
down
there,
there
are
many
procedures
to
shut
down
that
specific
device.
So
in
that
case
we
cannot
support
this
device
via
device,
plugin
interface,
so
I
think
that
could
be
something.
Maybe
that's
the
contribution
to
the
kubernetes
male
live
I,
don't
know,
but
that's
the
situation.
E
Yeah
and
thank
you
very
much
Kate
that
you
provided
this
information
and
I
was
trying
to
get
some
attention
from
yeah,
so
most
of
them
are
like
idiopathic
people,
so
I
was
trying
to
get
some
attention
from
them
to
to
join
Hey.
If
you,
if
you
are
interested
in
kubernetes,
iot
perspective,
biology,
join
us,
that's
that's
what
I
did
mention
yeah.
A
E
The
actually,
we
actually
tried-
probably
five
six
like
cameras
and
some
viewer
devices
for
this
moment,
but
for
the
business
I
think
it
could
be
a
thousand,
because
the
use
case
that
we
have
is
like
an
entertainment,
live
streaming,
this
kind
of
stuff,
so
that
could
be.
That
should
be
able
to
fill
up
like
a
thousands
devices
on
the
edge.
F
E
A
Yeah
I
think,
if
you're
already
having
trouble
with
consumption
of
CPU
and
things
I,
guess
that
consumption
is
down
at
the
edge
note
right,
not
a
not
at
the
supervisory
control
up
at
the
for.
Are
you
running
the
kubernetes
control
plane
in
a
cloud
or
on
some
other
physical
mode.
E
At
this
moment,
the
control
planes
running
in
the
cloud.
B
Just
a
clarifying
question
you
may
have
already
mentioned
this-
you
mentioned
in
your
challenges,
slide
that
last
bullet
about
what
kubernetes
provides
but
open
year.
It
doesn't,
it
seems
to
be
summarizing.
Is
that
saying
that
the
device
plugin
interface
isn't
supported
by
open
Year
and
that
they
have
a
different
way
of
abstracting
and
finding
devices.
E
And
now
device
plugin
the
open
yield
just
goes
with
kubernetes,
so
we
can
have
everything
provided
by
kubernetes
as
it
is
yeah.
So
the
hoping
you
have
to
provide
is
like
a
thumbnailing
for
the
control
plane
so
that
you
can,
you
can
I
mean
I
mean
like
HD
4H
devices
from
the
internet
can
connect
that
cluster
yeah
I
mean.
A
E
There's
a
many.
There
are
many
ways
to
address
this
issue.
I
mean
like
you
can
use
the
vbl
if
you
want
to,
but
we
just
take
advantage
of
the
opioid.
B
Okay,
so
you're
saying
the
cubelet
equivalent
on
open
your
Edge,
supports
the
device
plug-in
manager
and
that
I
guess
just
does
some
caching
before
it
sends
to
open
your
cloud.
E
Yeah
and
they
also
provide
like
an
edge
autonomy,
they
cache
data
on
the
edge
so
once
the
edge
device
rebooted,
it's
gonna
be
back
online
as
before.
So
something
like
that.
E
Yeah
I
mean
like
this
is
just
a
POC,
so
we
we
already
have
like
a
problem
so
like
optimization
and
the
memory
for
the
print,
it
does
not
match
our
requirements
and
besides,
if
it
scales
up,
maybe
I,
think
iptable
controlled
by
kubernetes
I.
Think
that's
going
to
be
the
problem,
because
I
think
it's
it
also
scales
up
as
the
number
of
its
devices
increase
so
yeah.
There
would
be
many.
E
You
know
like
issues
that
we
need
to
address,
but
this
is
the
first
step
and
we
just
wanted
to
you
know
like
if,
if
this
works
and
try
to
squeeze
out
that
problems
that
we
need
to
address,
that's
that's
the
main.
You
know
like
a
purpose
to
disappear.
You
see.
A
Okay,
well,
that
was
a
great
presentation.
I'd,
actually
love
to
see
that
in
the
future,
with
updates,
because
it
sounds
like
you're
still
pushing
forward
on
this.
So
I'd
encourage
you
to
submit
this
to
like
kubecon,
North,
America
or
kubernetes
on
edge
Day
event,
or
something
like
that,
because
I
think
it's
pretty
interesting
stuff
foreign.
E
It's
something
as
a
feedback
that
we
would
like
to
know
is
like.
Currently
we
just
you
know,
like
use
open
yard,
but
you
know
like
we
are.
We
are
open
for
discussion,
so
if
you
yeah,
so
if
you
know
like
anything
else
that
hey
this
could
be
work
for
you
or
you
know,
if
you
have
any
ideas
that
we
can
apply
instead
of
opener,
we
are
open
to
try,
try
that
out.
Well,.
A
We've
had
a
few
other
kubernetes
distros
present
in
this
group
like
super
Edge
is
another
one.
I
honestly
am
not
an
authority
on
what
the
distinctions
are,
but
you
know
Q
badge
open
your
super
Edge,
all
our
kind
of
variations
on
kubernetes
optimized
for
Edge
use
cases
and
then,
of
course,
there's
the
other
kind
of
low
resource
kubernetes
like
k3s
k0s,
microcates,.
A
I
don't
know
it.
It
unfortunately
has
been
rare
that
anybody
goes
in
there
and
gives
feedback
on
comparisons
because,
like
I
say
the
the
people
Behind
These
present
on
their
own
project.
But
historically
we
we
haven't
seen
that
many
people
giving
feedback
on
having
invested
the
time
to
compare
more
than
one
of
these
Alternatives
against
each
other,
so
yeah
that
it's
really
interesting
and
if
you
just
look
at
that
spreadsheet
I
think
you'll
find
that
there
are
a
number
of
kubernetes
distros
in
there.
If
you
wanted
to
go
there.
B
That
might
be
kind
of
interesting
as
a
project,
a
dichotomous
key
of
sorts
about
which
K,
which
kubernetes
distribution
for
the
edge
is
best
for
you.
That
could
have
kind
of
questions
along
the
lines
of
do
you
want
your
control,
plane
and
Edge
plane
separated,
and
if
so,
then
you
move
into
the
cube
Edge
super
Edge
open
your
sphere.
If
you
want
a
single
node
cluster
on
the
edge
you
move
into
the
k3s
k0
micro
Kate
sphere
and
like
kind
of
help,
people
navigate
it
that
way.
I
don't
know
that
would
be
interesting.
B
A
The
other
thing
is,
of
course,
the
moment
you
go
that
direction
to
really
have
good
coverage.
There
are
going
to
be
some
who
are
going
to
jump
in
there
saying
that
perhaps
the
solution
is
no
kubernetes
at
all.
You
know,
or
a
different
use
of
kubernetes
as
the
control
plane,
so
that
your
Edge
nodes,
rather
than
being
viewed
as
kubernetes
cluster
nodes,
are
controlled
in
some
other
way
with
crds,
where
their
perhaps
even
still
running
a
container
runtime,
but
not
managed
as
a
classic
kubernetes
worker
known.
C
A
What
I,
I
loved
out
of
this
was
adding
in
the
effect
of
the
collateral
too,
because
I
have
seen
presentations
measuring
the
metrics
on
CPU
and
memory
usage
of
the
kubernetes
components
themselves.
But
once
you
add
things
in
like
the
service
mesh
and
everybody
is
going
to
I
mean
the
the
whole
point
of
cloud
native.
Is
that
you
don't
just
use
it
for
the
control
plane
you
use
it.
The
whole
framework
of
collateral
tools
are
actually
where
the
value
lies
and
you're,
of
course,
going
to
use
those
things.
A
But
you
know
the
things
interact,
and
it
is
kind
of
fascinating
to
see
that
when
you
get
to
low
resource
it
isn't
just
kubernetes.
That's
involved,
it's
the
other
things,
and
if
one
of
these
kubernetes
distros
doesn't
play
well
with
a
tool
category
that
you
want
to
use,
that
can
be
a
showstopper.
A
I
think
even
going
to
Legacy
I.T,
when
you
got
to
things
even
pre-dating
things
like
kubernetes,
like
just
supporting
retail
edge
with
classic
VMS
people,
found
challenges
with
some
things
related
to
networks
on
very
low
resource
Hardware.
Just
because
things
like
TLS
termination
can
burn
CPU
cycles
and
if
you're
sitting
there
with
a
device
with
one
or
two
cores,
you
might
not
have
any
gas
left
in
the
tank
to
devote
to
that.
A
E
C
One
thing
for
you
Steve
for
for
the
list,
so
this
is
a
project
called
scorecard
that
actually
can
be
useful
to
to
to
show
the
the
metrics
of
of
the
project.
The
the
topic
that
we
discussed
earlier,
like
how
many
committers
and.
C
The
is
the
you
know,
activity
and,
and
are
there
GitHub
actions
passing
passing
and
things
like
that,
so
I'm
not
sure
I
haven't
played
that
much
with
it,
but
it
can
be
useful
for
putting
things
like
this
to
like
a
have
a
easy
to
to
access
score
for
for
a
project
like.
Is
it
alive
or
or
not,
and
you
know
how
well
maintained
it
is.
A
C
C
A
Rather,
have
something
like
that
too,
because
you
know
if
you
report
a
healthiness
of
projects
that
has
the
potential
to
get
controversial,
where
someone
would
be
offended.
But
if
you
can
say
hey,
it's
just
the
tool:
it's
reproducible,
if
you
don't
like
the
number,
it
came
up
with
change.
What
it's
measuring
on
your
project.
C
B
A
I've
got
a
couple
parting
items,
then
first
one
is
a
book
recommendation,
so
some
of
you
who
come
here
regularly
know
that
Frederick
from
the
eclipse
Foundation
shows
up
fairly
regularly
to
these
group
meetings.
You
know
maybe
once
every
month
or
two
and
he
runs
open
source
for
the
edge
projects
in
Eclipse
Foundation.
He
just
published
a
book
it
by
just
published
it.
A
It
just
became
available
about
a
week
ago
and
I
had
ordered
an
advanced
copy
and
actually
read
it
front
to
back
now,
and
it's
very
good
with
the
one
limitation
is
that
it
says
right
on
the
cover
it's
I
I
I'm
not
going
to
share
the
screen.
I
just
posted
the
link
in
the
chat,
but
the
title
is
building
Enterprise
iot
Solutions
with
Eclipse
iot
Technologies.
So
the
title
calls
out
that
it's
focused
on
the
eclipse
Edge
projects.
A
A
He
might
think
otherwise
and
I'll
give
him
the
form
if
he
shows
up
for
a
meeting,
but
I
think
that
the
coverage
there
is
a
little
light
and
maybe
even
controversial,
but
overall,
the
book
is
just
excellent
for
calling
attention
to
details
that
you
might
be
concerned
with
as
you're
trying
to
apply
these
things
to
Edge
use
cases.
So
you
know,
if
you're
doing
this
as
your
job
I
think
the
knowledge
you
might
gain
out
of
this
book
would
be
well
worth
the
forty
dollars
or
so
it
costs
and
I
recommend
the
book.
A
It
probably
is
one
of
these
things
like
anything
in
Tech
where
a
year
from
now
a
lot
of
it
might
be
stale,
but
since
it
just
came
out
a
week
ago,
it's
actually
really
fresh
too.
So
a
lot
of
these
Tech
books,
you
start
reading
them.
If
you
know
the
field
and
instantly
know
that
some
of
it
is
a
little
stale
or
miss
some
things,
but
at
this
point
in
time
time,
this
book
is
a
really
good
resource.
A
The
other
thing
that
came
to
me
regarding
kubecon
Europe,
if
we're
going
to
have
anybody
at
all
in
this
working
group
present,
there
would
be
this
idea
of.
Perhaps
even
if
we
don't
do
a
maintainer
track
session,
we
could
have
a
social
meet
and
greet
in
the
form
of
a
dinner
or
just
meeting
for
snacks.
After
the
conference
and
pre-covered
we
used
to
do
that
routinely.
We
could
often
get
some
vendor
to
even
sponsor
it.
A
A
You
know
I'm
talking
about
like
local
meetups
in
the
city
of
Amsterdam
I'm,
going
to
drop
a
link
to
an
event
last
year
as
an
example
where
this
is
just
a
a
local
meetup
group,
but
Amsterdam
I
think
was
the
sponsor
of
things
like
if
you're
familiar
with
that
Laura
based
things,
Network
I
think
that
came
out
of
there
and
they
have
a
a
lot
of
people
working
on
agricultural
government
projects.
You
know,
they've
got
the
system
to
reclaim
Land
from
the
sea
going
on
with
pumping
and
dikes
and
I
think
it.
D
A
That
is
kind
of
one
of
the
hot
spots
in
the
world
for
this
kind
of
activity.
So
I
think
if
we
were
to
declare
an
informal
group
beat
and
greet
social
dinner
or
whatever,
that
we
could
probably
get
a
a
pretty
good
attendance
out
of
that
just
from
people
who
live
in
the
area.
A
B
And
it's
1001.
F
Hi,
hello,
everyone,
yes
hi!
My
name
is
Juan
I'm
new
to
this
working
group,
I'm,
a
program
manager
from
Microsoft
I'm
working
for
Azure
Edge
platform,
so
I'm
currently
investigating
the
end-to-end
developer,
Journey
for
building
the
edge
application,
and
our
team
is
currently
developing
a
solution
to
accelerate
the
app
Beauty.
So
our
still
in
the
learning
mode.
Try
to
understand
the
pinpoints
and
blockers.
F
I
learned
this
nice
cncf
Edge
principles
from
the
website
as
very,
very
educational,
so
I
have
some
technology,
technical
questions
but
first
foremost,
I
would
like
to
join
the
select
Channel
and
would
love
to
join
this
work
group
and
start
to
contribute.
So
how
do
I
get
started?
Do
we?
Is
there
any
guidance
or
yes.
A
Let
me
repost
in
the
chat
the
link
to
the
agenda
notes
document
and
that
calls
out
how
you
join
the
group
and
that's
very
simple.
There's
a
Google
group
mailing
list
just
join
the
mailing
list
and
that
Gates
access
to
all
the
documents
and
things
we
have
so
oftentimes
when
people
try
to
edit
documents
they
you
know
they
end
up
submitting
a
permission
request,
but
I
think
if
you
just
have
a
Google
account
are
logged
in
on
that
Google
account
and
have
joined
this
mailing
list
group.
A
A
Technically.
This
group
started
out
historically
as
being
under
the
kubernetes
project,
so
our
slack
was
left
there,
but
we
are
technically
now
a
cncf
group
as
opposed
to
a
kubernetes
group.
So
one
could
argue
that
maybe
our
Slack
should
properly
be
in
the
cncf
slack,
but
it's
not
and
because
we've
got
a
history,
there
I
think
we're.
We've
made
a
conscious
decision
to
stick
with
the
kubernetes
slack,
because
people
know
that's
where
we're
found.
The
other
thing
I'll
reach
out
to
you.
A
If
you
will
with
links,
because
we
we're
trying
to
put
together
this
catalog
and
the
other
thing
I'd
invite
you
to
do,
if
you
want,
is
if
you
want
to
present
on
what
you've
got
with
Azure
iot
Edge
at
the
meeting
in
two
weeks.
We'll
give
you
time
to
do
that.
I
realize
you're
new,
but
maybe
you
you
know
you
kind
of
have
to
learn
what
you're
working
on
anyway
and
it
could
just
be
a
101
presentation
that
does
the
quick
overview.
A
Yeah
aerial
overview
in
15
minutes
or
something
and
I
think
we'd
all
appreciate
that
historically,
we've
had
a
little
bit
of
coverage
of
the
Microsoft
Solutions,
because
Kate
used
to
work
with
Microsoft
and
covered
it
well
when
she
was
there,
but
we
haven't
had
a
recent
update.
So
I
would
personally
appreciate
that
anybody
else
want
to
chime
in
on
whether
they're
interested.
F
Yeah
I
definitely
would
love
to
I
can
take
a
look
at
on
the
Azure
iot
I'm,
not
from
the
iot
org
I'm
from
The
Edge
platform.
So
my.
A
F
Is
on
the
Azure
stack,
HCI
I'm,
not
sure
if
you
guys
heard
about
that's
compared
with
iot,
we
are
kind
of
in
the
heavy
Edge
site,
but
I
can't
try
to
like.
Like
ask
my
ask
around
to
see
if
there
is
anyone
who
is
in
because
we
are
kind
of
like
next
to
the
iot
org,
so
I
can
try
to
find
someone
from
that
domain
yeah.
It
was
that
specific.
A
Topic,
you
can
just
talk
about
the
org
you're
in
too
I
think
there
are
plenty
of
people
if,
if
I
understood
correct,
if
you're
dealing
with
it
from
the
perspective
of
the
devices
I'm
very
much
interested
in
that
just
like
tamoya's
presentation
called
out,
you
know
the
challenges
of
what
you
face
out
on
the
edge
node
itself
are
often
the
more
difficult
ones
to
learn
about
and
and
overcome.
I
think
that
those
are
the
areas
with
more
rapid
movement
than
the
cloud
hosted
portion
of
this
yeah.
F
Definitely
we
we
have
a
bunch
of
our
researchers
and
the
PM
are
doing
the
actual
related
research.
We
we
like
talking
with
a
lot
of
customers
from
our
end.
Maybe
I
can
what
I
can
think
is.
Maybe
we
can
aggregate
those
feedbacks
and
then
present
the
existing
pinpoints
we
learned
from
our
customers
for
our
specific
domain.
Okay,.
A
And
if
you
could
even
get
into
like
Hands-On
of
how
you'd
get
started
to
try
it
out,
because
I
have
touched
on
the
Azure
iot
Edge
stuff,
that
is
I,
don't
know
what
kind
of
a
business
venture
there
is.
But
the
physical
Hardware
coming
out
of
seed
studios
in
China
actually
seems
to
have
a
tie-in
with
the
Azure
iot
Edge
stack
to
the
point
where
they
sell
some
of
those
devices
with
instructions
on
how
to
install
and
use
it
so
that,
if
that's
part
of
your
division,
I'm
definitely
interested
in
hearing
about
it.
F
So
actually
I
can
share
one
on
one
website.
This
is
a
project
we
are
currently
looking
into.
It's
called
jumpstart
I'm,
not
sure
if
you
guys
heard
about
that,
it's
kind
of
it's
building
the
learning
platforms
for
the
Azure
Arc,
actually
Arc,
is
kind
of
is
connecting
the
edge
as
connected
between
the
edge
and
Cloud.
So
but
I
have
a
hard
stop
right.
Now:
okay,.
B
I'm
Anderson,
just
to
jump
in
when
feel
free
to
reach
out
to
me.
I've
previously
been
talking
to
Vincent
Tam
from
the
AKs
Light
Group
about
presenting.
So
it
would
be
a
good
pair
up
of
you
talking
more
about
the
HCI
heavier
Edge
and
him
talking
more
about
the
more
constrained
Edge.
That
could
be
an
interesting
kind
of
full
stack
overview
of
what's
happening
in
that
area
of
azure,
so
yeah
feel
free
to
ping
me
yeah.
F
B
Yeah
I'll
I'll
drop
that
here
also
my
name
is
you
can
find
me
on
LinkedIn
too,
but
I'll
drop
my
email
right
here
in
the
chat
all.
C
There's
Steve
you're,
you're,
muted
in.