►
From YouTube: Kubernetes WG IoT Edge 20230726
Description
July 26, 2023 meeting discussed edge native white paper and moved to open discussion about various open source projects applicable to edge use cases.
A
A
So
maybe
we'll
I'll
invite
you
to
bring
that
up
again,
so
it
makes
it
into
the
recording
and
then
we
also
have
a
second
related
agenda
item
of
promoting
this
Edge
native
white
paper
at
the
kubernetes
kubecon
North
America
event
in
whatever
opportunities
might
be
available
to
do
that.
So
maybe
let's
go
to
the
the
first
thing
I
mentioned
in
the
second
item
in
the
in
the
written
agenda,
which
is
the
comments
related
to
LF
Edge?
Do
you
want
to
restate
that
again?
Just
so,
we.
B
Can
yeah
yeah
so
I
have
the
answer
so
the
rfh
they
actually
have
an
annual
state
of
the
edge
meeting
and
the
publication.
So
even
if
that
overall
is
not
that
active,
but
it
actually
is
ongoing
discussion
about
the
state
of
edge,
so
I
have
a
review.
A
lot
of
information,
there's
still
a
lot
of
good
informations
here,
I
added
one
comment
to
the
white
paper
and
I
believe
there
are
many
more
to
to
be
added.
B
I
started
that
one
comment
I
had
is
about
the
actually
bring
it
out.
I
I'm,
actually
looking
at
it.
I
cannot
from
my
phone
actually
calling
some
I.
Don't
see
that
my
comments,
so
basically
the
the
the
the
the
papers
right
now
says
that
about
the
statelessness
of
the
edge
Computing
right
so
I
believe.
If
you
look
at
the
12
Factor
app,
the
there
are
two
one
is
being
stainless.
B
A
Well,
I
think
I've
got
a
very
similar
comment
in
the
paper
now
about
statelessness
at
Edge.
So
I,
don't
know,
I
have
to
say,
I
agree
with
you.
Let's
see,
I
I,
don't
know
exactly
it's
under
the
data
persistence
section
of
that
document.
If
you
scroll
down
I,
don't
know
the
equivalent
of
a
page
or
so
so
I
don't
know.
There
already
has
been
a
comment
like
that,
and
you
know
there
are
limits
to
how
big
we
want
this
document
to
be.
A
You
know
like
it's
a
white
paper,
not
a
white
book
and
I
think
you
know
as
long
as
there's
an
acknowledgment
to
the
fact
that
things
going
on
at
Edge,
they
vary
there.
There
might
well
be
some
things
at
the
edge
that
could
afford
to
be
fully
stateless
but
yeah.
On
the
other
hand,
I
know
plenty
of
use
cases
where
people
are
looking
for
some
solution
to
store
State
at
the
edge.
A
How
deep
we
want
to
go
to
in
this
white
paper
other
than
calling
attention
to
people's.
You
know,
awareness
that
this
is
a
recognized
issue
and
you
know
I,
don't
know
that
we
want
to
produce
the
laundry
list
of
every
possible
solution
to
to
stateful
on
edge
just
because
once
again
it
can
get
too
big
and
the
other
danger
is
these
kinds
of
things
in
technology
get
stale
and
would
have
to
be
maintained.
A
So
if
you
purport
this
paper
gives
you
the
authoritative,
definitive
list
of
every
possible
solution
for
stateful
on
the
edge
you're
signing
up
to
maintain
this,
and
maybe
you
don't
want
to
go
there
rather
than
have
it
being
a
longer
lived
document.
Andy,
maybe
I'll.
Let
you
make
some
comments,
but
we
have
discussed
this
before.
C
Many
times
from
my
work
with
IBM
research
versus
my
work
in
this
specific
working
group.
When
we
talk
about
edge
here
in
this
working
group,
we
talk
about
very
small
constrained
devices
for
the
most
part
that
are
capable
of
a
limited
subset
of
things
and
and
potentially
could
do
more.
If
they
were
given
more
GP,
maybe
they
were
given
gpus,
they
could
potentially
contribute
to
building.
You
know,
models
for
AI,
processing
and
so
forth,
right,
but
they're
intended
to
play
a
part
as
a
larger
part
of
orchestration.
B
Yeah
I
I,
I,
I,
I
I
actually
agree
on
that
and
the
the.
If
you
look
at
my
comment,
yeah
I
actually
saw
you
Steve
yeah,
it's
on
the
same
line.
Basically
your
comment,
the
the
I
mean.
Definitely
if
we
shouldn't
be
like
any
technology
specific,
for
example,
the
example
I
give
in
the
comment
is
AWS
the
snow
family
right.
So
some
of
the
the.
D
B
Edge
that
have
actually
a
local
kind
of
aws3
service
which
serve
it
as
I.
Believe
is
it's
like
a
local
back-end
service.
So
you
know
whatever
you
do.
The
the
active
data
will
be
saved
on
there
right
so
they're.
Actually
in
the
rfh
community,
there
are
other
research
really
going
into
the
details
on
each
category.
B
B
You
know
either
captured
or
even
if
you
lost
the
problem
is
not
big
big
deal
right,
but
there
are
also
system
data
which
is
critical
for
you
know
if
the
system,
if
the
server
goes
down,
if
you
need
to
rebuild
the
whole
environment,
that
system
data
need
to
be
sort
of
protected.
So
this
way
when
you
need
to
restart
the
whole,
whether
it's
if
his
equipment
is
cluster,
the
system
data
need
to
be
durable,
otherwise
that
site
will
go
down
right.
So
there's
things
like
that.
That
need
to
be
further
discussed.
A
A
A
C
A
A
The
one
thing
Andy
I,
think
that
still
is
maybe
a
little
strong
I'll
just
directly
read
it
under
data
protection
or
data
persistence.
Any
data
that
needs
to
persist
must
be
stored.
I
think
the
word
must
is
just
a
little
too
strong.
You
know
tone
it
down
to
at
least
imply
that
this
isn't
like
delivered
in
stone
tablets
from
The
Mountaintop,
and
then
there
might
be
some
exceptions
and
because.
C
A
Yeah
or
even
recommended-
or
you
know,
there's
a
preference
that
it
is
something
along
that.
Well,
you
can
figure
it
out.
We
don't
need
to
come
up
with
the
exact
thing
in
this,
but
you
know
just
as
a
point
pretty
obviously
there's
there
there's
tiny
little
amounts
that
pretty
clearly
are
persistent
like
if
you've
got
Edge
Hardware
that
has
a
TPM
holding
securely
your
keys
and
things
like
that.
Those
things
are
clearly
persisted
on
the
device,
so
it
isn't
a
situation
where
there
is
zero.
A
C
A
This
one
in
particular,
like
I,
say
you've
now
got
a
few
comments
and,
frankly,
my
recollection
is
that
when
the
12
factors
came
out
over
10
years
ago
now
the
data
persistent
thing
has
been
a
Perpetual
hot
spot.
That
I've
been
to
numerous
comp
conferences
where
that
was
outright
ridiculed
by
speakers.
In
fact,
one
of
them
was
me,
and
you
know
it's
probably
a
clue
that
the
Ironclad
rule
of
no
persistence
maybe
is
controversial
depending
on
your
app.
A
A
More
okay,
great
yeah
and,
at
some
point
I
think
we've
been
now
discussing
this.
What
is
it
six
months
is
maybe
longer
it's
probably
time
to
move
on
and
get
it
published
and
leading
in
segwaying
into
point
two
on
the
agenda
of
building
awareness
that
this
document
even
exists
at
the
kubecon
North
America
conference,
which
is
in
early
November,
so
I
don't
want
to
cut
it
short.
A
If
anybody
wants
to
jump
on
the
mic
on
mute
yourself,
oh
Des,
Moines
I,
see
I
thought
it
was
a
hand
raise,
but
it's
a
thumbs
up,
but
any
last
minute
comments
on
the
document
itself.
Otherwise,
we'll
move
on
to
plans
to
publicize
this
and
declare
it
either
a
version
one
publication
or
a
version
or
publication
of
the
draft
to
the
General
Public.
A
A
A
But
we
have
the
opportunity
to
do
it
and
missed
it,
and
we
still
have
another
venue,
which
is
the
kubernetes
on
the
edge
day.
So
this
is
a
pre-event
to
kubecon.
It
is
these
day.
One
events
are
extra
costs.
So
when
you
register,
you
have
to
pay
a
little
extra
to
be
able
to
to
attend
those
day,
one
events
and
they
tend
to
be
half
day
or
full
day
Focus
sessions
on
particular
use
cases
or
niches,
and
there
is
one
devoted
to
kubernetes
On
The
Edge.
A
This
has
ever
since
they
started
doing
these
kubernetes
on
the
edge
days.
They've
been
one
of
the
more
popular
pre-events,
getting
hundreds
of
people
attending
and
the
cfps
for
that
are
still
open
for
at
least
another
week.
I
think
it's,
the
the
due
date
on
cfps
is
in
August.
On
top
of
that,
there's
a
group
of
people
who
volunteer
to
be
on
the
organizing
committee
for
those
pre-events
and
there
is
typically
an
opening
keynote,
as
well
as
a
closing
address.
A
Those
two
are
pretty
short,
but
historically,
we've
used
those
to
particularly
the
close
is
something
where
we
deliver
some
slides
on
activity
within
this
group
to
build
awareness
so
that
people
are
aware
that
this
group
even
exists.
We
tell
them
where
the
meetings
are,
how
they
can
get
involved
and
I
believe
we
have
called
out
that
previous
white
paper
we
did
so
we
could
do
it
again,
and
this
would
be
short.
A
A
It's
it
it.
It
isn't
fully
resolved
yet
because
we
technically
don't
have
the
schedule
posted,
but
I,
think
I'd
have
a
pretty
high
level
of
confidence
that
we
can
get
this
in
there.
This
opportunity
would
probably
be
to
just
put
up
two
or
three
slides
kind
of
a
lightning
talk,
kind
of
thing
that
wouldn't
read
somebody
the
paper
but
would
tell
them
that
it
exists,
maybe
quickly
show
it
on
the
screen
and
give
them
the
link
on
where
to
go.
A
To
get
this
and
find
more,
probably,
wouldn't
be
a
bad
idea
to
put
up
a
QR
code,
and
then
that
would
end
up
being
preserved
in
the
recording
for
the
event
and
historically,
the
recordings
of
this
event
get
a
lot
of
eyebrows.
Eyeballs
I
even
looked
at
the
last
one
in
Europe,
but
a
month
after
the
event,
I
think
some
of
those
sessions
had
400
views,
so
they
they
get
a
pretty
high
level
of
visibility.
Probably
the
recording
gets
more
total
eyeballs
than
number
of
people
actually
in
the
room
at
the
event.
C
Yeah
I
mean
based
upon
your
comments.
I
wanted
to
be
clear
and
upfront
transparent
about
the
I
hadn't
been
pushing
for
any
other
track
and
the
reason
why
these
sound
like
reasonable
ways
or
Outlets
to
work
with
is
because
I'm
already,
you
know
like
most
of
us
here
that
are
attending
we're
already
at
the
two
talk
limit
right,
yeah,
so
I
could
go
ahead
and
say:
hey
yeah,
let's
put
a
session
together,
but
I'd
have
nobody
to
to.
A
C
A
D
A
You
know,
depending
on
the
content,
I'd
be
happy
to
propose
that
and
then
maybe,
if
we
get
that
slide
deck
Andy,
you
can
do
a
trial
run
presentation
of
it
to
this
group
for
comments.
So
you
know
we
can
perfect
the
delivery
of
it,
but
that
seems
like
that
should
be
a
pretty
good
way
to
do
it
and
then
once
we
have
that
slide
deck,
whoever
ends
up
taking
on
the
closing
for
that
kubernetes
on
the
edge
day.
It's
a
simple
matter
to
just
tell
them:
hey
here's
some
slides
mentioning
the
deck.
A
C
A
Like
okay
sounds
good
and
I'd
be
happy.
If
you
want
a
reviewer,
just
send
me
a
link
to
it
when
it's
there,
we
can
just
put
probably
a
good
way
to
do
it
to
allow
future
speakers
to
easily
get
to.
It
is
just
throw
it
up
there
as
a
shared
Google,
slide
deck
or
something,
and
that
can
be
cut
and
pasted
into
somebody
who's
ever.
Whoever
might
want
to
use
it
in
the
future.
A
A
E
That's
a
good
question,
but
depending
on
what
target
audience
you
have,
but
but
certainly
I
think
having
one
or
several
webinars
or
maybe
a
panel
you
know,
custom
made
for
the
occasion
would
be
would
be
a
good
idea.
Maybe
you
bring
a
few
of
the
authors.
Moderated
by
you,
Steve
or
even
I
would
be
happy
to
to
play
that
role.
Maybe
if
you
want
an
outsider
so
to
speak,
and
you
know
that
that's
a
great
way
to
get
additional
mileage
of
of
of
good
content.
E
Well
one
one
one
strategy
that
we
we
use
for
some
content
was
to
have
as
well.
You
know
slice
and
dice
it
in
different
ways
and
and
publish
it
somewhere
else,
or
you
know
you
you,
you
pick
interesting
chunks
of
it
and
then
maybe
on
a
in
a
different
medium.
Let's
say
in
our
case
we
literally
created
a
publication
on
medium.com
and
we
had
short
short
form
versions
of
some
of
the
main
points
and
always
pointing
back
to
the
full
document,
and
things
like
that.
So
those
are
two.
E
A
D
A
E
A
A
I
think
we're
out
of
the
agenda
items
so
we'll
morph
this
meeting
into
a
general
purpose.
Birds
of
a
feather
free
form,
confirm
conversation,
so
I'd
invite
anybody
who's
come
across.
Anything
interesting
lately
like
in
the
past
couple
weeks
or
months,
it
just
unmute
yourself
and
tell
us
about
it.
It
could
be
blog
post,
a
YouTube
video
whatever
or
advances
on
some
project,
you're
working
on.
B
Yeah
I
think
the
edgy
is
very
interesting.
Actually
today,
I
actually
watched
interesting
video
about
satellite
security,
which
is
you
know.
Satellite
is
also
iot
problem
right,
so
yeah
I
think
it'll
be
I.
I
it'll
be
interesting
to
actually.
Let
me
post
that
share
that
video
in
the
chat,
so
it
will
be
I
I.
Think.
Definitely
that
I
mean
the
white
paper.
We
can
probably
say
that's
just
the
first
version.
It'll
be
I
mean
a
lot
of
interesting
stuff.
B
B
For
example,
they
have
a
very
detailed
Matrix
on
what's
available
to
build
a
edge
Network
right
network
computing,
a
lot
of
good
information
there
and
they
publish
a
state
of
edge
every
year,
but
they
I
believe
they
also
have
trouble
trying
to
promote
that.
You
know
what's
available,
for
users
to
for
companies
to
your
organization
to
adopt
Edge,
Computing
infrastructure,
but
I
think
what
this
group
can
actually
help
them
by.
Actually,
you
know
the
use
cases,
that's
really
suitable
for
what
they
have
I.
Think
that's
a
an
interesting
cross-section
of
the
two
communities.
D
B
It's
a
matrix,
so
basically
they
actually
tells
each
component
of
that
infrastructure
what
they
are,
what
like
freedom
for
the
cncs
right
in
cncf,
there's
so
many
different
projects.
So
it's
really
hard,
sometimes
to
say
what
what
project
does
and
how
to
choose
the
product
they
actually
sort
of
not
resolving
the
problem,
but
make
it
easier
for
people
to
choose
by
defining
the
details
of
each
each
project
subproject.
What
does
it
do?
So,
let
me
see
if
I
can
find
out
Post
in
the
chat.
A
A
Let
you
correct
me,
but
I
think
that
the
Project's
living
over
there
you've
kind
of
cornered
the
market
on
Industrial
iot
and
some
of
the
Technologies
related
to
that,
where
I
think
you've
really
got
it
strong
presence
there.
But
given
that
you've
got
50
plus
projects,
there's
a
lot
more
as
well
and
from
the
perspective
of
a
new
open
source
project.
E
Yeah
I
can
I
can
certainly
cover
that
and
we
in
fact
just
shuffled
things
a
bit
around,
but
essentially
we've
got
something
called
the
eclipse,
Edge
native,
which
is
the
Gathering
of
people
that
really
care
about
Edge.
But
we
given
our
past
in
development
tools
and
developer-centric
things.
Of
course,
we
don't
have
so
many
projects
that
are
pure
infrastructure
or
things
like
that.
E
In
fact,
the
edge
platforms
that
we've
got
are
really
designed
to
work
with
kubernetes
or
to
orchestrators
in
the
field
right,
but
they
can
work
in
Standalone
mode
as
well,
and
that's
where
our
interest
is
to
have
something
that
really
embodies
our
philosophy
for
what
we
call
Air
jobs
in
the
sense
that
well
doing
pure
devops
at
the
edge
is
a
bit
it's
a
bit
dangerous.
So
we
try
to
have
platforms
that
are
really
built
from
the
ground
up
for
the
edge.
E
The
way
the
eclipse
Foundation
typically
operates
is
through
a
general
concept
called
industry
collaborations,
and
we
have
various
things
that
you
know
belong
to
that
category,
such
as
walking
groups,
interest
groups
and
anyway,
I
I,
want
to
enter
you
to
the
details
of
that.
But
in
the
end,
what
happens
is
that
we
have
a
complete
decoupling
between
our
open
source
projects
and
our
working
groups
and
other
types
of
collaborations.
E
So
an
open
source
project
is
always
Project,
Lead
committers
and
they
set
the
technical
direction
that
they
want
as
a
community
right
and
the
staff
is
there
to
support
them.
So
we
are
member
founded,
like
the
Linux
Foundation
is
so
the
money
serves
to
deliver,
build
servers
and
things
like
that
and,
of
course,
support
staff
like
myself
to
accompany
the
projects
and,
on
the
other
end
we
have
our
collaborations.
E
So
in
the
in
the
case,
let's
say
of
the
eclipse
iot
working
group,
then
this
is
a
gathering
of
organizations
that
care
about
iot
and
Edge
in
the
context
of
having
a
set
of
building
blocks,
to
build
commercial
platforms
on
the
top
right.
So
most
of
our
members
do
that
and
the
way
this
works
is
that
projects
decide
which
working
groups
they
interact
with
and
working
groups
decide
which
projects
they
care
about
right.
So
we
have
a
number
of
projects
that
are
relevant
to
both
automotive
and
pure
Edge
infrastructure.
E
So
they
deal
with
those
two
distinct
walking
groups,
for
example,
so
that's
slightly
different
than
from
the
LF,
Edge
or
or
Linux
Foundation
model,
generally
speaking,
where
the
consortia
and
the
project
are
in
most
cases
the
same.
In
our
case,
we
dig
up
all
those
two
concepts
which
means
that
you
can
start
an
eclipse
Project
without
being
an
eclipse
member,
for
example,
and
we
have
a
number
of
companies
that
do
just
that,
and
that's
perfectly
fine
with
us.
So
that's
that's
the
short
version
of
the
way
that
we
operate.
Let's
say.
A
So
why
my
summary
is
I
think
there
are
three
major
organizations
but
not
to
leave
anybody
out.
I
think
there's
another
one
for
webassembly
tamoya
is
here
and
there's
another
independent
organization,
I
believe
for
the
Robot
Operating
System,
and
there
are
others
out
there
as
well.
So
I
threw
a
link
in
the
chat.
That
is
my
attempt
to
summarize
a
list
of
Open
Source
projects
related
to
Edge
and
column.
Two
there
is
the
sponsoring
organization.
A
D
That's
what
I
was
going
to
mention
is
is
that
outside
the
LFC
and
CF
sphere,
there
are,
you
know
dozens,
if
not
hundreds
of
special,
highly
specialized
work
groups.
You
look
at
like
the
IEEE
power
group
is
where
you
know
you
find
the
big
Builders
of
Power
Equipment.
You
know
GE
Siemens,
you
know
those
those
kind
of
things
so
I
think
that
the
definition
that
you
find,
the
challenge
that
we
continue
to
butt
heads
against
in
this
meeting
is
defining
the
edge.
D
It's
a
little
bit
more
I
think
it's.
You
almost
have
to
define
the
edge
in
the
context
and
operating
demand
that
you're
in
because
you
know
in
the
satellite
one
of
the
ones
mentioned
earlier.
You
know
you'd
be
dealing
with
the
ddil
concept.
You
know,
which
is
the
dod
acronym
in
them,
for
what
a
disconnected
disrupted
intermittently
latent
infrastructure.
You
know,
you
know
things
like
that,
and
then
power
and
systems
like
that
you
end
up
focusing
on
things
that
I
think
are
a
little
bit
more
stateless
in
their
definition.
D
There
is
some
intersection
of
of
needs
if
we
were
to
build
a
table.
One
of
the
features
would
probably
be
State
or
stateful.
Another
one
would
probably
you
know
would
be.
You
know
you
could
like
almost
like
a
capability
Matrix
for
some
of
these
various
you
know
themes
of
devices,
whether
it's
power,
you
know,
satellite,
you
know
all
the
way
across
I
think
to
kind
of
dial
in
and
help
that
discussion
happen
a
little
bit
more
productivity,
because
I've
always
found
that,
and
we
continue
to
hit
this.
D
D
G
So
hi
everybody
so
just
to
add
on
that
satellite
thing,
I
I
found
an
there
was
a
open
source
and
open,
embedded,
I.
Think
Summit
in
Europe
in
Prague,
somewhere
and
I
saw
a
talk.
I
can't
find
the
talk
at
the
moment,
but
I'll
put
it
in
the
the
channel
afterwards,
where
actually
as
Liam
was
mentioning.
There
are
there's
a
talk
where
how
you
can
actually
update
software
on
satellites,
where
you
don't
have
the
possibility
of
sending
any
nothing
more
than
three
MBS
or
something
over
the
Uplink
channel.
G
So
there
are
solutions
out
there,
where
you
just
send
the
patch
up
to
the
satellite
and
the
satellite
that
compiles
your
code
so
yeah.
It
absolutely
is
the
fact
that
it
depends
on
application
to
application,
depending
on
who's
spending,
so
much
money,
who's
who's
willing
to
spend
decent
chunks
of
money
on
Uplands,
down
links
on
persistence,
on
edge
persistence.
G
So
it's
completely
oriented
and
I.
Think
to
point
to
to
make
a
note
of
certain
things
is
the
fact
that
edge
Computing
has
unfortunately
become
much
more
of
a
marketing
term,
so
it
has
broadened
its
Horizons,
but
on
the
other
hand
it's
also
become
a
bit
more
key
these
days
as
to
what
Edge
is
but
and
I
see,
Mark
is
actually
in
the
meeting.
I
was
I,
think
I,
remember
a
couple
of
months.
G
Back
Mark
was
mentioning
about
how
it's
worth
mentioning,
at
least
in
the
community,
about
what
are
the
potential
alternatives
to
day
rice.
Dare
I
clean
this
Blast
for
me
in
kubernetes
meetings,
but
are
there
any
alternative
solutions
to
to
orchestrators
out
there
that
are
probably
I,
don't
know,
maybe
might
be
well
suited
for
very
constraint?
Edge
Solutions,
I
think
Andy
might
probably
have
some
good
opinions
about
it,
and
I
was
actually
very
curious
about
this
because
we
do
have
flavors
of
kubernetes.
But
what
about
other
tools
out
there?
G
F
Yeah
yeah
I
still
wonder
about
that,
as
well,
so
to
Steve's,
Point,
so
I
I
work
with
Susa
and
we
have
a
solution
to
deploy
an
entire
operating
system
kubernetes
and
whatever
your
app
is
on
top
of
all
that
to
the
edge,
and
you
know,
sort
of
this
full
stack
solution,
and
yet
there
are
parts
of
kubernetes
that
work.
There
are
parts
that
don't
and,
and
what
I
keep
coming
back
to
personally,
is
that
we
we
have
this
term
now
Edge,
but
Edge
has
been
around
forever.
F
Some
just
need
to
be
data
producers
and
just
need
to
pump
data
back
to
some
other
things.
Some
need
to
actually
do
compute
and
actually
communicate
with
other
things
in
that
local
space,
some
of
them.
You
know
a
satellite
needs
to
have
different
functionality
than
a
assembly
line
in
a
factory
and,
and
so
I
think
that
part
of
the
challenge
here
is
that
we
kind
of
with
Edge
we've
exploded
back
to
what
used
to
be
a
problem,
which
is
that
an
edge
also
has
devices
that
we
seem
to
have
really.
F
You
know
we
don't
have
that
problem
in
the
data
center
of
of
all
these
sensors
and
actuators.
So
so
one
of
the
things
that
I
I
keep
coming
back
to
and
I
know
that
kubernetes
solves
some
of
these
problems,
but
when
compute
is
when
the
edges
just
goes
back
to
being
those
core
components,
how
would
we
solve
the
problem?
F
And-
and
you
know,
can
we
come
up
with
creative
ways
to
to
getting
everything
below
your
application?
Is
overhead
and
I
think
that's
why,
with
embedded
systems,
we
just
build
the
application
into
the
system,
and,
and
it's
sort
of
been
this
set
it
and
forget
it
mentality
that
no
longer
works.
We
need
visibility
into
these
things.
We
need
them
to
communicate
with
other
things
we
need.
We
need
a
higher
level
of
control
and
security
over
these
systems.
How
do
we
get
there?
A
A
D
D
Oh
sure,
I'll
go
Andy.
Thank
you.
Look,
I,
I
think
the
I
think
that
there
are
all
of
these
reasons
why
execution's
distributed
and
some
of
the
fundamental
jobs
to
be
done
transcend.
You
know,
kubernetes
and
the
edge
things
that
transcend
both.
Are
you
know
the
need
and
desire
to
keep
things
up
to
date
and
secure
the
need
and
desire
to
update
things?
Those
are
long
life,
but
there's
all
of
this
focused
on
the
edge,
because
that
is
the
next
epic
of
computing.
D
The
last
decade
was
this
great
lifting
shift
into
the
cloud.
The
next
decade
is
bringing
things
out
to
the
edge,
and
it's
not
just
that's
where
the
data
is.
You
know
it's
latency
and
determinism,
it
is
regulatory
reasons,
limited
deliberate
autonomy
or
privacy
and
security.
All
of
those
reasons
are
fundamental
reasons
that
we
might
easily
come
up
with
examples.
D
You
know:
I
want
I'm
a
skin
chair,
checking
ml
app
wants
to
keep
a
naked
picture
of
Liam
on
my
phone
for
everyone's
sake
right
and
defining
the
those
those
those
various
Edge
devices
I
think
what's
Universal
about
them
is
would
be
themes
like
Mobility
remoteness,
and
that
would
be
all
we
could
agree
on.
Everything
else
would
be
a
capability,
including
State.
You
know,
including
persistence,
network
connectivity
offline.
D
D
D
Both
of
them
have
blogged
consistently
about
how
they're
adopting
webassembly
to
solve
some
of
these
problems
and
I'm
going
to
put
myself
on
the
schedule
for
next
week
and
I'll.
Do
you
know
whatever
the
group
wants?
I'll
do.
I
I
did
a
talk
for
the
open
source
Summit,
where
I
was
like
wasm,
WTF
and
I'll
do
a
wasm
WTF
and
talk
through
like
the
big
Ideas
the
standards
and
then,
where
is
webassembly
today,
we
actually
just
yesterday
published
the
web
assembly
roadmap
on
the
bike
code.
D
Alliance
blog-
if
you
didn't
see
that
yet
we
produced
a
nice
15
minute,
YouTube
video
to
walk
through
all
of
the
things,
but
I
think
what
we're
going
to
do
in
a
follow-up
post
is.
What
does
this
mean
and
I?
Think
that's
the
post
that
this
group
would
find
most
interesting
and
that's
what
I
can
touch
base
on
next
week?
So
I'll
I'm,
assuming
I,
can
double
check
my
schedule,
but
I
think
it
works.
D
I'll
I,
don't
know
how
much
time
I
can
spend
plan
to
spend
I
could
I
could
spend
as
much
or
as
little
as
people
want.
We
just
want
to
dial
that
in
15
minutes
or
30
and
I'll
I'll
get
everybody
up
to
speed
on
what
are
the
standards
who's
using
it?
What
are
the
great
reference
use?
Cases
like
Amazon,
Prime,
video
and
and
then
maybe
point
point
you
at
like
a
whole
bunch
of
projects,
including
the
one
ones
I
I
work
on
in
the
cncf
that
are
cool
and
about
wasn't.
A
It
sounds
great
to
me
Liam
and
maybe
take
the
full
30
minutes.
You
suggested
and
then
leave
some
time
for
free
form,
discussion,
q,
a
afterward
and
maybe
we'll
have
some
other
group
members
come
and
Tack
on
a
little
bit
too.
I
know
Kate
isn't
here
today,
but
she
works
in
the
web
assembly
space
and
maybe
we'll
invite
a
few
other
people
to
follow
you
up
or
something.
C
D
I'll
link
the
the
road
map
which
I
we
put
a
ton
of
work
into
this
into
this
roadmap.
It
includes
all
the
key
sort
of
projects
across
the
webassembly
ecosystem.
You
know
where
we
need
help.
What's
going
well,
what's
being
committed
now
what
happens
before
kubecon
what
happens
before
the
end
of
the
year
and
then
and
then
I'll
I'll
do
some
of
the
interpretation
of
the
you
know
of
like
the
hey?
What
does
resource
handles,
or
you
know,
file
system
virtualization?
What
does
that
mean?
D
D
I
can
point
you
at
some
of
their
stuff,
what's
important
to
them,
like
their
big
they've,
been
driving
a
lot
of
the
adoption
of
threads
into
webassembly
and
then,
and
you
really
try
to
tell
a
comprehensive
story
to
get
everybody
into,
because
this
is
incredible
stuff
and
it's
moving
at
a
lightning
Pace.
What
we
could
really
need
is
just
more
smart
people
that
bring
their
diverse
ideas
and
perspectives
to
the
table.
C
Your
hand
yeah
so
I
think
you
know
one
of
the
things
I
wanted
to
mention
is,
and
it's
an
area
that's
been
of
interest
for
me
for
a
while
and
as
an
extension
of
the
open
source
project
I
represent
Coupe
stellar
is
that
you
know
we
had
this
idea
right
out
of
the
gate
of
supporting
Hub
and
spoke
the
ability
to
to
deploy
things
across
multiple
clusters,
whereas
kubernetes,
you
can't
do
that
by
default
right.
You
do
that
in
a
single
cluster
with
multiple
nodes,
but
you
don't
do
it
across
clusters.
C
So
that's
what
Kube
style
is
good
at,
but
going
ahead
is
it's
all
I
think
where
we
get
to
is
this
idea
of
not
depending
on
a
single
hierarchy
that
starts
at
the
top
and
ends
at
the
bottom,
and
so
I
don't
see
Edge
so
much
as
as
a
leaf
node.
So
I
want
to
be
careful
here.
The
paper
that
I've
been
helping
and
contributing
to
on
edge
native
application.
Behaviors
is
not
the
hat
that
I'm
wearing
right
now
I'm
wearing
the
next.
C
This
is
the
next
phase
of
things
so
I
think
in
the
next
phase.
Is
this
this
connection
between
what
is
considered
traditionally
as
a
hub
and
what
is
a
spoke
and
that
an
edge
could
be
very
well
responsible
for
directing
workload
and
probably
should
be,
and
there
there
should
be
this
interactivity
up
and
down
this
stack
rather
than
there
being
you
know,
a
single
one-way
specification
being
sent
down
to
be
done.
C
C
C
Couldn't
do
this
because
of
a
specific
constraint
or
I'd
like
to
see
someone
else
do
this
because
I'm
not
in
a
state
where
I
can
execute
so
there
I
think
there
has
to
be
a
little
bit
more.
You
know
in
the
past,
where
we
looked
at
from
kubernetes
is
a
very
Simplex.
You
know
top-down
approach
to
how
jobs
and
execution
of
workloads
absent,
happens,
I
think
in
the
few
future.
We
need
that
to
be
duplex
where
things
happen
up
and
down
the
stack
and
then
that's
where
I'm
headed
with
coopstellar.
G
Right
to
Apple
I
think
what
Liam
was
mentioning.
I'm
I
mean
I'm,
looking
forward
to
the
wasm
thing,
because
from
my
personal
experience,
at
least
with
customers
who
have
some
insanely,
High
remote,
so
so
to
say,
infrastructure
I
think
it
becomes
quite
a
big
bit
of
an
hassle
when
it
comes
to
working
with
CNC
f2s
at
times,
because
the
air
gap
situations
sometimes
become
extremely
difficult
to
bridge.
Even
though
you
have,
let's
say
you,
you
might
have
a
very
Mighty
so-called
powerful
Edge
somewhere.
G
That
does
talk
to
your
constrained
Leaf
nodes,
but
it
sometimes
some
some
point
of
time.
It
starts
to
break
because
it
just
becomes
burdensome
so
I'm
looking
forward
to
having,
if,
if
actually
wasn't,
actually
can
solve
certain
problems.
What
I'm
actually
curious
about
in
this
case
would
be
if
and
wasn't
itself
requires
a
new
form
of
orchestration
logic.
I
mean,
of
course
it
can
piggyback
on
on
the
container
runtimes
at
the
moment,
with
some
shims
and
everything.
But,
oh
my.
G
Would
it
be
interesting
to
to
have
something
that
goes
from
bottom
up
and
bottom
up
in
the
sense
that
we
can
come
up
with
some
ideas?
What
Mark's
mentioning-
and
that
goes
up
to
the
cncf
logic-
maybe
Frederick-
might
also
have
some
projects
out
there
in
Eclipse,
who
might
be
looking
to
add
to
these
kind
of
things,
because
what
I
understand
from
at
this
point
of
time
is
having
kubernetes
full-blown
kubernetes
and
then
trying
to
strip
it
off
to
make
it
lean
is
not
always
the
right
solution.
G
I'm,
not
saying
it's,
it's
not
raw,
it's
not
wrong
per
se.
It's
generally,
it
works
wonders
for
most
scenarios,
but
in
very
specific
cases
it
starts
to
hit
a
wall
which
is
very
difficult
to
overcome
in
certain
cases.
So
I
think
it
might
be
an
interesting
opinion
to
look
into
these
kind
of
things
where
we
can
start
to
go
from
this
bottom
to
top
approach
from
this
community,
all
the
way
to
cncf
on
the
long
term.
C
A
Yeah
there
are
those
you
mentioned
whether
wasm
needs
a
new
orchestrator,
but
there's
plenty
of
people,
who've
voiced
an
opinion
that
even
traditional
Docker
containers
need
a
new
orchestrator
depending
on
the
use
case.
You
know
when
you
look
at
it.
Kubernetes
was
intended
for
this
cloud
cloud,
hosted
data
center
that
had
multiple
servers
that
were
homogeneous
and
you'd
maximize
efficiency
by
loading,
your
workloads
onto
them
to
shoot
for
100
utilization
as
a
goal.
A
You
look
at
Edge
and
there
are
certain
Edge
nodes
that
have
enough
resource
for
exactly
one
cluster
node
and
if
you
look
at
Cooper's,
so-called
single
load,
kubernetes
cluster
I
think
it's
a
legitimate
question
to
ask
yourself
whether
it
makes
sense
if
it
is
a
kubernetes
node
of
one
cluster
granted
the
attractiveness
is
that
the
tool
chain
is
phenomenal.
People
are
used
to
it.
There's
a
lot
of
collateral
tools
and
people
are
comfortable
with
its
reliability
and
the
health
of
the
open
source
project.
A
Single
node
does:
are
you
paying
for
something
that
is
not
delivering
any
value?
Is
the
scheduler
delivering
any
value,
legitimate
questions
and
maybe
they're
there
are
and
probably
should
be
alternate
Solutions
and
in
a
big
system
approach
that
you're,
a
worldwide
organization
or
Enterprise
I
think
there's
plenty
of
room
to
be
using
both
Solutions.
You
know
kubernetes
at
the
top
level,
for
your
Cloud
hosted
I
think
you're
almost
to
a
point
where
you're
crazy
not
to
be
using
it
there.
A
E
D
I
got
pulled
down
a
couple
times
there,
but
I
just
want
to
say
that
cncf
wasn't
Cloud,
does
webassembly
orchestration,
that's
compatible
with
kubernetes,
but
not
dependent
upon
it.
So
I'll
demonstrate
that
next
week,
when
you
think
about
an
edge
native
orchestrator,
I'll,
also
demonstrate
the
ml
with
a
routable.
You
know
routing
ml
to
different
nodes
depending
on
privacy,
availability,
security
and
those
other
things.
That's
something
that
we've
done
at
arm.
Summit
and
BMW's
done
I'm
using
wasenbob
before
so.
D
D
I
know
that
there's
a
huge
bodies
of
work
and
prioritization
happening
in
that
in
that
area.
The
biggest
one
of
the
challenges
there
is
is
that
most
of
the
those
the
real-time
arm
coolers,
you
know,
there's
three
different
families.
Some
are
only
32-bit
arm,
so
it
has
sort
of
a
A
reduced
availability
for
some
of
the
tools
tool
chains.
But
I
can
speak
to
that
next
week.
A
little
bit
when
I
chat
about
web
assembly
and
I'll
put
myself
on
the
schedule,
then.
A
D
D
D
Seattle
yeah
yeah,
we
just
announced
those
those
are
linked
off
that
blog
post
right
off
the
bike.
Put
Alliance
page
that
I
put
in
the
in
the
right
earlier.
That's
got
the
webassembly
road
maps
I'll.
Do
that
link
those
here,
I'll
link
those
now
here's
the
webassembly
roadmap
with
the
video
walkthrough,
but
that's
more
like
like
project
updates
across
you
know
the
I
mean
there's
hundreds
of
companies
that
work
on
webassembly,
so
that
was
quite
a
challenge
to
pull
that
together,
but
I'm
very
happy
with
how
it
came
out.