►
From YouTube: Kubernetes WG IoT Edge 20201007
Description
October 7, 2020 meeting of the Kubernetes IoT Edge Working Group - unstructured Kubernetes at Edge User perspective "bords of a feather" discussion.
A
Okay,
it
looks
like
we're
recording
welcome
to
the
october
7
meeting
of
the
kubernetes
iot
edge
working
group.
We
have
a
fairly
light
agenda
today,
but
any
members,
or
even
prospective
members
who
popped
in
for
the
few
first
time
are
welcome
to
add
discussion
topics
right
now.
If
you
like,
let
me
po,
let
me
paste
the
agenda
doc
in
the
chat.
A
A
We
did
have
one
item
that
I
put
on
the
agenda,
which
is
the
upcoming
presentation
at
kubecon
north
america,
that,
like
every
conference,
these
days
is
going
to
be
online
and
this
group
got
a
session
accepted
where
we're
going
to
present
on
event
driven
at
edge.
This
is
likely
to
be
a
dionne
and
I
haven't
prepared
the
deck
yet,
but
we're
planning
on
this
being
a
deeper
dive
into
what
we
got
started
with
in
the
talk
I
did
with
cindy
at
kubecon
europe
on
we
called
it
serverless
at
edge
in
that
one.
A
You
do
eventually
have
to
get
up
to
a
tier
where
you're
running
kubernetes,
or
at
least
you
do
for
the
purposes
of
this
group,
which
is
kubernetes
oriented,
but
you
can
actually
land
code
capable
of
generating
events
on
quite
small
systems.
You
things
all
the
way
down
to
arduinos
and
embedded
systems
that
aren't
even
running
linux
or
perhaps
aren't
even
running
in
os
at
all
and
anyway.
That
was
the
plan
for
this
upcoming
talk
in
november,
and
I
put
a
link
to
the
the
item
in
the
agenda
doc.
A
I
think
that
if
you
want
to
join
that
conference,
they
had
a
discount
period.
I
don't
know
if
it's
still
open
or
not,
but
it
was
like
25
or
50
bucks.
They
also
have
scholarships
available
and,
in
my
experience,
I've
had
friends
who
applied
for
those.
So
if
anybody
is
financially
constrained
student
out
of
work
whatever,
I
think
you
know
it's
just
a
matter
of
writing
a
plausible
reason
into
your
application
and
you
can
probably
get
a
free
conference
registration.
A
B
I
I'm
yeah,
I
haven't
been
dropped
during
our
meeting
for
a
few
times.
I'm
just
curious,
like
from
in
the
past,
like
we've
done
several
reference
architecture
designs
and
then
I'm
just
throwing
the
feather
to
here.
Like
a
people
here
like,
are
you
experiencing
any
like
new
customer
scenarios
or
demands
any
progress
made?
Like
maybe
kilton
has
a
company.
You
guys
have
done
some
improvement
in
the
security
area.
A
Yeah,
the
other
thing
cindy
is:
we
do
have
some
people
who,
like
you,
who
haven't
been
here
regularly.
I
wonder
if
we
should
go
around
with
introductions,
since
we
don't
have
much
of
a.
We
don't
have
a
lot
on
the
agenda
anyway.
Some
of
these
people
might
not
know
who
the
other
faces
are,
if
they're
seeing
him
for
the
first
time.
B
C
Okay,
sure,
so,
even
though
this
is
on
public
record,
I'm
not
I'm
not
scared
of
having
everyone
see
that
I'm
in
my
car
it's
my
mobile
meeting
place
for
the
day,
so
at
a
customer
site,
so
my
name
is
kilton
hopkins
and
I've
been
working
with
edge
computing
tech
since
about
like
2012,
but
2014
is
when
it
really
started
to
kick
off
with
the
iot
boom.
I'm
a
technologist-
and
I
also
am
co-founder
of
a
company
called
edgeworks.
C
The
thing
that
we've
done
that
you
all
might
be
familiar
with
is
create
this
eclipse,
io
fog,
which
is
with
kubernetes
or
not
kubernetes,
and
edge,
orchestration
and
runtime
framework
that
you
can
use
for
doing
microservices
at
the
edge.
So
like
everyone
on
this
call,
there's
a
lot
of
different
things.
I
know
about
and
a
whole
lot
more,
I
don't
know
about
so
we
all
show
up
here
to
to
give
our
two
cents
and
try
to
figure
stuff
out.
C
E
Barry
kaplan
I
work
for
a
company
called
vamana.
We
built
vertical
applications
in
the
connected
industrial
space.
We
started
out
with
mainly
manufacturing
plants,
but
now
we
kind
of
generalize
into
more
bits,
we're
much
more
in
the
the
user
space
of
the
edge
technology
than
the
developing
of
the
edge
technology.
So
we
focus
on
delivering
applications
and
a
lot
of
it
is
trying
to
figure
out
how
to
utilize
some
of
the
various
technologies
and
modernize
standardized.
All
that
kind
of
fun
stuff.
A
E
A
Yeah
well,
I'd
I'd
be
really
curious
to
hear
more,
maybe
not
in
this
intro
phase
and
if
you're
not
prepared
today,
maybe
at
a
future
meeting
the
rules
with
the
cncf.
Is
it
shouldn't
turn
into
a
commercial
presentation,
but
if
you'd
be
willing
to
just
go
into
kind
of
things
that
are
going
well
things
that
are
going
badly
and
you
know
we'd
love
to
hear
the
that
kind
of
conversation.
A
A
F
Good
morning,
thank
you.
Sorry,
I'm
sitting
here
trying
to
juggle
two
different
conference
calls
one
on
my
mobile
one
here.
So
I
apologize
the
life
of
coven
too
many
meetings,
virtually
jerry,
schumann,
I'm
the
founder
of
a
company
called
flight.
I
participated
a
little
bit
on
the
on
this
call
a
few
times
and
I
have
been
on
the
slack
channel,
often
we're
basically
a
company
that
is
specifically
iot.
Centric,
really
developed
a
platform.
F
In
fact,
in
my
previous
incarnation
of
another
company,
I
started,
which
is
still
in
operation
on
the
east
coast
and
is
a
kind
of
a
major
player
now
in
the
industrial
utility
iot
marketplace
and
the
platform
that
we
ultimately
developed.
There
is
something
I
wanted
to
take
to
a
more
horizontal
play
to
a
larger
iot
audience,
so
that's
kind
of
what
we've
done
with
flight
and
just
getting
started
right
now,
building
a
large
whirlwind
network.
F
Here
in
my
hometown
of
long
beach,
we
just
put
a
large
a
large
set
of
antennas
on
top
to
start
doing
some
data
telemetry
coming
from
sensors
and
we're
starting
to
work
with
this
local
city.
Here
in
the
port
of
long
beach,
port
of
la
and
start
looking
at
how
we
can
get
more
data
telemetry
flowing
in
and
how
we
can
take
advantage
of.
You
know
the
current
climate
that
we're
in
we're
all
kubernetes
based
been
kubernetes
based
since
the
beginning
I've
built.
F
I
don't
even
want
to
count
the
number
of
clusters
I've
built
at
this
point,
but
and
everything
from
across
you
know
large
large
cloud
implementations
to
you
know
very,
very
much
bleeding
edge
arm
kind
of
edge
implementations
before
it
was
ready
before
we
had
opportunities
in
kubernetes
on
the
edge
but
anyways
a
little
bit
of.
In
my
background,
so
thank
you.
A
F
A
F
You're
starting
to
see
a
lot
of
stories
about
that-
and
you
know
the
whole,
you
know
creative
ways
to
use
the
digitalization
and
you
know
to
supplement
and
or
you
know,
augment
kind
of
what
we
were
normally
doing.
You
know
as
meat
puppets,
so
it's
gonna.
It's
gonna,
be
interesting
to
see
kind
of
how
it
flushes
out
here.
F
In
the
end,
that's
right:
I'm
currently
active
very
active
at
with
the
economic
development
department
here
in
long
beach
discussing
you
know,
what's
the
benefits
of
having
a
you
know:
a
long
distance,
low
power,
telemetry
network
available
to
citizenry
small
business
large
business.
You
know
we
have
a
very
big
burgeoning
aerospace
here
now
in
long
beach
we
have-
I
don't
know
most
of
the
major
rocket
companies
now
are
set
up
shop
here.
F
We've
got
rocket,
lab
and
relativity
space
and
we've
got
spin,
launch
and
virgin
orbit
and
there's
I
understand,
there's
like
three
or
more
three
or
four
more
rocket
companies
that
are
moving
in.
So
it's
going
to
be
in
long.
Beach
is
kind
of
in
a
renaissance
right
now
for
from
an
aerospace
perspective,
and
I'm
hoping
we
can
bring
kind
of
you
know
edge,
iot
and
specifically
kubernetes
into
that
play.
I
know
they're
looking
so.
A
A
I
choose
that
word
because
a
lot
of
this
is
because
of
concerns
of
earthquake
taking
out
infrastructure
where
a
lot
of
these
leading-edge
things
have
gone
in
here.
I
think
this
old,
the
old
ricochet
network,
is
still
in
operation
at
one
time.
It
was
done
by
a
startup,
and
I
think
it's
southern
california
edison
was
a
user
originally,
but
is
still
maintaining
that
it's
either.
F
Yeah
yeah,
I
think,
there's
still
a
couple
ricochet
nodes
up,
not
many
but
I've
industry-wise
been.
You
know
many
decades
at
this
point.
35
years
so
been
been
around
the
track
100
times
so,
but
yeah,
no,
the
low
power
networks.
I
mean
I'm
really
I'm
pushing
hard.
I
mean
laura
is
phenomenally
amazing.
If
you
not
play
with
laura.
I
highly
highly
suggest
you
guys
do
because
it's
it's
it's
pretty
much
a
game.
F
Changer
I
mean
it's
a
no-cost
network,
pretty
much
public
oriented
the
things
network
out
of
I
think
they're
in
norway
actually
or
netherlands,
it's
kind
of
where
it
kind
of
originated,
but
it's
pan.
It
goes
across
the
world
now
and
and
I'm
currently
working
with
the
helium
network,
which
is
a
blockchain
implementation
and
another
company
called
layer
zero.
But
you
know
it's
pretty
easy
to
stand.
One
up,
it's
pretty
easy
to
buy
sensors
to
to
integrate
with
them.
You
know
you
get
them
for
50
bucks,
40,
bucks,
100
bucks.
F
F
Well,
you,
you
know,
one
of
the
latest
announcements
is,
I
don't
know
if
you
guys
heard
about
this,
and
this
is
another
interesting
play
for
wgi.
You
know
for
the
edge
implementations.
Amazon
just
announced
that
they're
they're
going
to
be
pushing
out
their
what
they're
calling
their
sidewalk
network
amazon
sidewalk.
It's
basically
a
low
power
network.
F
It's
going
to
be
embedded
in
every
one
of
your
alexa
devices,
anything
amazon's
out
there,
shipping
rings
the
cam
blink
cameras
all
of
them,
and
the
idea
is
that
they
are
going
to
be
building
based
on
their
current
product,
offering
a
low
power
data
network
for
telemetry
information.
So
it's
it
it.
This
is
a
burgeoning
thing.
F
Yeah
we
could
get
into
the
whole
privacy
thing,
but
I
I'll
stay
away
from
that
particular
subject
matter.
I
know
what
amazon's
motivation
is:
it's
not
my
motivation,
but
you
know,
but
the
idea
that
we
could
have
a
fairly
transparent
ubiquitous
you
know.
Network
specifically
built
for
data
coming
from
sensors
could
be,
could
be
really
phenomenally
groundbreaking,
especially
when
you
start
talking
about
machine
and
machine
economies.
F
Right
I
mean
that's
where
this
really
starts
to
play
in
the
and
then
you
start
looking
at
what
iota
is
doing
and
has
been
doing
with
what
they
call
the
tangle,
which
is
a
non-blocked
chain
implementation.
It's
a
giant,
it's
a
merkle
tree,
but
it's
about
merkle
tree
implementation
of,
and
they
have
a
token
called
the
the
iota
token.
C
F
F
Yeah
I
mean
my
take
on
it,
and
this
is
only
my
take
on
it.
You
know
if
you've
done
any
any
large
number
of
embedded
systems
that
are
ultimately
deployed
to
a
large
region,
and
you
want
to
manage
them
in
some
singular
fashion.
It's
an
impossibility.
I
mean
it's
hard
enough
just
to
do
upgrades.
You
know
that's
the
whole
ota
phenomena
that
you
end
up
having
to
deal
with,
but
I
mean
the
idea
of
having
a
you
know.
An
orchestration
fabric
that
incorporates
your
edge
devices,
I
think
is,
is
groundbreaking.
F
I
mean
the
idea
from
an
I.t
perspective.
You
know
not
from
necessarily
a
product
vendor
like
myself,
but
if,
if
I'm
an
individual
in
enterprise
and
I'm
sitting
there-
and
I
I'm
have
I'm
already
having
a
hard
time
managing
and
managing
my
150
individual
people
using
you-
know
personal
computers
on
my
network,
the
idea
that
my
company
starts
spawning
off
a
whole
slew
of
iot
edge
devices.
You
know
how
do
I
manage
that?
How
do
I
watch
cyber
security?
How
do
I
change
you
know
if
my
workflow
changes
internally?
F
How
do
I
change
that
implementation
at
the
edge?
If
that,
if
the
edge
is
actually
you
know,
impacted
in
that,
so
that
that's
where
I
think
kubernetes
you
know
has
a
has
a
phenomenal
role
to
play
and
and
it's
what
I've
been
experimenting
with
for.
Oh
my
god,
it's
been
probably
two
years
now,
since
I
stood
up
a
small
odroid
cluster
on
arm
that
I
tried
to
basically
build.
F
You
know:
kate's
cluster
on
and
failed
miserably,
but
regardless
it
was
the
attempt
to
start
demonstrating
that
and
the
re,
the
other
reality
for
a
product
vendor
that
is
kubernetes
oriented.
You
know
you
don't
always
get
great
internet
wi-fi
in
the
basement
of
a
hotel
when
you're
doing
demos.
So
the
idea
that
I
could
have
a
tag
along
kubernetes
cluster
with
me,
based
on
small
single
board
computers,
which
I've
now
done
a
few
times
over.
F
You
see
lots
of
raspberry
berry
clusters,
raspberry
pi
clusters
out
there,
but
the
idea
of
transporting
that
feature
functionality
directly
with
you
on
an
airplane
into
a
hotel.
You
know
basement
somewhere
without
internet
connectivity
and
be
able
to
demo
your
product
and
demo
demo.
The
the
power,
including
the
you
know,
I
bring
drag
along
a
low
power
network
with
me,
laura's
already
on
board.
So
I
can
have
my
sensor
sitting
there
doing
everything
I
want,
and
you
know
it's
all
self-contained
I
mean
and
that's
a
pretty
powerful
reason,
for
you
know
a
smaller
single
board.
E
Are
tens
of
thousands
of
devices,
but
trying
and
those
are
across,
maybe
400
plants
across
the
world
is
just
a
pain
in
the
ass
I
mean
it's,
it's
an
untenable
problem
and
basically
we
have
version
drift
all
over
the
place
and
configuration
it's
a
it's
a
major
obstacle
to
be
able
to
continue
to
grow
and,
as
now
we're
getting
into
not
just
factories,
but
all
kinds
of
industrial
devices.
That's
going
to
go
up
in
order
of
magnitude.
F
B
Yeah,
like
you
know,
you
look
at
the
industry
like
saying
aws,
green
grass.
They
use
london
or
even
they
use
docker
compose
for
for
container
and
then
for
azure.
We
have
kubernetes
solution,
but
majority
is
doctor
as
well.
So
I
wonder,
like
you
know,
like
from
a
commercial
perspective
like
cloud
providers,
we
are
all
doing
our
own
approach
from
the
community.
What
I've
heard
is
people
are
waiting
to
use
kubernetes,
so
helping
help
me
understand
what
exactly
the
the
needs
are
in
the
in
the
community
or
in
the
industry.
From
from
your
perspective,.
E
Yeah,
well
I
mean
that's
one
thing
too:
it's
just
the
fact
that
they're
different
too,
because
we
run
on
on
google,
amazon
and
azure
for
different
parts
of
our
deployments
and
different
customers.
So
we
really
can't
say
we're
going
to
only
be
you
know,
azure
or
only
be
you
know,
microsoft.
If
I
I
I
I've
always
predicted-
and
maybe
you
could
tell
me
this
is
true,
but
I
always
predicted
at
some
point.
A
If
you
get
some
kind
of
density
like
that,
you
could
actually
for
free
get
customers
to
pay
the
power
bill
and
the
physical
hosting
of
enough
to
make
a
mesh
network
that
you
can
get
for
free.
There
aren't
many
people
in
a
commercial
position
to
do
that,
but
amazon
and
google
are
maybe
even
some
of
the
car
makers
are,
if
they
get
a
lot
of.
You
know,
locality,
density
and
spread
to
where
these
things
can
just
communicate
to
with
each
other,
to
build
a
mesh
network
but
they're
keeping
all
of
that
proprietary.
A
You
know
it's
pretty
much
a
given
that
you
can
look
at
the
resource,
because
people
have
done
tear
downs
on
those
alexa
speakers
and
there's
no
way
they're
running
kubernetes
nodes
in
there.
Nor
could
they
because
they
just
don't
have
enough
cpu
and
memory,
but
there's
a
whole
niche
in
commerce,
where
you
don't
really
want
to
give
control
of
that
critical
aspect
of
your
business
over
to
an
amazon
or
google.
I
suspect
I
mean
you
could
make
the
decision
to
do
it,
but
hey
if
there's
an
open
source
infrastructure,
that's
portable
in
community.
C
And
it's
just
going
to
take
time-
and
I
agree
with
you
steve
that
at
some
point,
ex-version
of
you
know
microservices
at
the
edge
or
whatever
it
is,
or
x
version
of
iot
or
edge
gateway
device
management
is
going
to
drop
their
slightly
different
proprietary
version
to
go
with
the
standard
that
solidified,
but
it
won't
happen
until
the
standard
solidifies.
So
it's
for
right
now
it
makes
sense
to
kind
of
do
your
own
thing,
embed
it
in
your
products,
so
that
you've
got
kind
of
your
own
walled
garden.
C
That's
lightly
extendable,
but
eventually,
all
of
that
always
comes
down
to
whatever
open
standard
took
over
and
so
yeah.
I
just
I
know
I
know,
there's
so
much
to
juggle
there
that
there's
nobody
is
going
to
be
able
to
solve
it
in
a
single
framework
or
a
single
project
or
anything,
but
eventually,
maybe
three
more
years
from
now,
we'll
have
some
standards
that
we
all
agree.
You
know
kind
of
encompass
everyone's
needs
very
well.
Yeah.
E
It
probably
won't
be
all
kubernetes
because
again
when
we
get
to
the
constrained
devices,
you're
not
running
kubernetes
right,
you
know
on
the
small,
so
there's
so
the
kubernetes
is
one
part
of
it
for
for
when
we're
we're
literally
running
gateways
that
have
horsepower
we're
doing
edge,
processing
we're
doing
our
entire
filtering
down
sampling.
You
know
stuff,
that's
low
latency,
but
we
still
have
to
talk
to
literally
the
the
smaller
edge
of
light,
and
I'm
mainly
only
from
the
industrial
context
that
we
don't
really
do
commercial
stuff
other
than
connected
vehicles.
E
C
Yeah
and
if
you
can,
if
you
can
orchestrate
using
kubernetes
at
the
top
level,
you
know
this
kind
of
master
master
grade
orchestration
and
have
have
it
have
minions
that
do
the
work
down
the
lower
levels.
You
can
orchestrate
things
like
firmware
rollouts
as
long
as
you
have
the
ability
to
extend
on
the
top
level
and
something
to
carry
it
out,
execute
it
on
the
lowest,
but
it
has
to
merge
together.
Really
seamlessly
is
the
current
problem
we
have
is.
C
A
F
I
agree
that
standards
adoption
is
a
big
giant
primary
driver
here.
I
think
the
other
part
of
it
is,
you
know,
widespread
open
source
support
of
those
standards
in
in
implementations
that
are
lower
cost
than
what
these
you
know.
Outlier
implementations
from
vendors
are,
are
going
for
and,
and
that
is
what's
turned
the
tide
in
a
lot
of.
What's
happened
in
enterprise
software,
I
mean
I,
I
will
say
we're
still
very
early.
F
I
my
my
personal
view
of
this
world
only
having
to
try
to
educate
the
north
american
utility
market
over
the
course
of
the
past
six
years
about
you
know
the
value
of
containers
when
they
were
very
much
still
locked
into
running
windows.
Nt
servers
and
you
know,
monolithic
programming,
but
you
know
once
they
started
to
see
the
power
of
you
know:
ci,
cd
and
containers
and
kubernetes
and
and
all
the
capabilities
that
we're
bringing
to
bear
and
the
performance
of
nature.
Of
that
you
know
of
microservices,
you
know
all
of
a
sudden.
F
You
know
we've
got
a
groundswell
of
people
that
now,
finally,
after
especially
now
being
demonstrable
where
we've
got
a
widespread
acceptance
of
yeah.
This
is
the
one
this
is
the
future
or
even
though
the
future
was
started
10
years
ago.
This
is
the
future
now
for
them,
but
I
think
that
enterprise
computing
has
still
got
a
long
ways
to
go
when
it
comes
to
containers
there
is,
they
are
embracing,
at
least
from
a
docker
level,
obviously,
because
of
the
big
vm
push
initially.
E
It
works
great
for
the
startup,
that's
doing
it,
it
works
perfectly
again.
We
just
took
over
a
connected
vehicle
platform
and
they
were
running
on
google
and
we
took
their
costs
down.
I
think
it
was
a
300
300
times
reduction
and
cost
for
the
same
capabilities
just
by
switching
off
to
managed
services.
C
Yeah
that
makes
sense
that
that
would
be
a
sticking
point,
because
holding
the
data
means
that
there's
there's
something
that
keeps
you
on
that
platform
right
and
so
that's
a
key
way
to
monetize
jerry.
What
you
mentioned
earlier
about
laura
and
the
revolution
of
laura,
and
that
is
essentially
the
story
that
you
were
telling
about
enterprise.
You
know
in
the
shift
to
containers
happening
in
lopez,
radio
when
semtech
open
sourced
lorawan
it
it
moved
from,
will
sig
fox
take
over
the
world
to
nope
no
way
everyone.
C
F
I
I
do
too,
and
I
I
I
kind
of
left
the
iot
environment.
I
was
doing
some
consulting
you
know,
seven
years
ago,
early
early
on
and
then
basically
left
to
start
this
other
company
ping,
things
that
ended
up
doing
the
utility
marketplace,
stuff
and
kind
of
lost
sight
of
what
was
happening
inside
of
iot.
When
I
got
when
I
came
back
after
I
left
ping
things
and
started
flight,
I
started
looking
around
at
what
you
know
what
had
changed
in
the
environment
and
all
the
rest
of
it.
F
The
amazon
engineers
figured
out
basically
how
to
run
laura
on
top
of,
or
at
least
the
basically
writing
a
protocol
for
ble
so
that
they
can
extend
the
range
of
bluetooth
low
energy.
So
I
don't
know
much
more
than
that
about
it.
I've
just
been
watching
watching
the
press
releases
and
I
haven't
really
done
a
technical,
deep
dive,
but
it's
an
interesting
take
on
you
know.
That's
I
mean
that
it's
got
that
much
momentum
that
you
know
it's
got
people
like
amazon
going
interesting.
Maybe
we
should
extend
our
own
tentacles
using
the
technology.
G
Hi
this
is
in
I
mainly
working
on
the
coop
edge
project
right
now,
so
yeah,
that's
about
it
and
I
log
in
at
the
crino
first
then
I
realized
that
and
log
out
and
I'm
logging
again
with
my
real
name,
yeah,
the
crino
community.
They
have
the
the
common
account.
You
need
to
use
that
login.
I
was
logging
last
night
well.
A
G
G
He
mentioned
the
pre-doctoral
or
pre-kubernetes
world
everybody
use,
embedded
or
lower
language
and
there's
not
much
change
on
the
edge
or
iot
side,
not
like
the
cloud
evolution.
So
we
were
thinking
with
that
kubernetes.
So
we
really
can't
use
a
cloud
to
control
or
manage
the
edge
and
iot
device
so
that
really
lowered
a
bar
for
the
application
development
and
all
the
ota
and
the
life
cycle
management.
H
D
E
Just
just
as
a
plug
since
you're
from
microsoft,
cindy
I
mean
one
of
our
biggest
pain
points
has
been
at
least
in
manufacturing
plants.
They
run
windows
for
a
lot
of
things
and
some
of
the
machines
we
connect
to
literally
require
windows,
95
beta,
2
dlls.
You
know
for
machines,
but
not
being
able
to
to
effectively
run.
You
know,
docker
on
all
windows
and
stuff
has
been
just
just
killed
us
right.
E
So
that's
still
a
problem
for
us,
because
if
because
even
if
we
can
run
kubernetes-
but
you
know,
there's
no
kubernetes
yet
that
runs
on
windows,
not
even
k3s,
so
we're
still
in
a
really
bad
position
to
be
able
to
deal
with
that
particular
space.
So
we're
really
waiting
for
for
microsoft
to
catch
up
to
the
rest
of
the
world
and.
A
Yeah
barry
it
isn't
it
isn't
just
microsoft
by
the
way
I
didn't
introduce
myself
if
you
haven't
met
me
before,
but
I'm
with
vmware
and
windows
containers
actually
have
been
released
very
recently,
but
there
is
such
a
thing.
Some
might
allege
it's
kind
of
bleeding
edge
at
this
point,
but
I
think
we'll
get
there.
Yeah.
E
Well,
yeah.
I
know
we
track
it
pretty
closely,
but
it
is
you
know.
Now
I
have
to
convince
300
plants
to
upgrade
their.
You
know
their
windows
servers
which
they
don't
want
to
do.
Yeah
you
talk
about
industrial,
you
know,
like
you
know:
local
utilities
manufacturing
a
clipboard
is
state
of
the
art
for
a
lot
of
these.
A
A
F
A
Know,
unlike
cloud
the
tr
one
of
the
big
issues
with
edge
is
that
the
device
life
cycles
are
easily
10
years.
I
mean
you
know
you
go
into
a
data
center
people
routinely
replace
servers
every
five
years,
but
you
go
into
the
physical
industrial
control
world.
Frankly,
they
they'd
replace
them
every
50
or
100
years.
If
they
could
get
away
with
it
and.
A
A
B
I
think,
meanwhile,
like
some
of
the
windows
applications
they're
really
hard
to
containerize,
so
potentially
in
the
community.
We
should
in
think
about
how
kubelet
can
support
a
native
process
that
way
like
even
you
can
run
the
kubernetes
cluster
with
the
hybrid
workload,
including
both
container
and
the
host
process.
E
We
still
need
to
do
that,
so
then
we
end
up
having
to
tell
customers.
Okay,
you
got
to
have
this
linux
box
over
here
and
then
we'll
have
another
windows
box
over
here
to
run
the
things
that
require
windows,
workloads
and
for
small
plants
that
doesn't
work
right
because
they
only
have
100
machines
or
50
machines.
C
Yeah,
what
you
really
need
is
you
need
you
need
an
extension
and
like
an
agent
that
runs,
you
know
all
the
way
back
win32,
you
know
the
most
basic
wins,
32
calls.
So
it's
when
95
98.,
you
know,
god
help
you
emmy.
What?
If
you
still
come
across
that
and
can
allow
you
to
do
trigger
native
processes,
but
also
give
you
you
know,
maybe
the
ability
to
run
a
certain
type
of
of
new
process.
E
E
Have
our
own
little
agent,
similar
to
what
what
the
I
fought
gauge
is,
but
it
literally
just
it
interfaces
with
the
native
windows
process
manager
to
install
things,
make
sure
they're
running
and,
and
you
know
it
could
do
it
over
the
wire.
So
it's
it's
kind
of
our
home
grown
cockpit.
If
you
will,
but
talking
to
you,
know,
windows,
processes
and
again
that
that's
still
going
to
be
part
of
it
because
they're
probably
for
us,
I
don't
think
some
of
these
native
apps
are
never
going
to
go
away
in
my
lifetime.
E
E
Go
down
there
and
give
the
commands
and
control
it's
an
mqtt
app
that
basically
listens
to
our
cloud.
That
tells
it
what
to
do
it's
in
a
way
that
that
we
would
open
it's
not
documented
well
enough
or
anything
for
anyone
else.
Usually
we
could
do
that,
but
you
know
there's
so
many
other
things
going
on
right
now.
It
seems
like
there's,
probably
other
homes.
That
would
be
better
for
us
to
contribute
to
than
trying
to
do
that.
A
Yeah
well
a
lot
of
these
things.
The
way
these
open
source
projects
have
started
up
is
a
couple
of
companies
or
organizations
have
a
common
need
and
they
do
decide
to
get
together.
So
maybe
you
know
it.
It
seems
like
that
should
be
generically
useful
across
a
lot
of
boundaries
to
the
point
where
an
open
source
project
like
that
should
exist,
whether
it
does
or
not.
E
F
How
are
people
I
want
one
thing,
I'm
running
into
now
with
a
client
in
chicago
and
I've
and
haven't
really
started
dipping
my
toes
into
what's
out
there
from
a
project
standpoint,
one
of
the
areas
that
I
think
is
a
potential
barrier
for
a
wider
spread.
Adoption
is
the
fact
that
most
of
these
enterprise,
large
corporate
enterprises
are
obviously
very
security
concerned,
and
but
they
tend
to
do
all
their
work
through
a
single
pane
of
glass
in
the
form
of
a.d
right
I
mean
that's
where
they're
basically
dealing
with
all
their
authentication.
F
So
you
know
how
do
we
extend
that
that
single
pane
of
glass
look
into
our
back
controls
inside
of
kubernetes,
or
you
know
any
other
kind
of
im
kind
of
capabilities
inside
of
kubernetes?
I
think
that's
going
to
be
the
next
major
barrier
that
we're
going
to
have
to
jump
over
as
a
community,
because
yeah.
A
F
But
not
modeling
it
on
a
d
but
yeah
allowing
for
that.
You
know
I
don't
want
to
change
somebody's
tool
belt.
The
worst
thing.
The
last
thing
we'll
do
in
a
major
enterprise,
the
most
cost
most
expensive
thing
you
could
ever
do
going
out
selling
to
corporate
enterprise
customers
is
trying
to
get
them
to
change
their
tools.
E
A
But
I
think
another
avenue
that's
related
to
kubernetes.
That
is
big,
I
mean
kubernetes,
could
do
orchestration
of
containerized
apps
and
given
the
control
plane
is
extendable,
you
can
do
device
management,
but
I
think
the
third
leg
of
that
is
network
management,
where,
if
you
could
take
concepts
like
service
mesh
down
to
the
edge
to
where
you
could
have
it
manage
the
network
for
you,
rather
than
that,
be
home
grown
tools
and
a
lot
of
things
you
roll
yourself
do
the
edge
leaf
node
to
leaf
node
command
and
control.
A
C
I
would
think
that
a
a
layer
to
interpret
between
whatever
is
your
center
of
identity,
whether
it's
active
directory
or
shoot.
What's
the
used
to
be
was
that
ldap,
yeah
or
just
just
plain
old
ldap
or
whatever
it
was
that
was
a
sun
micro
systems
thing
that
that's
now
open.
I
can't
remember
the
oh,
my
god,
what's
called
open
id,
I
think.
F
C
Oauth
yeah,
all
of
that,
so
whatever
your
provider
is,
it
sounds
like
we
need
a
layer
to
basically
translate
between
that
and
all
of
the
permissions
that
you'd
like
to
expose
going
down
and
that
way
you
keep
the
same
tooling
up
top
whatever
you're
using.
But
you
can
write,
maybe
an
adapter.
If
it's
something
you
know
rare
and
if
it's
your
regular
single
sign-on
stuff,
it
just
merges
right
in
with
your
identity,
it's
an
active
directory,
but
then
it
interprets
downstream
in
a
way
that
you
know
that
you
can.
C
You
can
then
issue
commands
out
with
the
security
check
and
stuff
such
as
I
have
the
permission
to
see
the
status
of
an
edge
gateway,
that's
running
in
a
factory.
I
do
not
have
the
permission
to
reboot
it,
and
so,
even
though
I
might
want
to
try
to
issue
the
command
via
the
api
directly,
because
I'm
trying
to
do
something
nefarious,
my
identity
just
simply
doesn't
allow
it,
so
the
command
gets
received
but
blocked,
and
that
stuff
that
you
don't
have
present
in
any
of
the
existing
permissions
and
role
systems.
C
C
B
B
A
Yeah,
I
think
one
of
the
issues
is
that
a.d
is
really
originated
to
authenticate
people,
but
there's
a
lot
of
stuff
going
on
where
a
lot
of
this
is
actually
intended
to
replace
people
and
it
doesn't
or
didn't
wasn't
architected
to
authenticate
devices.
Should
this
app
be
authorized
to
run
on
this
device?
Did
my
device
get
physically
stolen
and
hijacked
or
cloned?
You
know
those
kinds
of
issues
go
far
beyond
ad,
but
you
need
to
look
at
the
big
picture
to
integrate.
All
of
that
now
barry.
A
I
can
hear
him
saying
already
that
these
people
will
never
change
the
old
stuff
and
you're
right
about
that,
but
somehow,
maybe
if
there
was
a
way
to
even
get
into
the
old
stuff,
to
at
least
prove
that
you're
talking
to
who
you
think
you're
talking
to
not
not
as
a
person
but
as
a
process
or
location.
I
think
that's
important.
Well,.
E
E
D
E
E
E
E
D
A
A
Okay,
so
one
thing
that
at
running
groups
like
this
you're
always
out
to
get
content,
that's
interesting
to
the
community,
so
anybody
got
any
suggestions.
I'm
I'm
willing,
if
I
get
enough
leeway
several
weeks,
hopefully
to
go
out
and
try
to
land
a
speaker
if
somebody
wants
something
wants
to
nominate
a
topic,
you're
curious
about
and
we'll
go,
try
to
line
up
speakers
or
if
you'd
rather
have
a
birds
of
a
feather,
just
open
discussion.
We
could
do
it
that
way,
but
we're
always
looking
for
topics.
B
A
Well,
one
of
the
one
of
the
constraints
we
have
is
that
we,
this
group,
is
funded
by
the
cncf
under
kubernetes,
okay,
so
they're,
not
really.
The
topic
should
be
at
least
somewhat
related
to
kubernetes,
rather
than
just
generic
iot
edge.
I
mean
there
might
be
other
groups
that
would
be
free
form,
but
I
think
technically
we're
constrained
to
be
subjects
that
are
related
to
kubernetes,
and
I
don't
know
if
sidewalk
qualifies
in
that
regard.
A
A
B
I
I
think
it
will
be
really
useful
like
because
I'm
just
thinking
from
my
perspective,
I
really
I'm
I'm
not
very
close
to
the
end
customers
so
like
jerry,
berry
or
kyoto,
you
guys,
if
you
can
like
share
the
pain
points
our
customer
needs,
and
then
we
can
all
think
about,
like
as
a
team
or
in
this
group.
What
are
the
things
that
I
can
do
like,
or
people
can
do,
or
we
can
get
speakers
we
talk
here
that
way
like
we
can
all
enable
everybody
right.
A
F
B
B
E
If
we
were
going
to
do
that,
it
might
be
interesting.
The
packaging
aws
other
I've
got
the
name
of
it,
but
the
network
that
lives
just
on
the
other
side
of
the
telco,
so
the
one
hop
or
the
zero
hop
for
look
for.
You
know
basically
they're
moving
their
their
their
cloud
down
closer
to
the
edge.
Oh,
you
don't.
F
Just
I
just
read
about
it
the
other
day
you
know
you
know
about
vapor.
No,
I
don't
vapor
has
basically
gotten
into
bed
with
crown
castle
crown
castle
is
actually
the
number
one
owner
of
cell
towers
throughout
the
united
states,
north
america
and
what
they're
doing
is
they're
basically
putting
containerized
data
centers
at
the
base
of
the
antenna
yeah
so
that
in
their
so
you're
going
to
be
able
to
basically
float.
F
A
C
C
F
B
C
B
A
C
C
C
F
C
F
Yeah,
I
think
ipa
yeah,
and
I
think
it
does
seem
to
have
a
little
bit
of
waning,
not
waning
interest,
but
I
I
do
think
on
the
on
the
development
side
of
it
it
kind
of
tapered
off,
and
I
think
it's
because
the
kid
that
that
was
at
stanford
went
on
to
do
something
else,
but
but
I
I
still
think
ipfs's
underpinnings
are
significant
and
specifically
in
edge
environments
could
potentially
be
some
phenomenally.
Amazing
implementations
done
so
I
was
I'm
glad
you.
Someone
took
a
look
at
it.
Thank
you.
C
I
think
it
would
be
great
to
review
it
here,
because
it
it
really
has
an
impact
on
the
way
that
you
would
carve
up
your
resource
locations
in
a
kubernetes
driven
system,
because
you
might
be
high
off
of
an
iot
gateway
somewhere,
and
that
might
be
a
privileged
resource
that
you
want
to
get
things
connected
to.
Well,
how
on
earth
do
you
address
it
when
you've
got
3
000
such
gateways
out
there.
A
B
F
This
would
be
impossible
without
kubernetes,
on
what
we've
been
able
to
do
in
changing
the
scope
and
size
of
what
a
historian
is
capable
of
doing
so.
I
mean
obviously,
with
the
underwear
background
you
get
the
the
historian
side
of
it,
but
you
know
with
that
amount
of
data
things
like
osi,
softs,
pie
and
wonderware,
and
you
know
schneider's
whatever
they
called
it
aviva
or
whatever.
It
was
bottom
line,
fall
down
and
burn
with
that
kind
of
input.
F
You
know
without
much
data,
but
kubernetes
was
absolutely
critical
for
our
development
and
and
production
and
of
the
of
the
application
that
basically
sustains.
You
know
one
of
the
largest
utility
companies
on
the
eastern
seaboard
called
dominion:
software
dominion,
energy,
my
god,
software
in
the
head,
but
anyways
I
would
go-
take
a
look
at
pink
things
because
I
think
it
it
is
an
internet,
interesting,
kubernetes
play
and
it
definitely
it
kind
of
highlights
kind
of
the
the
benefits
and
and
and
power
of
you
know
what
okay.
F
A
I
think
one
of
them
I've
come
across
before
was
electric
utilities?
Have
these
massive
grids
like
the
ones
that
go
across
country
and
when
they
have
faults,
it's
sometimes
not
easy
to
establish
root
cause
like
a
whole
region
goes
down,
and
if
you
could
get
super
accurate
sampling
and
accurate
time
stamps,
which
has
always
been
an
issue
you
could
you
could
you
want
to
establish?
Well
what
happened
first,
because
that's
going
to
be
the
cause,
you
know
the
thing
that
happened.
Second
couldn't
have
couldn't
have
gone
back
in
time.
F
And
this
is
all
nanosecond
scale,
timestamps
yeah,
so
we
basically
it's
the
federal
government
during
the
recovery
act
and
spent
almost
a
half
a
billion
dollars
in
conjunction
with
another
half
a
billion
from
the
utility
matched
the
utilities
match
to
put
out
these
things
called
phaser
measurement
units
pmus,
and
these
things
now
permeate
most
of
the
high
voltage
transmission
system
in
north
america
and
around
the
world.
It's
starting
to
find
its
way
in
into
distribution
today.
But
it
was
you
know,
scada
can
only
you
know.
F
Skinny
gets
picked
up
once
every
you
know,
15
seconds,
you
know
it's
a
small.
The
sample
is
nothing
so
you
get
a
big
basic.
You
know
stair
step
with
phasor
measurement
units.
You
actually
look
at
the
actual
wave,
almost
the
wave
form
I
mean
it's
a
they're
primarily
track
and
phase
angle,
but
anyways
the
the
it
was
a.
It
is
a
major
shift.
I
you,
if
you
would
have
told
me
10
years
ago,
that
our
our
north
american
power
grid
had
no
situational
wide
area
situational
awareness.
F
I
would
have
thought
you
were
a
loony
come
to
find
out.
We
haven't
had
white,
we
still
don't
have
wide
area
situational
awareness
with
the
national
power
grid,
and
here
we
have
an
interconnected
grid.
So
it's
kind
of
it's
kind
of
hairball,
but
that's
why
they
put
these
things
in
because
the
north
american,
the
north
northeastern
blackout,
that
happened
in
like
2003,
so
this
investment
started
in
2008
and
they're
now
starting
to
permeate
the
grid,
but
the
whole
the
whole
back
end
system
had
to
change
out
because
pie
melted
down.
F
A
F
A
Had
elected
to
use
mesos
just
because
they
gave
them
low
level
control-
and
I
remember
from
the
talk
they
did
have
to
use
some
dedicated
hardware
devices
to
get
the
time
accuracy
they
really
needed
for
that.
So
you
were
kind
of
they.
They
even
had
difficulties
with
interrupt
response
time
and
things
coming
out
of
the
linux
operating
system,
so
in
a
way
mesos
with
supervisory
control
over
some
specialized
hardware.
But.
F
Well,
I
thought
you'd
get
a
kick
out
of
it
and
I
started
to
bore
the
rest
of
the
group
but
yeah,
but
kubernetes
made
a
major.
It
was
a
major
implementation
change
that
really
changed
everything
for
to
this
day.
You
know
they
just
did
a
big
demonstration
of
this
to
the
industry
and
pretty
much
everyone's
jaws
dropped.
So
go
take
a
look
at
paintings,
pink
things,
dot,
ai.