►
From YouTube: KubeEdge Community Meeting 20190710
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
C
C
This
is
just
basically
an
overview
of
what
we
want
to
do,
but
if
the
community
feat
is
that
we
can
have
some
more
things,
please
feel
free
to
add
and-
and
we
intend
to
basically
let
community
members
own
more
and
more
features
and
improve
our
contributor
base.
Our
first
feature
that
we
want
to
support
is
support,
assistant
volumes
position
and
volume
claims
so
that
any
stateful
applications
that
want
to
basically
use
any
local
storage
or
any
any
persistent
volumes
to
store
their
data
at
the
this
would
be
a
very
useful
feature.
C
The
second
feature
is
related
to
network
right
now,
if
you
see
1.2,
we
support
something
called
edge
mesh.
This
feature
allows
you
to
basically
have
four
pod
communication
via
services
headless
services
in
kubernetes,
but
it
currently
is
only
limited
to
communication
within
the
same
subnet
and
intend
to
basically
move
further
and
provide
communication
between
any
apps
running
on
the
cloud.
How
can
they
reach
to
applications
running
on
edge
nodes
and
further?
C
The
third
area
that
we
want
to
support
is
like,
with
respect
to
your
devices,
we
introduced
the
device
management
feature.
It
is
in
the
alpha
stage,
so
we
hope
we
get
some
impact
on
the
community
as
to
what
what
is
missing
with
respect
to
device
management
or
how
we
can
take
this
forward,
and
one
of
the
consumers
of
this
fair
feature
is
the
mappers
or
the
applications
which
which
run
on
edge
nodes,
connect
to
devices
via
different
protocols
and
and
then
basically
get
the
data
into
the
platform
and
and
do
further
processing.
C
So
we
implemented
Bluetooth
and
Modbus
protocol
mappers
in
1.0
going
ahead.
We
see
that
OPC
UA
protocol
is
also
very
popular,
so
we
intend
to
provide
support
for
OPC
UA,
so
if
the
community
feels
that
they
also
can
contribute
to
OPC
UA
or
any
other
protocols
that
they
are
immediate
need
of,
please
feel
free
to
propose
new
mappers
for
those
protocols.
C
The
next
item
that
we
we
are
looking
at
is
basically
how
do
you
improve
the
reliability
of
the
platform?
So
right
now,
if
you
see,
if
you,
if
you
have
communication
between
edge
nodes
and
the
cloud,
we
don't
have
any
reliability
built
into
the
platform
like
if
we,
if
we
send
some
messages
from
the
edge
to
the
cloud,
if
they
are
not
delivered
physically,
we
don't
have
any
mechanism
to
determine
whether
the
message
was
actually
delivered.
C
So
there
is
an
open
issue
for
that
and
other
than
that,
if
you
see
we
have
the
next
item
in
the
list
for
separating
the
path
of
the
control
plane
and
the
data
plane
right
now,
the
device
twin
metadata
is
also
passed
on
the
control
plane.
We
we
wish
to
have
two
separate
channels
for
sending
metadata
and
data,
and
currently
we
have
supported
quick
protocol,
which
is
a
UDP
based
protocol,
which
also
provides
a
concept
of
streams.
C
So
we
can
have
different
streams
for
different
kinds
of
data
that
is
passed,
but
we
have
seen
some
performance
performance
improvement
that
we
were
expecting
with
quick.
We
are
not
seeing
until
now
we
have.
We
are
still
evaluating
and
performing
the
performance
tests.
So
if
we
feel
that
you
know,
quic
is
not
as
useful
as
it
was
supposed
to
be
like
with
respect
to
performance,
then
we
need
to
think
of
some
other
way
to
provide
this,
the
separation
of
the
control
path
and
the
data
path
channels.
C
Other
than
that.
The
third
item
in
the
list
is
we:
we
want
to
switch
to
protobufs
based
messages
right
now.
All
the
messages
are
based
on
JSON
and
serialization,
and
deserialization
is
a
little
costly
when
you
send
messages
over
the
wire.
So
what
we
intend
to
do
is
like
try
to
compress
or
make
this
message
format
a
little
more
smaller,
so
that
we
can
have
lesser
overloads
or
lesser
kind
of
impacts
when
we
serialization
and
deserialization
of
the
data.
So
we
think
that
protobufs
is
a
good
option
for
for
this.
C
So
we
intend
to
replace
the
JSON
based
format
with
the
protobuf
paste
format,
so
I
think
Dec
is
working
on
that
from
the
community
and
the
last
item
that
is
there
in
the
list
is
ingress
support.
So
we
have.
We
have
seen
some
requests
from
the
community
members
that
they
are
trying
to
deploy
engine
X
ingress
controllers
and
they
are
not
able
to
get
it
to
work.
C
C
C
D
C
Yeah
right
now,
basically,
if
you
see
about
performance
framework
right,
so
we
are
basically
having
we
are
having
one
control
plane
for
the
for
the
cloud
and
and
we
treat
each
of
the
right.
We
bring
each
of
the
edge
nodes
as
a
single
border
inside
the
containers.
So
we
are
using
the
concept
of
for
darker
inside
darker
kind
of
concept
and
we
are
trying
to
basically
mimic
edge
nodes,
basically,
which
are
actually
VMs
our
raspberry
PI's.
C
We
are
making
them
as
containers,
so
each
edge
node
will
be
basically
a
container
that
comes
up
and
you
will
connect
to
the
control
plane.
So
the
design
for
this
performance
test
framework
is
there
in
github.
If
you,
if
you
want,
we
can
have
a
separate
discussion
on
how
we
achieve
this
kind
of
framework.
D
Ok,
so
which
mindspace?
If
we,
if
we
took
into
simulation,
is
a
container
or
a
guest
of
am
so,
which
one
is
better
I
think
it's
a
VM
should
be
better
VM.
C
Is
C
VM
will
be
limited
to
numbers,
I
think
we
are
targeting
some
right
now
hundred
of
edge
nodes
so
because
of
you
know,
constraints
related
to
number
of
VMs
what
we
thought
was
like.
We
just
need
to
test
the
interface
between
the
cloud
and
the
edge.
How
long
does
it
take?
You
know
to
deploy
a
pod
on
edge
node,
or
how
long
does
it
take
for
a
device
to
report
its
status
back
to
the
cloud,
so
we
are
not
basically
targeting
first
to
test
on
an
edge
node.
C
How
much
is
the
performance
gain,
but
the
interface
between
the
edge
and
the
cloud?
How
can
we
test
that
so
for
testing
the
interface
we
can
still
have
you
know
edge
nodes,
running
inside
containers
and
then
let
those
edge
nodes
generate
some
metadata
or
data,
and
then
you
test
this
interface
first.
So
once
the
interface
is
fine,
I
think
we,
you
can
easily
scale
to
you,
know
thousands
of
edge
nodes
or
to
thousands
of
edge
nodes,
since
they
are
all
running
in
containers.
C
D
D
C
F
I
think
we
can
divide
this
work
into
several
phases,
because
it's
constantly
we
we
just
have
a
one,
a
model
of
the
the
one
model
of
deployment
for
the
scalability
testing
and
we
know
yet
have
any
number
of
visually.
They
were
coming
out
from
the
community,
so
I
think.
First
of
all,
we
can
do
this
testing
to
to
get
some
number
about.
So
actually
this
was.
The
current
model
is
to
evaluate
Oh
from
the
control
point
perspective.
How
many
know
the
education
that
the
page
can
be
managed
with
the
other
set
of
requirements
there?
F
C
Moving
on
version
upgrades
support,
a
braids
of
you
know,
kubernetes
virgins,
and
the
second
topic
is
over-the-air
updates.
So
if
you
have
edge
nodes
which
are
running
edge,
coal
software's
and
you
want
to
basically
upgrade
them
on
the
fly
over
the
air
which
is
commonly
provided
by
you-
know
IOT
kind
of
solution
providers.
So
we
need
to
think
about.
How
can
we
do
these
kind
of
you
know?
Firmware
updates
or
any
other
security
patches
need
to
be
updated,
or
you
know
any
major
version
has
to
be
upgraded.
C
You
need
to
basically
think
about
how
to
do
that,
so
this
can
help
our
customers,
who
are
currently
using
cube,
h
1.0.
If
they
want
to
move
on
to
1.1,
then
what
is
the
approach
that
can
take
so
moving
on
to
the
node?
So
on
the
edge
node
side,
we
currently
have
the
CRI
interface
implementation
supported
and
we
support
container
D
as
one
of
the
remote
container
runtime
and
we
also
support
docker.
C
So
the
support
for
darker
via
the
doctor,
shim,
is
currently
in
progress,
and
you
can
see
some
peers
are
there
which
which,
which
will
allow
you
to
basically
use
any
any
container
runtime
which
is
CRI
compliant.
So
support
for
docker
and
container
ID
is
there
if,
if
we
have
any
more
other
container
when
times
like
cRIO
or
any
other
runtime
that
customer
needs,
we
can
consider
that
and
add
support
for
that.
C
C
Basically,
how
do
you
validate
whether
you
are
CR
D
input?
Data
is
correct,
but
still
there
is
a
lot
of
work
to
be
done
with
respect
to
managing
the
sanity
of
your
data,
and
how
do
you
basically
allow
deletion
of
for
device
models
if
there
are
devices
which
are
referring
to
it
and
there
is,
and
there
are
mappers
who
are
already
using
this
metadata?
How
do
you
manage
them?
How
do
you
do
you
restart
them,
or
do
you
watch
on
this
config
maps
and
then
reload
them
stuff,
like
that,
so
yeah.
D
E
D
D
Something
should
follow
the
schema,
validation
and
but
it
cannot
capture
all
the
missed.
Configuration
is
in
the
llamo
file.
So
besides
is
a
basic
CID
validation,
feature
I'm,
I,
think
Easter
know
the
mission
webhook
framework.
It
might
be
a
good
option
to
also
enable
for
the
cooperage
project,
so
current
I'm
still
investigating
into
this
feature.
This
feature-
and
this
feature
is
also
provided
by
the
communities
and
the
concern
I
have
is
that
this
my
man
needs
some
more
dependency
that
will
be
dropped
in
the
bender,
the
realtor
II.
D
D
D
C
G
A
F
D
D
A
F
A
A
Hey
guys
like
like
that,
he
will
be
know
like
we
are
getting
more
and
more
yes
and
committee.
Participation
has
also
got
increased
so
now,
like
we
should
start
looking
into
the
having
a
more
organized
manner
or
like
we
can
see.
The
kubernetes
community
is
how
they
are
evolving,
like
in
a
very
organic
ulm,
an
ER.
So
when
I
organize
my
methodology,
so
I
refer
some
of
the
documentation.
A
I
understood
like
it's
time
like
we
should
start
form
some
of
the
groups,
especially
they
have
some
different
groups.
So
I
thought
like
we
can
have
some
sig
groups,
basically
I
thought
like
I
thought,
like
even
with
discussing
with
the
keywords
maintain
us,
we
can
have
some
documentation
group
sig,
release
group.
A
So,
keeping
this
in
mind
like
we
wanted
to
have
some
people
from
the
community
get
into
the
system
and
they
may
be
having
some
inclination
towards
there
are.
Some
people
may
have
inclination
towards
improve
the
documentation,
part
from
people
on
the
release,
document,
released
packaging
or
agreeing
scripts
towards
that,
and
some
people
may
be
interested
in
the
architecture
designing
and
the
development
of
that.
So
keeping
these
points
in
mind
wanted
to
have
like
cupboards.
A
A
A
So
any
like
short-term
objective,
like
which
we
want
coverage,
maintain
us
think
like
yeah,
we
can
have
it
as
a
work
group
and
we
can
achieve
this
creation
of
work
group
right
with
a
perfect
purpose
so
which
can
be
a
shortly
for
the
objective
so
that
will
be
handled
by
the
work
group.
So
we
can
see
there
are
several.
A
Actually,
this
I
have
that
imprudently
written
is
a
little
bit
reform
and
checking
to
air-dry
cubed
G
Drive
and
share
the
link.
Okay,
thank
you
because
none
of
the
things
that
these
are
also
referring
to
the
key
words
community
governance
of
the
policy
documents
only
so
respect
you
referring
to
that
I
am
defining
which,
which
keeps
better
taught
up
from
the
key
beds
as
of
now
so
have
to
refine
it.
A
A
This
group
will
be
basically
aiming
to
give
the
requirements
with
the
Qbert
sig
groups
so
and
they
will
not
be
influencing
the
race,
but
indeed
they
will
be
supporting
to
the
supporting
to
the
like
commercial
teams,
production
teams
which
they
which
they
are,
which
they
want
the
requirements
to
be
addressed
in
the
kibbutz.
So
this
way
like
having
some
different
set,
some
grooves
with
particular
objectives,
right
I
feel
like
I,
owe
like
we
feel
like.
We
can
slowly
start
having
the
structure
governance
in
our
cube
edge
project.
A
So
this
is
my
thoughts
based
on
that
respectively.
I
will
start
writing
some
other
documentation
of
the
governance
related
points
for
a
specific
sig
group,
and
we
will
hear
these
thoughts
in
the
channel,
so
I
will
check
in
this
one
after
some
a
little
bit
of
refinement
and
share.
This
part
details
in
the
slack.
A
Any
any
any
thoughts
you
people
have
which
want
to
suggest
to
us
right
so
that,
like
we
can
improve
the
community
participation
and
they
can
bone
certain
areas
which
they
like
to
do.
Like
you
see,
people
saw
like
a
lot
of
features.
There
are
categories
related
to
network
storage,
upgrade
several
things
are
coming
in
too
slowly.
We
have,
we
can
expand,
it
divide
it
in
two
different,
seek
group,
verticals
agenda
or
dental
six
kind
of
thing,
and
that
particular
group
can
work
in
that
area.
F
This
is
Kevin
I,
think
from
the
beginning,
from
my
perspective,
I
think
released
and
the
documenting
very
important
so
for
the
architecture
I
think.
Actually
we
today
we
have
met
unless
reviewing
all
the
design
proposals.
So
that's
why
I
am
open
either
we
want
to
make
it
to
say:
you're
not,
and
therefore
doesn't
man
and
the
release
I
think
that's
important.
I
also
heard
some
of
the
community
users
that
they
still
think
that
there's
there
are
not
enough.
Documents
were
used.
F
Either
users
were
contributors,
we
definitely
need
more
documents
to
guide
users,
to
torture
or
or
help
contributors
to
contribute
and
I.
Think
we
for
the
project
development
we,
whereas
we
have
many
things
on
the
list,
so
have
a
have
a
group
of
people
and
a
lot
of
effort
on
the
list
make
it
a
better
way
is
very
important.
F
But
just
as
currently
the
expression
is
been
a
little
bit
confusing
because
you,
you,
like
you,
have
the
description
under
the
approver
were
strictly
the
measures
only
one
community
member
so
I
want
to
suggest
to
you
maybe
turn
into
two
how
many
signals
we
would
have,
whereas
they
were
so
immersed.
Actually
it
should
be
open,
so
everyone
can
just
the
joy
or
any
other
special
roles
we
need
were
for
running
a
saying
word
or,
for
example,
when
running
a
released.
We
need.
A
F
A
F
A
A
A
A
B
Okay,
yeah
yeah,
yeah
yeah.
The
last
part
is
tom
response
and
the
user
kiss
from
Cudahy.
First,
his
summers
folks
and
you
you
said
to
me
that
Krueger
edge
as
the
only
while
communities
native
actual
here,
the
framework
they
are
very
willing
to
child,
but
they
said
we
need
to
improve
the
dog,
especially
the
part
of
install
example.
So
so
so
they
can
more
easy
to
try
to
China
yeah
yeah.
B
This
is
a
response.
Next,
in
some
conscience,
the
users
talk
to
me.
The
first
is
whether
the
crib
had
to
cover
the
situation
at
the
edge
part
move
from
Michael
what
well
being
to
another.
B
Not
yes,
don't
we
we
need
help,
we
don't
have
any
change
to
the
crowed
part
to
the
committed,
so
in
equipment
hits
when
the
node
of
life,
the
application
will
my
grades.
No.
Yes,
here
here
maybe
ring
to
Edison
at
made
admission
and
SRAM
cook
to
add.
Some
information
tends
to
know
that
application
to
you
to
make
that
don't
you
don't
migrate
to
the
edge
part
of
nine?
The
next
is
how
to
manage
the
edge
side
in
the
corner,
as
no.
We
already
have
the
other
side
in
in
the
edge.
B
B
D
B
F
I
A
F
Wage
because
we
think
that
not
necessarily
so
in
then
ways
people
may
want
to
try
use
vanilla
from
that
age,
so
they
so
they
just
know
everything
so
I
think
I
think
at
least
the
one
thing
clear
we
can
do
is
that
way
we
can
identify
exactly
the
features
the
core
data
currently
from
th.
The
document
set
I.
F
F
So
I
think
I
think
we
wins.
We
need
to
need
to
document
that,
and
some
people
may
not
quite
clear
about
what
the
what
the
functionalities
the
they
exactly
needed
age
is
so
so
in
one
case
today
is
the
one
want
everything
and
the
way
link
was,
will
expire
hours
below
vision,
but
what
what
are
necessary
a
day
will.