►
From YouTube: Network Service Mesh Meeting - 2019-07-16
Description
Join us for Kubernetes Forums Seoul, Sydney, Bengaluru and Delhi - learn more at kubecon.io
Don't miss KubeCon + CloudNativeCon 2020 events in Amsterdam March 30 - April 2, Shanghai July 28-30 and Boston November 17-20! Learn more at kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects
A
A
If,
if
anyone
has
anything
that
they
want
to
talk
about,
that
is
not
on
the
agenda,
please
feel
free
to
add
it
so
getting
started.
So
we
have
a
few
recurring
calls.
We
have
this
call.
We
have
the
NSM
document,
call
which
we
may
switch
to
monthly
and
we
have
the
MSM
a
use
case
call
which
appears
every
second,
fourth
and
fifth
Monday
of
each
call.
Prime,
do
you
know
that's
going
on
this
next
week
or.
C
A
B
A
Yeah
it's
up
to
it's
up
to
to
the
both
of
you
in
terms
of
how
you
want
to
how
you
want
to
run
it.
So
if
you
feel
that
your
that
the
community
is
getting
a
lot
out
of
the
current
cadence
and
feel
free
to
keep
it
as
it
is,
if
you
feel
that
it
needs
to
be
reduced,
then
you
know
three
to
four:
you
reduce
it.
A
And
third
Monday
at
8:00
a.m.
Pacific,
and
so
we
just
had
one
this
last
this
last
week
we
have
the
CN
CF
networking
working
group,
which
occurs
every
two
weeks
on
Tuesday
and
I
am,
and
there
is
a
so
in
terms
of
events.
We
have
DVD
K
user
space,
which
I
have
a
I,
have
a
talk
that
was
accepted
into
that.
So
that's
on
the
MSM
in
the
NDB
K
and
in
terms
of
what's
going
on
there
we
have
in
September,
19th
and
20th.
There
is
a
open
networking.
A
Sorry,
there
is
a
open
core
summit
in
San,
Francisco
I.
Don't
think
anyone
has
reached
out
on
that
right,
primarily
because
of
release
on
my
party
I'm
based
in
San,
Francisco
I,
think
I'm,
the
old
well
rather
from
from
the
innocent
commuters
I'm,
the
only
person
who
is
based
in
this
area,
but
perhaps
because
perhaps
we
could
get
someone
else
was
heavily
involved
in
the
community.
So,
like
pre
media
talks
on
an
SM.
This
is
that
something
that
perhaps
you
can
reach
out
in
advance.
You
can
organize
or
yeah.
B
A
A
D
A
On
October
8,
so
we'll
take
a
look
at
take
a
look
at
that
as
well.
We
have
open
source
summit
which
Nikolay
has
a
and
run
at
Slava
navona
have
talks
that
have
been
submitted,
so
we
should
hear
about
those
in
early
August
and
we
have
mystic
on
with
Ivana
stock
accepted
in
November
14
through
15
yeah.
A
We
have
tube
con
coming
up
in
1883
24
person,
so
we
have
a
few
talks
that
have
been
pushed
in,
or
rather
that
have
been
submitted
in
so
we'll
see.
If
we'll
see,
if
we
well,
we
will
see
which
talks
get
get
in
on
that
one.
So
we'll
know
more
soon
on
that
and
the
announcements
will
should
be
in
early
September.
A
There
was
a
co-located
event
with
on
the
way
con
I
need
to
check
that
there
was
a
team
that
have
reached
out
about
this.
Do
that
was
thinking
about
it,
putting
something
in
with
no
sorry
envoy.
They
were
sorry
about
putting
something
in
with
on
way
con
Nick
white.
You
know
they
ended
up
submissive
meeting
into
you
into
all
the
way.
Con
I.
C
A
C
A
A
We
don't
have
any
detailed
list,
we
will
have
to
circle
back
on
that
we
have
when
we
are
ready.
We
have
a
call
for
demos
at
the
kubernetes
community
meeting,
which
occurs
every
Thursday
demos
are
booked
at
a
month
and
a
half
or
more
out
or
about
a
month
out.
So
we
need
to
give
them
a
little
bit
of
RAM
or
ample
time
to
to
submit.
F
A
F
That's
before
zero
one
zero
release,
so
that's
really
great
can't
wait
to
see
what
happens
when
we
go
forward.
So
that's
23
in
a
week
and
I
followed
about
83,
more
counts,
I'm
surprised,
I'm,
even
finding
any
more
related
folks,
but
there's
a
lot
of
folks
in
this
space
and
just
keep
using
the
Twitter
who
I
should
follow
recommendations
and
posted
and
retweeted
24
times
this
week.
There's
a
lot
of
interesting
news.
F
So
today
the
300
followers
I
posted
a
thank
you
post
last
week
posted
a
reminder
about
the
Cellcom
user
group.
This
week
there
was
a
press
release
out
of
Verizon
and
Ericsson
first
in
the
world
to
introduce
cloud
native
technology
in
the
air,
live
wireless,
core
network
environment,
so
retweeted
that
I
also
reach
weeded
a
post.
The
now
is
the
time
for
telcos
to
adopt
cloud
native
micro
services.
That
was
a
pretty
interesting
article.
A
A
C
So
big
I,
don't
think
that
we
have
this
custom.
Maybe
I
have
skipped
a
couple.
So
I
don't
know
if
this
has
been
already
being
discussed,
but
because
there's
these
wallpapers
going
on
and
a
couple
of
us
are
attending
and
participating
I
know
that
Jeff
is
one
of
the
read
persons,
so
he's
not
here
I
believe
they
feel
not
so.
C
Essentially,
there
are
two
two
white
papers:
one
is
small,
QG
centric
and
the
other
one
is
more
supposed
to
be
like
more
of
service
providers
and
drink,
at
least
as
a
starting
point.
So
I
would
I
would
really
like
to
have
some
some
discussion
within
within
our
community
and
how
we
want
to
position
ourselves.
There
I
mean.
Are
we.
C
Are
we
perceiving
our
surface
as
and
are
we
actually
looking
for
being
the
one
of
the
main
drivers
there?
Do
we
want
to
push
NSM
somehow
deeper
into
the
into
the
into
the
ecosystem
and
kind
of
from
the
whitepaper
point
of
view,
or
we
somehow
want
to
to
just
be
one
of
the
many
possible
solutions
and,
as
an
example,
I
will
I
will
give
honest,
because
this
is
the
the
hot
news
that
actually
we
were.
A
couple
of
our
talks
were
accepted
so
essentially
I'm
on
to
talks.
C
Don't
know
Fred
if
you
would
agree
here,
but
I
think
that
this
this
is
how
how
we
are
going
to
approach
the
talk
saying
that,
okay,
even
even
if
you
give
an
example,
you're
just
going
to
say:
okay,
okay
or
I'm,
going
to
say:
okay,
so
you
have
the
endpoint,
then
you
can
change
it
to
this
something
else.
Then
you
can
help
the
network
service
being
composed
out
of
smaller
endpoints.
C
C
A
A
A
So
our
our
pitch
has
been
to
use
device
plug-in
for
injecting
special
pieces
of
hardware
in,
and
we
know
that
there's
issues
around
that.
So
that's
something
that
will
very
likely
come
up
but
another
one
with.
What's
what
it's
going
to
lead
to
as
well
as-
and
this
is
quite
understandable-
because
a
lot
of
vendors
they
want
to
take
what
they've
already
created
and
they'll
want
to
run
that
within
a
within
a
pod
or
within
a
VM.
A
A
Imperative
or
it
doesn't
follow
like
they
stand
back
cloud
native,
declarative
style,
then
you're
like
we're
gonna
run
into
issues
with
that
and,
and
so
I
think
part
of
some
part
of
the
goal
into
scenario
is
going
to
be.
How
do
we?
How
do
we
convince
the
community
that
cloud
native
is
more
than
just
a
lift
and
shift?
This
is
not
like
a
PMF
to
DNF
use
cases
is
this
is
equivalent
to
the
applet
back
to
many
applications
when
you're
personally
mundaka?
A
How
do
we
move
applications
from
being
giant
monoliths,
which
you
just
stick
in
a
container
and
then
deploy
and
say:
ok,
containerize
now
to
actually
building
to
kubernetes
and
and
being
able
leverage
current
any
strengths,
because
you
you
only
you'll
see
the
end
result
is
if
you
just
still
lifted
a
shirt
and
you're
not
ready
for
the
paradigm.
Is
your
you
do
end
up
having
you
and
a
brief
certifying.
A
Alright,
so
I
think
I
think
that's
where
the
I
think
that's,
where
the
real
benefit
that
we
can
try
to
push
is
is
to
say:
hey,
don't
don't
just
do
a
lift
and
search
like
actually
redesigning
or
you're,
really
architect,
redesign
redesign
your
system
and
start
to
it
doesn't
happen
all
be
at
once,
but
to
move
towards
a
more
cognitive
environment.
The
the
make
sense.
C
Yes-
and
this
is
this-
has
been
something
that
we
are
discussing
over
and
over
again
I
mean
that's,
that's
okay,
I
mean
I,
don't
have
any
objections
here,
but,
for
example,
yesterday
down
kick
came
up
with
a
link
and
then
we
discussed
I,
don't
remember
the
name
of
the
guy
from
Eric's
home
that
they
are
deploying
in
I,
think
it
was
Verizon,
they
have
a
cloud
native.
You
know,
Mme
I
think
it
was
so
essentially
they're,
saying:
okay,
we
can
do
this
with
motors
and
it
works,
and
we
don't
really
need
all
this
dynamism.
A
C
But
that's
what
what
actually
reality
shows
us
so
I
mean
if
we're
approaching,
and
then
we
want
to
say
all
the
things
that
you
you
you're
saying
now
we
really
need
to
go
with
some
strong
case.
We're
actually
on
the
same
is
really
because
the
best
thing
about
NSM
in
this
case
is
that
it's
dynamic
I
mean
like
in
this
specific
case
at
Eric's,
are
demonstrating.
C
There's,
essentially
you
deploy
it
once
then,
you
use
motors
to
just
statically,
you
know,
connect
all
the
interfaces,
I
guess
it's
SRO
VM
whatever,
and
you
connect
your
data
plane
tool
to
the
SLV,
and
then
I
mean
you.
Don't
move
this
much
right,
I
guess
that
I
mean
this
is
not
criticism
to
their
solution.
This
is
like
people
are
already
showing
some
real-world
solution
that
works
with
the
existing
technologies
and
doesn't
really
I
mean
it
does
its
job
without
really
using
the
the
nice
things
that
we
we
can
offer
here.
C
So
if
we're
going
to
the
white
paper,
we
need
to
come
up
with
a
strong,
strong
use
case
strong,
like
yeah
use
case
and
so
business
business
logic.
Behind
it
saying
yeah
I
mean
if
you
are
deployed
this,
then
you
will
save
this.
This
amount
of
effort,
this
amount
of
money,
even
if
you
can
mop
mop
it,
you
know
operations
whatever,
but
we
need
to
come
up
with
something
something
really
strong.
Otherwise,.
A
B
B
The
reason
was
I
validate
it,
some
of
these
use
cases
and
the
benefits
with
some
of
the
t1
service
providers,
but
by
and
large
the
issue
here,
a
couple
of
things
oneness
today,
as
we
speak,
the
industry
I
mean
there
is
a
real
need
for
such
architectures,
because
the
whole
infrastructure
has
to
be
dynamic
and
then
it
has
to
essentially
be
ready
to
accommodate
some
of
the
use
cases
and
also
Phi
G.
Pretty
much
drives
this
requirement.
I
would
say
so
what
I
was
thinking
was.
B
Probably
we
can
just
take
a
news
case,
a
bunch
of
use
cases
and
then
probably
highlight
some
of
the
benefits
on
how
this
new
architectures,
particularly
cloud
native,
will
help
out
in
the
in
realizing.
Those
use.
Cases
would
be
a
good
input
to
it,
and
also
probably
I
can
share
the
slide
which
I
have
presented
during
the
last
one.
A
A
They
have
to
to
touch
the
various
seer
knees
directly
to
specify
IP,
subnets
and
various
other
things,
and
so
we
can
that's
what
area
where
we
can
help
provide
that
that
and
dynamism,
and
once
you
have
that
that
process
automated,
then
you
can
do
more
interesting.
Things
like
Auto
cooling
is
similar
stuff,
so
I
think
we
can
we'll
be
able
to
push
in
that
direction.
A
A
What
doesn't
work,
and
my
intuition
on
this-
is
that
there
may
be
room
for
additional
experimentation
that
they
may
have
been,
maybe
willing
to
do
so
if
we
engage
them
and
try
to
collaborate
with
them,
it's
possible
that
we
could
convince
them
to
give
NSM
a
try
and
and
show
them
how
and
and
show
them
the
the
benefits
that
that
we
bring
to
the
table.
So
so
I
think
we
should.
C
C
B
So,
okay,
so
here
does
what
it
is
right,
so,
particularly
with
respect
to
5g,
it's
the
same
network
infrastructure,
but
it
has
to
provide
you
or
it
has
to
support
a
lot
of
different
use
cases,
the
use
cases
meaning,
for
example,
we
are
talking
about
ultra
reliable
low
latency.
We
are
talking
about
mass
of
IOT,
it's
going
to
the
same
infrastructure
that
has
to
be
providing
the
support
for
that.
So
in
that
case,
what
what?
B
That
is
number
one
number
two
also
the
users
are
changing
where
and
the
requests
from
the
users
would
be
done,
for
example,
dynamic
installation
of
services
right,
so
what
it?
What
dynamic
in
sensation
of
services
mean
is
that
the
infrastructure
is
assumed
that
it
has
to
meet
the
service
SLS
as
well
as
the
needed
requirements.
B
So
all
of
it
would
essentially
mean
that
the
whole
infrastructure
has
to
be
composable
when
I,
say
composable,
you
would
need
you
get
in
terms
of
the
intent
and
the
intent
has
to
be
directly
translated
into
what
is
that
that
is
needed
by
the
infrastructure
right.
So
that
is
where
the
concept
of
composable
infrastructure
place
in
it
comes
in,
or
it
plays
a
very
important
role
right.
So
now,
I
can
probably
just
give
an
example
right.
B
So
if,
for
example,
in
a
particular
area,
someone
wants
to
bring
up,
say,
100,
say
or
100
k,
subs
right,
meaning
subscribers
right.
So
in
that
case,
in
the
in
the
s12
world
where
and
when
things
are
static,
you
would
essentially
need
to
create
the
network
functions,
which
is
essentially
be
it
physical
into
a
function
or
a
virtual
into
a
function
and
then
ot.
Essentially,
it
would
essentially
take
like,
for
example,
a
few
months
to
get
the
infrastructure
up
and
ready
right.
B
But
now,
when
we
talk
about
these
new
use
cases,
they
would
need
these
services
to
be
immediately
available
right.
So
how
do
you
release?
Is
the
first
question
so
to
realize
that
what
you
can
do
is
you
can
always
contain
erase
all
of
these
Network
functions
and
then
probably
put
it
in
terms
of
helm,
charts
and
then
probably
instantiate.
Let
me
give
me
a
second.
Let
me
share
you
the
slide
so
that
it
makes
easy
for
people
to
follow.
Can
you
give.
C
B
B
These
are
the
five
say,
network
functions
that
needs
to
be
instantiated
right,
so
what
I
do
is
I
would
essentially
pull
it
from
the
home
charts
and
then
I
would
essentially
instantiate
it
where
kubernetes
now,
depending
upon
my
previous
data
I
know
what
has
to
be
the
current
state
and
what
has
to
be
desired
state
and
then
it
would
essentially
automatically
go
and
instantiate
right
now.
This
is
with
a
generic
with
generic
kubernetes
right.
B
You
would
essentially
need
to
look
at
these
network
service
functions
to
be
like
impossible
components,
and
you
would
essentially
compose
your
your
function
because,
for
example,
this
is
quite
common
in
the
network
service
mesh
world
right
because,
for
example,
the
network
service,
whatever
you
would
offer,
would
essentially
differ.
For
example,
if
it's
going
to
be
a
secure
internet,
it
would
mean
that
I
would
have
a
firewall.
I
would
have
a
DPA
and
then
instantiated
the
same,
secure
Internet
would
mean
a
different
context
in
different
scenarios
right,
so
what
I
mean
is
the
name?
B
C
C
To
implement
the
CNF
concepts
are
more
or
less
static,
and
not
that
dynamic
and
with
technologies
like
NSM.
Essentially,
you
are
able
to
be
a
lot
more
dynamic
and
more
declarative,
so
that,
essentially,
you
enable
more
use
cases
and
this
composable
infrastructure,
which
is
more
or
less
the
foundation
of
5g
implementations
in
the
future.
A
Like
the
in
terms
of
where
we
sent
that
were
able
to
really
help
in
so
one
of
the
big
problems
in
multiples
has
is
that
when
it
runs,
there's
no
file
system
available.
So
when
you
try
to
when
you,
if
you
want
to
run
something
that
certain
memory
based,
you
have
to
be
able
to
mount
the
inode
somewhere
so
that
you
can
then
active
so
that
you
can
access
it,
and
so
when
C&I
runs,
there's
literally
nowhere
for
for
that
mounts
to
to
occur,
and
so
you
end
up
having
to
do
out-of-band.
A
Some
significant
amount
of
man
configuration
in
order
to
in
order
to
land,
faster
paths
from
device
access,
and
so
that's
that's,
there's
a
there's
a
couple
areas
where
this
this
may
not
be
entirely
true,
like
you
can
do.
If
you're.
If
you
want
to
stick
in
a
device,
then
you
can
use
the
vise
plugin
to
land
that
device
directly
into
your
into
the
into
the
file
system.
A
When
C&I
runs,
it
runs
very,
it
really
runs
before
all
of
that
has
been
set
up.
So
if
you
wanted
to
do
some
set
of
configuration
to
that
device,
CNI
is
not
the
right
place
to
do
that
and
motifs
is
a
C&I,
so
it's
gonna
run
into
the
various
limitations
of
CNI
itself.
When,
when
this
happens,
the
only
thing
that
C&I
has
access
to
is
is
it
has
privileges
and
it
has
access
to
the
network
namespace
of
the
pond.
So
it
knows
what
the
name
of
our
game
space
of
the
pot
is
so.
A
A
Some
of
the
interesting
work
that
we've
brought
up
into
into
the
fray
and
we'll
be
able
to
we'll
be
able
to
win
support
with
with
those
on
a
technical
level
as
well.
But
it's
not
it's
not
just
like.
Oh
no,
it
just
needs
more
work.
It's
like
no.
This
is
spun
as
playing
in
a
space
that
requires
significant
changes
to
kubernetes
in
order
to
make
it
work,
which
some
of
those
patches
have
been
submitted
in
one
of
the
big
ones.
A
Whether
there
was
a
big
there
was
a
device
plug
in
and
multi
device
plug
in
and
see
an
eye
patch.
That
was
it
was
that
dim
and
melon
was
actually
blocked
by
Clayton
Coleman,
who
is
I,
think
the
chief
architect
of
openshift
over
at
Red
Hat,
because
of
the
complications
to
to
kubernetes
it
would
that
it
would
ring.
So
so
that's
that's
another
area
that
we
were
able
to
you're
able
to
talk
about
is
like
how
how
we
were
able
to
do.
C
G
So
I
think
it
was
the
that's
that
goal.
So
we
had
a
lot
of
work
to
clean
up
the
mess
of
which
white
paper
to
write
and
what
kind
of
write
paper
to
write.
So
we
kind
of
end
of
that
now.
The
next
part
was
actually
to
be
able
to
discuss
and
look
at
other
technical
issues
and
the
feasibility
and
which
it
makes
more
sense
and
I
think
over
by
discussing
it
and
exploring
those
and
I
using
the
CNF
testbed.
G
That's
the
purpose
to
be
able
to
actually
I
lied,
those
who
actually
makes
more
sense
on
the
engineering
side
in
their
feasibility
side
and
that
me
as
an
end-user
and
operator.
That's
what
I
want
from
that
white
paper
is
I,
don't
expect
there
will
always
be
a
single
solution,
I
think
by
design
it
will
always
be
some
sensitive
or
emotional
attachment
to
some
technologies
and
easier
streets
always
been
like
this.
G
So
a
but
it's
common
sense
and
requirements
and
features
and
gaps
are
going
to
be
shown
so
that
we
actually
know
which
one
makes
more
sense,
and
it
is
not
going
to
be
back
to
the
emotional
state
of
deciding
which
technology
you
pick
and
choose
so
I
think
there's
a
lot
of
potential
true
that
work
at
that
effort
of
figuring
out
those
gaps
to
see
well,
this
is
already
covered
to
@fm.
This
NSM
is
you
can
easily
provide
an
hour
or
adjust
so
I
think
that
the
that's
the
best
framework
to
prove
it
up.
A
Yeah
I
actually
see
the
I
show
the
same
view.
I
think
that
we're
gonna
hear
a
lot
of
noise
until
he
actually
starts
taking
these
things
into
test
beds
and
comparing
them
side
by
side,
and
once
we
once
we
do
that.
I'm
confident
that
I'm,
confident
that
NSM
brings
enough
value
to
the
to
the
table
that
that
it'll
end
up
dominating
the
space,
at
least
in
the
corner
of
the
nets
and
plays
with,
but
but
we're.
The
only
way
to
really
get
that
message
out.
A
It's
literally
going
to
be
go
implemented,
go
find
out
what
the
problems
apart
and
this
will.
This
will
be
good
regardless,
like
even
it,
even
if
it's,
even
if
it
was
not
understand
that
one
out
it'd,
be
good
to
get
this
type
of
thing
out
anyway,
because
in
the
long
run
we're
looking
for
we're
looking
for
solution
to
to
quad
native
networking
problems,
and
so
whatever
that
solution
looks
like
we're.
A
B
G
G
I
have
found
the
ways
that
are
being
proposed
to
us
I'll
find
a
way
to
have
like
books
to
a
specific
kind
of
hardware,
design
or
platform
design,
meaning
that
if
you
want
to
be
able
to
lead
truly
composable
and
M
platform,
agnostic,
we're
not
there
yet
and
that's
where
it's
challenging
to
say
to
have
a
vendor.
Xyz.
Give
me
your
give
me
your
L
chart
or
your
York
way
image
to
your
CNF,
not
there
yet
and.
A
G
A
challenge
that
that's
gonna
be
the
first
challenge,
I
think
it's
the
first
challenge,
we're
gonna
face
is
being
able
to
say
your
workload
should
be
agnostic,
whether
I
run
on
a
VMware
version
of
purities
or
open
ship
or
something
else
I
had
n
FM
or
not.
You
should
be,
you
should
be
agnostic
to
it
and
where
that's
the
first
first
first
big
challenge
were
I'm.
Seeing
coming.
Yes,.
B
I'm
so
fully
agree
with
you,
and
one
question
I
have
is,
would
clear
what
Rio
must
be
of
any
help.
Can
we
start
looking
at
it
so
that,
because
they
do
have
some
darker
limits
and
they
have
in
a
way,
I,
don't
know
fully,
but
they
have
containerized
it.
So
it's
just
curious
I
started
looking
at
it,
but
it
was
quite
complex,
so
just
wanted
to
get
your
inputs.
What.
B
B
C
We
were
able
to
enable
the
multi
host,
like
the
explained
connectivity,
multi,
hosts
and
now
we're
talking
about
like
this
is
among
seven.
Eight
months
later,
we
are
talking
about
interval,
main
meaning
multi
cluster,
being
able
to
spread
services
and
consume
them
and
declare
them
in
different
kubernetes
clusters.
I
think
that
this
is
a
huge
huge
milestone
for
us.
I
mean
it
will
probably
take
a
while
until
this
gets
in
it
will
probably
take
a
couple
of
iterations
until
it's
stable.
D
C
It
yeah
probably
some
some
more
work
here.
I
put
some
comments
around
documentation,
but
yeah
other
than
that.
It's
it's
it's
great
that
we
already
have
this
and
it's
probably
again
one
of
the
unique
features
that
that
we
can
offer
to
the
to
the
world.
C
C
A
Yeah
we're
coming
up
to
the
top
of
the
hour
as
well,
so
I
think
we
should
probably,
which
probably
ended
up,
ended
here
and
yeah,
just
just
for
one
last
comment
on
the
inter
domain
stuff.
So
I
think
this
is
one
of
the
area's
that's
gonna,
be
incredibly
powerful
for
us,
so
I'm,
really
looking
forward
to
seeing
the
inter
domain
stuff
come
come
up
with
its
first
full
iteration.
So
definitely
amazing
work
for
that
team
and
with
that,
is
there
any
last
agenda
items
we
want
to
cover
before
we
finish
or
any
last
bounce
ones.