►
From YouTube: CNCF Network Service Mesh Meeting - 2019-07-30
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
B
Right
so
welcome
to
the
network
service
match
meeting
yeah.
We
have.
We
have
three
general
meetings.
We
have
this
meeting,
which
occurs
every
Tuesday
at
8
a.m.
there
is
a
MSM
document
meeting
which
is
going
to
be
switched
to
every
month,
and
there
is
a
use
case
meeting
which
has
been
postponed
for
July,
since
people
running
it
have
been
have
been
out
for
various
reasons,
but
will
reconvene
on
August
13th,
which
is
approximately
two
weeks
from
now.
B
B
I,
don't
know
when,
where
the
other
I'll
check
up
on
that
we,
we
also
have
events
coming
up.
We
have
DVD
users,
DP
DK
users,
face
in
Bordeaux
France.
Where
we'll
talk
about
DVD
K,
we
have
a
potential
talk
in
September,
19th
or
20th,
with
open
core
summit
depending
on
depending
on
Prem
Prem
is
not
on.
Today
we
have
ons
Europe
or
we
have
a
extensive
schedule,
and
that
includes
discussions
on
CNX
discussion
on
the
CNF
test.
Bed,
kernel-based
boarding,
planes,
service,
mesh
interface,
interoperability
with
MSM.
B
We
have
the
CN
CF,
Telecom
user
group
widget
will
be
ran
by
then
and
hide
n
cone
and
tailor.
We
also
have
a
tutorial
on
driving
telco
performance
using
the
testbed,
please
preregister.
If
you
intend
to
go,
we
have
embracing
cloud
native
on
the
path
to
5g,
which
is
a
panel
discussion
and
I'm
sure
there'll
be
something
else.
It
will
pop
up.
B
B
There
is
open
source
summit
coming
up
in
October
28th.
The
call
for
papers
is
already
closed.
We
are
waiting
for
notifications,
this
Monday
on
whether
Nikolai
and
rattle
Slava
Nirvana's
talks
have
been
accepted
or
not.
There
is
the
edge
Congress
in
Austin
and
the
schedule
I
see
the
schedule
announced
with
a
link.
So
there
is
during
the
schedule.
B
I
don't
know
if
we
have
anything
on
MSM,
but
it
doesn't
look
like
it,
but
it
still
may
be
useful,
especially
if
you're
in
the
area
we
have
istic
on,
of
which
Ivana
has
a
talk
accepted,
not
only
in
Sofia
Bulgaria.
In
November
we
have
the
cube
con
and
late
in
November.
You
can
do
21st
coming
up
wish
to
call
for
papers
are
already
closed.
We
have
multiple
talks
that
have
been
submitted.
We
will
get
notification
on
them
on
September
3rd.
B
C
D
B
E
Thank
you
so
much
with
the
announcement
of
the
open
networking
summit
in
Europe
being
announced.
It
was
easy
to
find
a
lot
of
content
to
post
to
the
end
service.
Mesh
Twitter
account
so
posted
and
retweeted
35
times
gained
27
more
followers
followed
a
few
more
related
folks.
This
week's
plan
I'll
continue
announcing
the
ons
EU
events
I'm
doing
one
a
day
so
I'm
near
the
end
and
towards
the
top.
E
D
E
You,
yes,
that
was
a
surprising
number
for
this
week
as
well,
but
we've
got
a
lot
of
cool
things
going
on
and
exciting
events
to
look
forward
to
so
keep
on
posting
about
our
events
and
I'll
also
share
the
0-100
release
announcement.
When
ready
are
there
any
other
items?
You'd
like
me
to
look
into
for
this
week,
I.
C
Yeah
I
mean
I
would
like
to
use
this
in
my
immunity.
Essay
I
would
like
to
use
this
as
a
kind
of
move
to
the
next
topic
here.
If
you
don't
mind
losing
so
I
will
start
start
to
start
running
this
Asia
friendly.
If
we
want
to
call
it
like
this,
probably
not
the
will
group
code
that
actually,
okay,
as
I
wrote
here,
it
should
happen
every
second
Tuesday
10:00
to
10:30
a.m.
CET.
That
would
be
like
4:00
p.m.
in
Shanghai
and
the
zone
their
time
zone
there.
F
D
You
know
they're
not
staying
up
until
11
o'clock
at
night
to
attend
so
I
think
we
can
definitely
get
some
promotion
behind
this,
but
let's,
let's
get
it
up
in
all
the
usual
places
and
get
a
tweet
going
and
everything
else,
and
thank
you
so
much
for
being
willing
to
do
this.
You
know
it's
you
for
those
of
us
in
North
America
trying
to
do
something
at
4
p.m.
and
chain
gang
ties.
Time
is
unbelievably
painful
for
you,
it's
hard
to
say,
because
you're
always
up
so
late,
but
I'm,
hoping
that
10
a.m.
C
Idea
is
primarily
to
gather
feedback,
and
someone
is
if
they're
sitting
sketching
any
species
specific
topic
to
try
to
bring.
You
know,
questions
or
whatever
is
needed
to
the
workgroup
call
later.
So
what
I
would
like
to
ask
is
I'm
not
sure
if
I
have
access
to
the
right
calendar,
etc,
etc.
Just
help
there
to
set
this
up
properly
and
if
needed,
I
cannot
any
PRS
to
the
site
or
whatever
is
needed,
but
especially
for
the
calendar,
and
you
know
all
these
social
media
things
I
would
definitely
use,
can
use
something.
D
C
C
G
C
D
Awesome
so
I
was
sort
of
highlighting
a
little
bit
some
of
the
stuff
in
progress
and
some
of
the
specs
that
people
are
putting
together
as
part
of
this
meeting,
to
get
more
people
looking
at
them,
because
we
have
a
lot
of
exciting
things
happening
and
we
also
have
a
whole
bunch
of
PRS
and
going
on
at
issues
going
on.
So
it
can
be
a
little
bit
difficult
if
you're,
just
trying
to
do
a
linear
scan
to
figure
out
where
all
the
pieces
are
and
what
interesting
things
are
going
on.
D
So
I
wanted
to
go
ahead
and
take
a
few
minutes.
If
that's
okay
and
sort
of
highlight
a
few
of
the
things
that
I
know
are
happening
and
give
folks
the
opportunity
to
highlight
things
that
they
may
know
that
are
happening,
that
I
have
missed
a
linear
scan
of
the
issues
and
PRS
and
then
to
point
a
little
bit
at
some
of
the
specs
that
are
floating
out
there.
That
could
definitely
use
additional
comments.
D
That's
some
good
to
folks
yeah
so
and
I
wanted
to
say,
give
the
folks
who
are
working
on
the
stuff,
an
opportunity
if
they
would
like
to
speak
up
a
bit
about
the
stuff
they're
working
on
now.
I
know
not.
Everyone
is
super
comfortable
speaking
up
in
meeting.
So
it's
perfectly
fine
to
this
sort
of
demure
and
I
all
hum
a
few
bars,
but
I
did
want
to
get
folks
the
opportunity
to
talk
a
little
bit
about.
D
You
know
what
they're
actively
working
on
so
I,
the
first
one
up,
DNS
I,
think
Denise
is
working
on
that.
Do
you
want
to
say
a
few
things?
Today's.
D
Yeah
I
mean
so
I'll
go
ahead
and
hum
a
few
bars.
Then,
if
he's
not
wanting
to
speak
up
or
having
trouble
with
meeting
the
basic
idea
behind
the
DNS
stuff,
is
that
one
of
the
things
we're
going
to
have
to
be
able
to
do
is
if
you
start
a
pond
and
you
connect
to
a
network
service.
D
So
it's
a
really
small
thing
that
can
fan
the
DNS
out
to
normal
kubernetes
DNS,
but
also
to
different
network
services
that
have
indicated
that
they're
hefty
in
us
is
to
be
queried
as
well.
And
so,
if
you
can
imagine
a
query
goes
out,
it
gets
fanned
out
to
all
the
possible
folks
and
then
the
first
response
that
comes
back.
That
is
not
a
negative
response,
gets
returned
to
the
workload
and
so
Denise
has
been
working
on
that
bit
by
bit
and
I
think
he's
got.
D
D
I
mean
effectively
I
sort
of
set
a
few
things
going
by
about
the
idea
behind
it,
which
is
sort
of
multiplex
DNS,
but
the
net
net
result
from
a
user
point
of
view
is
that
if
I
connect
to
a
network
service
like
a
VPN
where
I
say
I've
got
DNS
inside
my
corporate
Internet,
then
it
will
result
in
me
getting
proper
DNS
server
from
my
corporate
intranet.
In
addition
to
getting
it
from
kubernetes
or
at
least
I've
stated,
the
attention.
D
True
true
I
mean
this
was
really
quite
excellent,
work
that
was
done
by
by
Denis
and
and
he's
trying
to
break
it
up,
I
know
and
just
sort
of
smaller
manageable
pieces.
So
I,
don't
think
you
know
all
is
fully
turned
on
with
this
blue
dispatch.
Lance
I
think
he's
breaking
it
into
sort
of
more
manageable,
more
reviewable
pieces
this
week.
C
One
thing
that
I
wanted
to
add
here
is
not
during
the
discussions
that
we
we
attend,
fred-head
con
Europe
with
some
members,
but
netis
teams
that
actually
this
could
be
very
useful
for,
like
as
an
experience
for
giving
a
sidecar
D&S
could
be
useful
for
other
projects
and
other
initiatives,
so
to
definitely
give
much
larger
impact
in
that.
But,
okay,
we're
doing
it
for
ourselves,
primarily
yep
yeah.
B
D
I
know
I
could
see
all
kinds
of
utility
for
it.
I
mean
it's.
It's
really
interesting.
One
of
the
things
that
sort
of
thematic
lis
coming
out
of
the
work
that
we're
doing
here
is
taking
everything
down
from
a
cluster
granularity
to
a
workload.
Granularity
I
think
this
is
just
an
extension
of
the
same
thought.
So
you
know
it's
actually
quite
funny
and
talking
to
some
of
the
more
senior
communities,
people
like
the
very
first
thing
they
brought
up
when
I
first
talked
them
about
it.
What
what
are
you
doing
for
DNS?
D
D
D
I
think
if
you
go
look
you're
looking
at
the
original
document-
and
there
was
a
second
one
that
was
done,
but
the
a
lot
of
these
things
takes
a
lot
of
code
to
actually
figure
out
what
you're
doing
right,
because
there's
a
lot
of
innovation
involved
and
so
for
a
bunch
of
them.
But
I
know
that
Ilya
did
this
for
security.
D
He
went
and
sort
of
did
the
whole
thing
soup-to-nuts
and
then
you
know
you
look
back
at
what
you've
done
and
you
realize
that
you've
just
written
a
really
really
big
patch
and
so
he's
been
kind
enough
to
try
and
break
it
into
pieces
to
make
it
easier
to
review.
And
so
what
are
those
yeah
you're
on
the
third
piece
now
Ilya.
D
Is
what
being
kind
of
cool?
Because
we
are
using
sort
of
industry
standards,
50,
Speyer
kinds
of
approaches
for
the
most
part
for
this,
but
one
of
the
things
were
actually
doing.
That's
kind
of
cool
is
around
provenance
because
you
can
end
up
passing
through
a
bunch
of
network
service
managers
and,
if
you're
trying
to
decide
whether
or
not
something
should
be
admitted,
it's
not
enough
to
know
do
I
trust
the
client,
because
the
in
the
intermediate
you
know,
network
service
manager
is
kind
of
changed.
D
C
So
I
set
of
questions
regarding
okay
documentation
is
one
thing,
but
then,
if
you
can
at
least
share
with
us
now
your
experience
like
have
you
tried
packet
cluster
to
Google
Cloud
Cult
cluster,
or
have
you
tried,
you
know
different
different
clouds?
What
your
overall
experience
with
setting
this
up
is
it
complex
to
set
it
up
like
just
more
for
general
description
about
your
experiences?
We.
I
C
F
I
K
C
I
D
D
The
other
thing
that's
going
to
be
interesting
with
it
is
you
know,
sort
of
it's
good
to
be
interesting,
seeing
the
kinds
of
interesting
things
that
people
try
and
do
with
it.
So
there's
a
lot
of
good
room
for
experimentation,
and
it's
super
exciting
and
I'm,
hoping
to
see
folks
playing
with
it
more
illicit
lands,
cool.
B
D
D
D
No,
it
absolutely
is
it's
all
kinds
of
exciting
and
we
should
definitely
look
at
how
we
want
to
go
about
promoting
at
it,
and
that
was
actually
one
of
the
reasons
I
wanted
to
talk
about
the
things
in
progress
right
now,
because
there
are
so
many
exciting
things
happening
by
if
you're.
Just
you
know,
if
you
aren't
real
close
to
the
ground
on
the
PR,
as
you
might
have
missed
them
going
by
so
on
the
share
I'm.
D
C
D
A
Here
is
all
clusters
specific
logic
from
an
SM,
so
we
want
to
get
rid
of
this
hard
dependency
between
kubernetes
sidecar
and
in
the
SEM
right
now.
The
kubernetes
sidecar
is
started
on
5,000
port
and
an
assumption
know
about
this
particular
port
and
we
want
kubernetes
sidecar
just
register
in
a
some.
So
there
is
no
this
hard
lane
and
also
we
want
all
classic
specific
stuff
be
in
this
set
car,
and
if
you
want
to
move
for
some
reason
from
kubernetes,
we
just
need
to
replace
this
point
and
implement
another
plug-in.
A
D
For
those
of
you
guys
for
a
little
less
familiar
one
of
the
things
that
work
service
mesh
that
is
particularly
within
kubernetes,
is
it
will
avoid
colliding
with
the
prefixes
that
are
in
use
in
the
cluster
and
so
breaking
that
out
into
sort
of
a
a
plug-in
and
I.
Think
correct
me.
If
I'm
wrong
Victoria,
when
you
say
plug-in,
you
mean
things
sort
of
in
the
same
spirit
as
what's
done
with
CRI
or
CSI
and
kubernetes,
where
things
register
themselves
with
G
RPC
and
in
the
system
in
metric
service
manager,
interacts
with
them
to
charity.
A
D
Yeah,
so
not
a
plug-in
in
the
traditional
sense,
because
Co
doesn't
usually
work
that
way,
but
more
on
the
kubernetes
sense
and-
and
hopefully
this
will
make
network
service
manager
a
lot
simpler
and
make
the
entire
system
a
lot
more
flexible
and
modular
so
and
I
know.
We've
had
several
people
speak
up
for
flexible
and
modular
lately,
so
good.
D
D
D
The
other
thing
that
actually
came
up
did
this
came
up
around
some
minutiae.
Around
Vienna
handling
for
the
inner
domain
stuff
was
I.
Think
Nicolai
observed
quite
correctly
that
we
probably
want
to
figure
out
a
way
to
move
some
of
the
things
like
the
NI
selection
down
into
the
NSM
data
plane,
but
I
think
we're
now
calling
the
NSF
order,
which
you
know
rather
than
having
that
done
in
the
MSM
manager,
because
it
makes
the
system
much
more
flexible
and
pluggable.
D
We
probably
gonna
have
to
make
some
small
changes
to
the
data
playing
API
to
do
that,
but
we'll
try
and
keep
that
to
a
minimum,
but
that
there's
that
issue
literally,
is
you
know,
fourteen
eleven,
that's
literally
a
conversation
that
started
this
morning,
and
so
definitely,
if
you're
interested
in
how
that
works.
Please
jump
down
on
that
issue.
D
J
C
C
J
D
We
it
shouldn't,
be
underestimated.
How
awesome
this
is
a
trace
of
work,
because
the
kernel
forwarding
plane
stuff
is
both
useful
and
excellent
foil
for
going
through
and
shaking
out
the
issues
with
supporting
multiple
data
points
and
multiple
data
planes,
particularly
multiple
simultaneous
data
planes
is
going
to
be
hugely
important,
as
we
start
moving
to
handling
critics
and-
and
it's
sort
of
literally
the
simplest
possible
thing
you
could
have
done.
That
is
interesting
that
actually
shakes
out
all
these
problems
and
gets
them
fixed.
So
it's
a
really
good
choice
of
thing
to
work
on.
C
D
So
cool
the
do
other
folks
have
things
that
they
have
in
progress.
They
wanted
to
race
here.
I
mean,
as
always,
feel
free
to
add
your
stuff
to
the
agenda
as
we
go,
but
I
I
do
want
to
make
sure
that
if
we've
got
other
people
who
are
working
on
interesting
chunks
of
things,
oh
I
actually
I
just
remembered
one
that
I
should
add.
I've
been
puttering
away
at
some
STK
stuff.
D
L
L
D
That's
super
cool
because
we
have
had
a
lot
of
people
think
various
things
about
how
to
add
visibility
into
network
service.
Mash
we've
got
some
metric
stuff
that
was
worked
by
them.
It's
you,
and
we've
also
got
people
who
have
used
a
bit
about
using
IOM
headers
and
that
kind
of
stuff,
but
more
stuff
to
learn
from
would
be
excellent.
C
Yeah,
we
also
have
a
link
on
the
SMI,
which
is
more
or
less
also
at
least
full
for
the
first
attempt
we're
mostly
looking
at
she's
looking
at
the
metric
apart
bit.
So
that's
I
mean
kind
of
in
the
same
direction
of
trying
to
be
able
to
get
more
information
and
probably
presented
itself
format
to
the
outside
world.
D
D
B
C
We're
doing
some
similar
for
the
CNF
test,
but
with
Taylor
and
Michael
at
least
we
tend
to,
but
it's
not
as
formal.
So
whatever
you
have
proposed,
like
it's
more
or
less
manual,
and
this
is
not
yet
for
the
time
being,
but
at
least
maybe
we
will
help
pave
the
way
for,
but
I
know
that
there
are
various
ideas
here
like
and
although
that
he
has
proposed
something
we
can
consume
the
external
hardware
as
a
service.
C
D
Of
the
thing
you
might
want
to
take
a
look
at
Matthew
and
the
stuff
you
do
it.
You're,
looking
at
for
gateway
stuff,
is
go.
Take
a
look
at
the
other
domain
spec
because
I
believe
architectural
space
in
there
for
what
you
were
talking
about
doing
with
the
gateway
stuff,
in
the
sense
that
you
can
have
proxy
network
service
managers
and
proxy
network
service
registrar's.
That
can
link
up
with
whatever
you're
doing
for
gateways
in
terms
of
data
play.
Does
that
make
sense?
Okay,.
K
D
D
Cool
and
so
specs
to
review
we've
got
quite
a
few
specs
that
are
sort
of
running
around
right
now.
That
I
think
folks
would
like
to
you
know.
In
addition,
due
to
the
ones
that
are
actually
somewhat
in
progress,
where
you
can
input
is
welcome.
What
are
the
ones
running
loose
was
there's
a
general
desire
to
right
now.
H
Also
at
I
suppose
form
in
my
case.
We
could
have
moved
with
logic
to,
for
example,
EPP
pays
it,
but
the
plane
only
so
it
was
also
possible
and
security.
We
could
have
just
one
similar
to
kubernetes
a
unique
socket
to
connect
to
an
s
manager
and
if
the
client
need
ma'am
I
have
a
request
from
the
data
plane
workspace.
H
D
C
C
F
The
spec
zone
is
not
up
to
date
with
the
latest
thing,
but
I'm
currently
on
the
matrix
part
in
the
prometheus
integration.
We
had
some
discussion
with
rather
often
requires
sharing
the
approach
I
taught
on
and
what
we
agreed
on
that
we'll
have
well
representative
matrixes
to
match
them
with
the
cloud.
Namespace
is
its
unique
so
that
you
can
make
curious
in
prometheus
and
it
is,
it
will
be
well
integrated
with
the
SMI
requirements
and
for
the
other
part
of
this
will
need
to
integrate
in
the
community.
There.
F
I'm
still
by
I,
see
some
movement
there
still
waiting
for
them
to
announce
some
weekly
meetings,
so
that
for
this
pecs
knee,
we
need
to
what
network
service
route
because
they
are
working
with
HTTP
only
for
the
moment.
But
this
is
not
the
first
focus.
We
are
first
focusing
on
the
met.
Expert
I
will
I'm
going
to
update
these,
because
I
didn't
update
the
spec
very
much.
D
This
is
for
something
to
look
at
with
some
of
the
plugins
stuff
as
well
that
we
discussed
earlier
because
there's
effectively
when
you
should
get
down
to
it.
What
this
guy
is
actually
looking
at
wanting
to
do
is
to
be
able
to
plug
in
a
different
selection
criteria
among
candidates
other
than
round-robin
right,
so
Rob
Robin
is
a
is
basically
wonderful
in
that
you
can
do
it
really
easily,
but
there
are
lots
of
more
sophisticated
things.
K
D
D
C
Examples,
repo
site,
we
have
more
and
more
examples.
There
and
I
have
recently
added
number
of
improvements
just
to
make
life
of
the
people
that
are
using
examples
there.
So
we
can
list
the
examples
with
some
simple
descriptions.
You
can
browse
browse
that
you
can
record
description
which
essentially
shows
specific
examples.
Readme.
J
C
J
C
C
D
D
J
Hi,
this
is
Taylor
and
I.
Think
we'll
probably
have
other
options
in
the
test
bed
specifically
because
other
things
are
being
worked
on
and
anything
that
someone
wants
to
contribute.
I
I
do
believe
that
most
of
the
upcoming
use
cases
will
be
connected
for
all
the
service
chains
and
driven
within
us
and
so
I
think
it's
going
to
be
a
core
part,
even
after
other
test
cases
that
may
use
other
options.
I.