►
From YouTube: CNCF Network Service Mesh Meeting - 2019-08-27
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
Okay,
let's
go
ahead
and
get
started
so
first
off.
Please
add
yourself
to
the
attendees
list
list
on
the
on
the
agenda,
and
so
we
have
a
few
recurring
colds
that
are
that
are
on
so
we
have
the
MSN
documents
call
we
have
aa
moment.
We
have
a,
we
have
a
use
case
call
and
also-
and
also
on
hiatus.
We
have
a
document
call,
so
we
will
resume
those
heads
as
necessary.
A
A
B
A
So,
in
that
scenario,
4:00
a.m.
Pacific
time
for
September
16th
all
right,
so
we
have
a
few
major
events
coming
up.
We
have
ons
Europe
in
Antwerp,
where
we
have
for
talks
accepted
there
is
a
telecom
user
group
meet
up
and
the
CNF
test
bed
tutorial.
We
have
the
open-source
summit
coming
up
which,
where
they
talk
accepted
on
on
the
introduction
of
NSM
by
Ivana
and
Robin
South,
we
have
Q
Khan,
incarnated
Khan
coming
up
still
is
still
being
decided
upon.
A
We
have,
but
we
do
have
MSM
Khan,
and
so
we
definitely
request
everyone
here
joining
us
at
NSF
Khan.
If
you
should
be
in
the
area
and
simultaneously
we're
also
looking
for
for
proposals
for
for
talks
at
the
at
the
MSM
con
as
well.
So
if
you
have
something
you
would
you
wouldn't
love
to
talk
about
that
is
NSM
related,
definitely,
definitely
bring
it
up
here.
A
B
Followed
15
more
accounts
and
we
did
Anna
retweeted
25
times
this
week
to
follow
up
from
last
the
VMware
open
source
teams,
part
two
of
mess
of
mess
meshes
breaking
down
service
mesh
ecosystem,
including
network
service
mesh,
was
reposted
from
168
different
accounts
on
Twitter
and
that's
kind
of
slowed
down,
but
last
Tuesday
through
I
think
last
Thursday
160
reposts
of
that
blog
post.
So
congratulations
Nicolai
for
your
successful
post
people
really
liked
it.
I
reposted
it
twice
from
our
account.
B
I
could
have
reposted
160
times,
but
that
would
be
spammy
so
I,
just
posted
it
twice
and
I
also
posted
a
reminder
of
today's
working
group.
The
call
for
sponsors
for
network
service
mesh
con
@q
con
North,
America
and
I
pinned
a
new
tweet
for
network
service
mesh
con
on
to
the
top
of
the
Twitter
profile
page,
and
this
week's
plan
I'll
send
a
24
hour
reminder
for
bi-weekly
networks
or
smash
call
on
Tuesday
the
September
3rd.
B
The
earlier
time
now
continue
announcing
the
open
networking
summit,
open
source
summit
and
cube
con
talks,
as
well
as
searching
for
any
mentions
and
reposting
I'll
announce
the
next
release.
When
available-
and
it
did
have
some
questions
this
week
and
service
mesh-
was
tagged
in
a
post
and
I
read
through
their
github
repo
to
see
if
they
had
Forks
or
mentioned
network
service
smash,
as
they
tagged
and
thanks
network
server
smash
in
their
products
release.
B
B
I
believe
that
nonce
and
CF
posted
projects
must
have
300
stars
on
their
github
repo
in
order
to
qualify
to
be
displayed,
and
so
they
were
looking
for
more
stars
and
that
just
led
a
question
we
have
two
hundred
and
fifty
four
stars
in
our
repo.
Do
we
have
any
goals
to
reach
any
number
of
stars
by
any
certain
day?
If
so,
I
can
request
otherwise
may
not
request.
A
Yeah,
that's
that's
gonna,
be
one
of
the
interesting
things
about
it,
because
in
the
long
run,
I
suspect
that
NSM
is
going
to
be
more
like
I
guess
you
would
say
like
the
thing
that
helps
things
run,
but
not
necessarily
the
thing
on
the
front
and
so
like.
Like
you
know,
if
you
go
see
a
concert,
everyone,
everyone
always
cheers
for
the
lead
front
guy,
but
they
never
look
at
the
sound
guy
behind
them.
B
B
A
I
think
that'd
be
a
great
idea
and
just
that
the
we
want
to.
We
want
to
be
at
a
point
where,
like
right
now,
you
can
download
and
you
can
run
an
MSM.
These
steps
to
do
it
are
still
non.
Arcs
are
still
not
as
trivial
as
we
want,
but
we're
getting
down
to
the
point
where
you
can
run
home
install
and
it
just
it
just
installs
it
and
so
a
little
bit
more
work
and
I.
A
Think
I
can
talk
to
this,
but
once
we're
like
once
we're
getting
close
to
a
release,
candidate
or
production,
we
get
people
to
to
start
to
start
playing
around
with
it
and
I
think
that'd
be
the
perfect
time,
and
that
way
the
people
win
their
first
interaction
with
it
is,
if
they
decide
to
try,
it
is
hey.
This
is
so
rather
than
hey,
this
looks
cool,
but
this
doesn't
work
and-
and
so
so
once
we're
close
to
production,
I
think
that'd
be
a
fantastic
time
to
do
that.
B
Absolutely
are
there
any
other
shareable
milestones
with
that
stars.
Waiting
on
hold
we've
got
the
1000
commits,
which
we
had
passed
so
next
will
be
two
thousand
commits,
we're
almost
to
four
hundred
followers
on
Twitter
I
think
we
need
fewer
than
ten.
Are
there
any
other
milestones
that
you'd
like
to
try
to
reach
or
share
once
we
do
reach.
C
A
A
Well
so
Ana's
not
here
today,
because
he's
not
feeling
particularly
well
so
he's
taking
that
short
break
but
I,
but
I
think
we
can
do
as
well
is
so
depending
on
who
we
have
on
the
call
we
can.
We
can
ask
for
a
few
updates,
and
so
our
laps
were
so.
The
first
thing
is
circle.
Ci
has
had
some
some
instability.
Do
we
have
anyone
here
who
has
been
focusing
on
circle
CI
in
order
to
in
order
to
increase
the
stability.
D
C
A
A
Thank
you
so
on
the
security
aspect,
so
we're
looking
at
bringing
in
spiffy
inspire
so
for
those
who
are
unfamiliar
with
spiffy
and
and
fire.
So
spiffy
is
a
specification
of
how
to
establish
identity
of
four
codes.
Inspire
is
a
implementation
of
that.
If
you
have
used
this
do
to
do
workload
identity.
You
are
using
that
spiffy,
but
your
use
Citadel
instead
of
spire.
In
that
scenario,
the
reason
for
using
spire
is
that's
fireworks
without,
without
its
do
so,
Citadel
is
designed
to
work
specifically
only
and
only
in
this
geo
and
is
tightly
integrated.
A
So
in
terms
of
what
we're
looking
to
do
is
we're
looking
to
use
52
to
try
to
work
out
exactly
who
is
the
workload
on
the
other
side,
so
you
can
think
of
it
as
establishing
problems
like
establishing
provenance
establishing
a
connection
we
can
use,
because
you
get
tokens
on
each
side
that
are
graphically
verifiable.
You
can
check
to
see
you
know
who
who
is
the
next
person
in
the
chain
and
simultaneously
you
can
also
check,
as
each
of
the
tokens
gets
populated
as
you
connect
through
your
chain.
A
You
can
look
at
the
list
of
tokens
and
work
out.
Do
I
trust
everyone
in
the
in
the
chain
or
do
I,
not
trust
everyone
in
the
chain.
So
from
a
security
perspective
this
it's
not
the
only
thing
we
need
to
do,
but
it's,
but
this
certainly
is
a
excellent
first
building
block.
So
with
fad
we
have
the
security.
Spec
is.
D
A
D
D
C
Was
checking
the
testing
today
that
we
had
the
include
mean
and
lose
very
interesting
I'm
actually
inclined
to
somehow
have
some
example
in
Indiana
replay
example
straight
to
where
we
can
demonstrate
this
with
two
kind
deployments,
or
something
like
that,
but
yeah
I
have
to
figure
it
out.
I
mean
we
never
designed
for
multi
cluster.
C
A
D
Yeah
you'll
talk,
a
bit
have
murdered
few
requests,
and
few
is
some
in
some
progress
and
I.
Think
Denise
who's
working
on
genius
started
to
walk
over
them
and
will
provide
some
updates
in
few
days.
I
suppose
so
in
general
idea
is
to
have
plugins
to
be
able
to
do
all
kind
of
cluster
specific
stuff.
At
the
moment,
plugin
support,
exclude,
prefixes
and
upcoming
stuff
is
will
provide
support
for
finding
can
points,
and
some
parts
of
inter
domain
implementation
will
be
also
implemented.
D
It's
also
John
he's
not
here,
as
I
know,
he's
already
heavy.
We
seek
support
between
local
clusters
and
most
of
the
tests.
A
stable
and
right
now
he's
working
to
add
support
of
series
6
between
inter
domains
and
I
hope
it
will
be
nurtured
and
not
noted,
but
poor
quest
will
be
ready
in
a
few
days,
so
you'll
be
able
to
review
for
36
supports
also
for
36.
A
C
Another
troll
stuff
is
working
on
the
Inc
matrix
support
and
we
have
been
discussing
with
it
about
next
steps
and
all
that
under
that
some
people,
from
from
your
team,
is
going
to
also
join
this
effort
to
update
network
service
manager
to
service
major
in
the
data
plane
selection,
to
give
a
better
support
for
multiple
data
planes.
Because
today
it's
a
bit
opportunistic,
like
whatever
you
find.
First
to
just
select
it
and
call
yeah.
C
Yeah
so,
depending
on
the
older
stuff
they're
registered,
you
can
get
different,
essentially
scenarios,
but
I
mean
it
works
for
whatever
it
was
designed.
We
have
to
move
forward
and
and
already
had
been
some
discussions
around
this
some
people
from
your
team.
So
we
have
to
figure
out
the
spec
and
more
more
concrete
implementation
in
mind
and
in
forward.
A
Yeah,
so
one
of
the
nice
things
on
this
as
well
is
it
definitely
helps
us
work
out
if
any
VPP
isms
have
crept
into
the
into
the
codebase.
So
definitely
definitely
looking
forward
to
that
this.
This
also
helps
me
out
in
a
couple
other
efforts,
because
I'm
looking
at
I'm
currently
engaging
a
couple.
Other
companies
who
who
have
data
planes
and
it's
making
sure
that
this
that
this
effort
is
smooth,
is
critical
to
to
engaging
them.
D
Okay,
in
few
words
at
transfer
to
its
work
to
me,
so
at
the
moment
we
have
a
pull
request,
which
will
reverse
logic
in
ways
DK
when
we
define
a
chain
of
components,
I'm
working
to
stabilize
it
and
fix
link
linking
issues
during
his
work.
I've
found
few
issues
with
the
open
tracing
will
be
separate,
requests
and
found
very
a
bit
of
weird
issue
with
our
VPP
example
in
case
of
yours,
with
fall
into
infinite,
recursion
and
step
by
timeout.
D
So
it
also
will
be
a
separate
request
for
a
bit
of
fixes
for
this.
So
I
think
it's
in
few
days.
It
will
be
ready,
including
examples.
Repo
changes
so
mostly
also
in
step
right,
paw,
request,
open
tracing
will
be
much
improved
and
it
will
be
possible
to
look
for
all
chain
of
operations
perform
it
starting
from
client
to
all
offer
endpoints
we
go
across
so.
C
C
D
C
C
C
C
D
E
D
A
Yeah
for
those
who
are
unfamiliar
with
this
one
of
the
nice
side,
effects
of
this
patch
is
that
it
simplifies
the
path
for
adding
multiple
data
plane.
So
if
you
want
access
to
a
special
NIC
card-
or
you
have
maybe
you're
using
VPP
for
certain
certain
act
actions,
but
you
want
to
use
something
something
else
for
performing
a
different
action
based
upon
the
need,
then
this
this
helps
out
tremendously
because
it
gets
rewritten
on
the
fly
to
the
to
the
correct
order.
So
so
this
definitely
this
definitely
helps
out
tremendously
in
that
space.
A
C
C
C
C
At
some
point,
should
we
be
able
to
actually
expose
the
the
networking
resources,
the
high
speed
network,
for
example?
We
have
some
50
gig
network
or
contradict
network,
and
you
expose
it
as
a
service.
Then
you
consume
it
this
way,
and
so
various
very
discussion
was
there,
but
one
of
the
ideas
that
mutt
actually
from
Mirage
I
started
this
a
couple
of
months
ago
was
this
English
gateway
and
all
that
he
is
working
and
I
think
that
he
is
on
the
call
today.
Matt
are
you
here?
No.
C
C
A
Just
yet
because
I
mean
I'm
still
thinking
for
a
lot
of
them
and
there's
a
couple
approaches
that
I
that
I've
thought
up
of
the
one
that
I
think
is
the
most
promising
is
I
was
thinking
of
when,
when
we
want
to
when
we
want
to
gain
interoperability
with
NSM,
what
we
can
do
is
we
leave
the
CMI
as
it
is,
but
what
we
can
do
is
once
the
interface
has
been
created,
CNI
as
a
process
exits,
and
so
that
gives
us
the
that
gives
us
the
interface.
That's
there.
So
what
I'm
thinking
of
is?
A
That
is
valid
only
only
for
that
specific
request,
so
that,
when
you
make
the
the
call
in
it
has
a
registered
and
I
may
space
that's
there,
and
it
would
only
move
things
on
on
the
man
there's
some
things
that
we
have
to
think
through
in
terms
of
making
in
terms
of
making
this
work.
This
has
implications
on
on
IP
table
rolls
I,
think
a
manned
up
working
out.
Okay,
because
the
IP
tables
will
still
be
there.
They'll
still
interact
on
on
that
specific
interface,
and
so,
but
we
we
should
still
be.
A
We
should
still
be
careful
with
it
in
terms
of
how,
in
terms
of
how
it
works,
there's
a
couple
other,
simpler
ideas
that
have
popped
up
as
well,
but
I
think
that
one
may
be.
It
may
be
the
simplest,
and
so
the
the
goal
at
the
end
of
the
day,
is
to
be
able
to
interact
with
an
existing
CNI
and
to
be
able
to
interpose
an
SM
between
the
network
namespace
and
the
actual
CNI
SDM
in
such
a
way
that
we
can
inject
network
services
in
them
in
the
middle.
A
So
just
some
just
some
spots
that
that
are
that
are
there
anyways
I'll
have
I'll,
have
a
few
paths
that'll
be
that
I'll
write
up
in
the
next
week
or
so
of
different
possible
approaches
and
I'm
going
to
ask
people
for
help
with
reviewing
this
and
in
more
detail,
because
this
is
something
that
there's
gonna
be
a
lot
of
edge.
Cases
on
and
I
actually
don't
think
we'll
be
able
to
solve
all
the
edge
cases
in
the
first
pass.
C
Yeah
I
mean
I
I
would
prefer
to
see
some
some
more
concrete
writing
and
maybe
one
pictures
but
yeah
it
don't
sound,
sounds
very
interesting
and
I'm
not
on
a
lot
of
interest
and,
as
we
said
a
lot
of
times,
maybe
it
brings
us
closer
to
make
any
same.
One
potential
help
like
our
one
potential
solution
to
the
kubernetes.
A
A
A
What
I'm
going
to
aim
for
is?
Is
we
pick
a
couple
of
the
the
most
commonly
used?
See
a
nice
to
start
with
so
I
think
we
can
pick
things
like.
That's,
let's
see,
let's
make
it
work
for
calico
and
in
flannel,
and
so
we
can
start
with.
You
know
start
simple,
but
then
we
we
can
build
up
from.
You
know
we
can
build
up
from
there
over
time.
A
A
D
F
F
C
A
A
A
C
A
Fantastic
well
with
fat
when
a
couple
a
couple
last
reminders:
NSM
call
for
proposals
and
number
two.
If
you're,
considering
sponsoring
NSM
con,
the
the
link
is
on
the
website
in
the
events
in
the
events
section
so,
and
so
with
that
I
want
to
thank
everyone
for
your
time,
and
we
will
see
you
all
again
next
week.
At
the
same
time,
you
all
have
a
good
day.
Thank.