►
From YouTube: Network Service Mesh WG Meeting - 2019-02-19
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
All
right
so
events
coming
up,
we
have
in
February
25th
to
29th.
We
have
Mobile
World
Congress
in
Barcelona.
This
tends
towards
demos
and
it's
more
Expo
focused
very
heavy
on
energy
use
cases,
and
so,
if
you
are
heading
over
there,
let
me
know
because
I
will
be
there
as
well.
We
can
meet
up.
We
have
serviced
mesh
day
which
is
going
to
be
in
San.
Francisco
I
am
going
to
submit
a
talk
to
this.
A
The
call
for
papers
is
going
to
close
on
March
first
I
swapped
it
around
for
future,
because
health
of
paper
is
more
important,
April
3rd
through
5th.
We
have
open
networking
summit,
oh
and
that's
in
North
America
in
San,
Jose,
so
again,
very
service,
very
envy.
Centric
again
the
Culver
papers
has
already
closed
and
the
schedule
has
been
published
as
of
this
week.
So
we
have
a
few
talks
on
network
service.
Mesh
that
have
been
accepted
in
the
talks.
I
believe
are
mainly
on
the
Wednesday
and
Thursday
I.
A
There
was
also
a
panel
discussion
that
was
accepted
in
that
I
believe
was
putted
together
by
romkey,
which
I,
don't
think
was
was
NSM
centric,
but
will
very
definitely
include
discussion
about
MSM
based
upon
the
people
who
are
in
that
we
have
in
April
9
through
12.
We
have
the
MPLS
and
Sdn
and
in
and
at
the
event.
So
if
you're
in
the
area
feel
free
to
attendance
in
April,
17th
through
19th
would
have
container
world
2019
with
a
talk
accepted
by
Prem
from
Lumina,
he
will
be
talking
on
network
service
mesh.
A
We
have
cube
con
EU.
The
call
for
papers
is
closed.
We
will
not
know
about
what's
what's
happening
on
that
until
around
March
5th,
the
recommendation
is
to
book
hotels.
Now
the
dates
are
May
21st
or
May
23rd.
There
will
also
be
co-located
events
at
Q,
Khan,
EU,
Fido,
Mini
summit
is
do
we
do
we
have
any
information
yet
on
Fido
final,
many
many
event
called
papers,
not.
A
A
I'm
guessing
that
in
this
particular
scenario,
and
if
anyone
if
anyone
is
attending,
let
it
snow
and
we
can
get
you
ready
if
you
want
to
talk
about
network
service
mature
or
rather
discuss
with
people.
Finally,
we
have
ons
Europe
in
September
23rd
to
25th
coming
up.
The
call
for
paper
window
is
currently
pending.
So
once
we
get
more
information
on
that,
we'll
publish
it
announcements,
we
have
Circle
C,
iirc
announcements
that
were
disabled,
so
we
found
that
the
IRC,
an
announcements
were
we're
a
bit
flaky.
A
Sometimes
it
would
send
a
message
sometime,
they
wouldn't
Nikolay
has
disabled.
It
is.
Are
there
any
other
announcements
that
I
should?
Actually
there
is
an
announce
my
should
do.
We
also
have
a
new
documentation
group
where
we
are
documenting
and
and
solidifying
the
the
architecture
and
wording
and
so
on
around
it
in
in
our
Wednesday
meeting.
So
it's
at
the
same
time
as
this
meeting
just
tomorrow
and
that
is
being
read
by
our
veg
being
led
by
by
Jeffrey
from
charter.
So
definitely
it's
it.
A
Also,
don't
forget
to
star
the
project:
if
we
can
manage
to
get
300,
then
we
are
able
to
to
add
ourselves
to
the
cloud
native
compute
foundations,
landscape.
Being
a
part
of
their
landscape
would
be
a
good
thing
because
we'll
be
able
to
point
to
it
and
get
them
and
get
other
people
to
or
increase
the
number
of
people
who
look
at
our
project.
A
D
D
D
D
We
also
are
working
hard
on
identifying
the
potential
issues
with
the
data
plane.
Instability
that
we
have
identified
in
several
kind
of
specific
scenarios
at
and
step
on,
are
kind
of
working
on.
This
I
know
that
not
we
have
the
very
first
debug
output
already
sent
to
the
dingus,
which
is
which
is
a
good
start.
I
guess
I
hope
that
we
have
some
automation
around
this
being
able
to
generate.
D
Generalities,
the
poke
out,
puts
in
case
of
instabilities
and
be
able
to
submit
them
so
that
for
faster
I,
say
the
resolution
resolving
of
the
of
the
data
plane
related
problems,
and
so
with
this
we
also
today
like
as
of
yesterday
at
least
European
time.
Yesterday
we
had,
we
had
merged
a
couple
of
peers
which
made
master
failing
CI.
D
There
was
some
short
discussion
going
common
I'll
see
if
someone
was
following,
but
essentially
it's
around
the
lines,
how
we
ensure
that
master
is
never
failing
with
the
CIA,
and
the
fact
today
is
that
you
are
not
I
mean
after
after
merging
consultant
PR.
We
essentially
have
to
enforce
the
other
pending
peers,
to
merge
master
or
to
the
base
unit.
Okay,
actually,
merge
is
a
better
practice
to
just
in
kind
of
rerun
the
tests
and
verify
that
that
the
the
second
PR
to
be
merged-
it's
not
failing
after
the
first
they
are
being
merged.
D
So
this
is
something
that
I'm
working
on
in
one
of
the
in
one
of
the
peers
here,
which
should
be
in
the
backlog,
I
believe
750.
No,
it's
not
here,
but
it's
something.
That's
that's!
On
the
table,
I
think
that
that
we
are
having
a
good
pace
for
the
release.
For
now.
One
thing
that
title
is
worth
maybe
maybe
discussing
is
that
maybe
some
time
middle
next
month,
like
in
middle
of
March,
we
should
try
to
to
focus
more
on.
D
You
know,
testing
and
stability
instead
of
you
know,
merging
huge,
huge,
huge
features
in
master,
but
my
guess
this
is
also
part
of
the
of
the
discussion
that
we
hit
around
the
spec
for
the
release
process,
but
yeah.
That's
him!
That's
Morris
in
the
statement
today,
I
don't
know
if
there
are
any
questions,
any
suggestions
on
how
we
can
drive
this
better
possession,
my
approach,
because
I
am
I'm
driving.
D
D
D
B
E
E
B
D
B
I
think
probably
we
may
have
to
paralyze
some
things.
And/Or
I
know
right
now
our
tests
are
currently
standing
up.
All
of
the
infrastructure
is
a
fixture
and
then
tearing
it
all
down
and-
and
there
may
be
an
interesting
question
as
to
whether
some
tests
can
be
run
in
parallel
on
the
same
infrastructure,
I,
don't
know,
and
then
whether
we
need
to
simply
more
infrastructure
and
run
testing
hurtle.
That
way,.
B
Around
they're
probably
zombies,
oh
yeah,
they
should
be
after
the
test
about
they're,
probably
zombies,
but
I.
Think
the
underlying
point,
which
is,
could
we
run
two
of
them
two
clusters
in
packet?
Yeah
we
serve
that
good
yeah.
Would
it
possibly
make
sense
to
explore
keeping
some
number
running
in
packet
all
the
time
and
then
allocating
them,
and
then
it's
already
allocated
and
you
come
in-
you
need
one
or
allocating
them
and
then
creating
a
new
set
when
you've
allocated
them.
B
D
D
B
F
Yeah,
so
basically,
my
idea
was
to
to
somehow
consolidate
this
and
to
make
it
make
it
available
for
future
reference,
because
it's
a
bit
troublesome
to
to
keep
track
of
all
the
google
docs
that
are
there
and
yeah
I'm
I
have
created
a
template
that
can
be
used.
So
it's
easier
to
to
fill
such
a
spec
file
and
publish
it
along
with
your
PR.
B
So
hopefully,
that
gives
us
a
good
way
to
capture
these.
Some
of
these
faxes
were
going
to
come
later,
because
someone
asked
the
questions
like
okay.
Well,
what
are
we
doing
about
security?
Well,
okay,
great,
there's,
the
security's
back
and
it's
living
with
the
code,
and
so
while
it
has,
it
has
a
somewhat
better
chance
of
actually
being
up
to
date,
then
something
that
is
not
living
with
the
code.
B
B
You
know
I,
think
Nikolai.
We
can
probably
capture
the
spec.
You
have
for
the
release
process.
At
this
point,
or
at
least
the
dates
should
probably
be
captured.
We've
got
something
on
the
sessions
payload
type,
which
I
put
together
to
sort
of
relate
to
the
Envoy
as
a
network
service,
and
then
we've
got
some
things.
Looking
at
proxy
network
service
managers,
physical
NICs
and
inter-domain,
and
a
sentiment
I
think
those
are
have
some
lively
discussion
going
around
them
right
now.
B
D
B
B
B
D
B
B
B
B
The
rest
of
the
world
and
I
mean
that
that's
sort
of
the
thing
that
I
think
you
could
take
or
leave
I
know
the
opendaylight
guy
is
because
of
this
decided
to
name
a
thing
made
their
releases
in
the
order
of
the
elements,
because
you
know
we
all
agree
on
the
order
of
the
elements
and
I'm
not
suggesting
that
here.
But
you
know
so
me
we
could
do
the
alphabetical
thing
or
we
could
just
pick
names
that
we,
like.
Oh.
B
B
F
B
A
Just
one
area
that
people
can
look
at
is
if
we
end
up
using
things
that
are
standard
for
the
control
plane,
so
we
have
a
couple
areas
in
security
of
control
and
and
and
data
plane
and
so
on.
But
if
you
look
at
just
even
gaining
access
to
the
MSM
component
in
itself,
there's
there
are
a
few
tools
that
we
could
use
within
kubernetes,
mixed
in
with
kubernetes
secrets
and
and
rotating
keys,
and
so
on.
A
That
might
be
really
really
nice
to
to
add
in
so
and
then
we
can
tie
them
into
GRP
see
through
the
authentication
component,
so
that
would
give
us
the
ability
to
revoke
to
revoke
keys
the
ability
to.
If
someone
were
to
grab
one,
it
would
not
be
able
to
use
it
and
we
can,
even
depending
on
the
type
of
keys
that
we
use.
If
we,
if
we
give
them
names,
then
we
could
also
verify
that
them
that
the
original
comes
from
the
comes
from
the
name.
That
is,
that
is
on
the
key.
A
If
it's
going
through
a
socket
to
the
MSM,
B
and,
of
course,
intent
or
inter
domain,
is
this
whole
thing
as
well?
There's
a
lot
of
things
that
we
can
do
even
just
to
establish
authentic,
to
authenticate
and
authorize
the
user
at
the
beginning,
like
there's
a
lot
of
stuff
there
that
that
needs
to
be
looked
at.
So
if
anyone
has
an
interest
in
this
kind
of
stuff,
like
definitely
step
up,
there's
a
huge
amount
of
stuff
here.
B
A
D
H
G
So
yeah
we
had
our
first
meeting
last
week.
You
know
there
was
a
lot
of
people
on
that
call
that
are
on
this
call
so
I'll.
Let
them
speak
if
they
thought
it
was
worthwhile.
From
my
own
perspective,
I
thought
that
it
accomplished
what
I
was
ultimately
hoping,
and
that
was
pointing
out
that
maybe
some
of
these
terms
aren't
as
clearly
defined
as
everybody
thinks
they
are,
or
it
might
be
clearly
defined,
but
that
knowledge
is
trapped
in
one
individual's
head.
D
D
G
Stuff
will
have
to
survive
the
Google
Docs
lifecycle
before
it
moves
down
to
the
repo,
so
yeah
go
ahead
and
click
this
so
I
mean
even
with
Fredrik
and
Nicolai
on
the
call
I
think
they
would
agree
that
you
know
there's
application
people
who
have
application
type
definitions
floating
in
their
head.
G
There's
networking
people
who
have
Network
definitions
floating
in
their
head
and
I
think
there
was
a
decent
amount
of
collaboration
on
trying
to
figure
out
what
the
NSM
definition
is
for
a
lot
of
these
things,
we've
tried
to
keep
this,
as
you
can
see
like.
When
did
you
sentences
per
like
this?
This
isn't
meant
to
be
like
the
deep
dive
document,
because
then
you
know
you
could
probably
write
a
small
document
just
on
network
service
endpoint
but
I'm,
giving
a
clear
definition
of
what
this
is.
G
So
that
way
when
people
come
in
and
look
at
the
documentation
of
this
and
that
we
can
link
this
glossary
to
everything
and
people
in
a
single
sitting,
you
can
know
exactly
what
a
network
service
client
is
versus
what
a
network
service
in
point
is.
You
can
see
in
the
comments
there
was
debate
during
the
call
we
went
back
and
forth
I
think
we
only
got
through
about
one
third
of
it,
so
we're
going
to
continue
to
tackle
this
tomorrow.
G
B
B
G
Yeah
I
think,
if
you
just
give
it
a
glance
over
and
kind
of
when
you
come
in
in
the
back
of
your
mind,
kind
of
be
ready
to
talk
about
like
what
the
data
plane
is
from
the
innocent
perspective
of
what
the
control
plane
is.
You
know
there's
still
some
fuzziness
around,
like
just
the
nebulous
concept
of
what
a
network
service
is
right
and
I
mean
this
is
once
again
what
I
like
I've
got,
what
I
think
it
is
from
a
network.
G
You
know
standpoint
and
then
the
application
people
have
like
what
they
kind
of
think
it
is
so
I'm
I
think
you
know.
Once
we
get
this
glossary
done,
then
we
can
start
picking
out
individual
buckets
out
of
this
and
start
diving
in
even
deeper
starting
to
you
know,
go
into
how
the
code
interacts.
You
know,
with
what
the
concept
is.
I've
been
constantly
going
back
to
the
specs
pages
and
trying
to
make
sure
that
this
document
aligns
with
what
we're
calling
out
in
the
specs.
G
There
are
a
few
things
where
maybe
some
of
the
specs
reuse
terms-
and
you
know
that's
tried
to
be
called
out
here
so
yeah
I
mean
this
is
not
the
sexy
work.
This
isn't
like
new
feature
releases,
but
as
far
as
getting
like
wide
adoption
getting
into
CNCs
things
like
that,
I
think
this
is
ultimately
where
we'll
make
our
money
so.
A
In
order
to,
in
order
to
like,
like
what
level
of
detail,
we
were
probably
go
into
a
bit
more
detail
than
they
would
because
we're
describing
a
natural
product,
but
but
it
would
be
good
to
give
them
a
quick
look
over,
and
we
can
also
make
them
aware
of
us
as
well
in
time,
because
we
would
I
would
love
to
be
able
to
get
things
like.
What
is
a
network
service
like
actually
into
her
into
their
definition,
and
it
looks
like
that
Nicolai
trying
to
try
to
connect
them.
A
A
G
Probably
so
yeah
short-term
goal
is
to
finish
the
glossary
and
then
I'm
gonna
kind
of
I'll
probably
do
like
another
poll
or
something
there's
people
in
my
camp
where
I
want
to
then
start.
You
know
really
diving
deep
into
individual
terms
in
here
and
building
out
its
documentation.
You
know
along
sup
the
code.
I
know
other
people
are
kind
of
anxious
to
look
at
use
cases
and,
as
opposed
to
you
know,
just
basically
building
out
like
there's
a
framework,
and
maybe
we
have
two
parallel
efforts.
G
D
I
H
I
G
B
D
B
G
D
G
The
next
bullet
my
threw
up
there
and
I
just
wanted
to
like
maybe
throw
this
out
at
the
group
I
know
we're
very
heavy
kubernetes
focused
right
now,
but
the
more
I
think
about
it
and
the
more
and
more
I
wonder
how
we
push
adoption
into
telcos
early
I
mean
especially
if
you
scroll,
and
you
see
all
of
the
different
talks
other
than
like
you
know,
service,
mesh
days,
etc,
etc.
Like
80%
of
the
conference's
all
say
that
there's
a
NFV
sub
context
or
a
telco
sub
context.
G
H
G
Them
a
service,
a
network
service
that
allows
me
to
host
all
of
the
web
apps-
and
you
know
just
basic
utility
services
that
I'm
running
in
my
data
centers
and
then
service
chaining
them
via
NSM
to
a
V&S.
Via
you
know,
virtual
csr,
virtual
firewall
or
whatever
in
like
a
VMware
or
an
OpenStack
environment
I
think
would
catch
people's
attention
right,
so
I
mean
trying
to
boil
the
ocean
right
now
and
solve
the
CNS
thing
is
probably
you
know
a
long-term
goal,
but
showing
immediately
how
all
these
companies
can
use
their
brownfield.
I
You
know
I
think
this
is
an
important
topic,
I
believe
it
Frederic
and
myself
own
facility
when
discussing
about
it
and
some
of
those
can
be
probably
courier.
There's
a
protocol,
OpenStack
Korean
tries
to
bridge
kubernetes
and
OpenStack
world.
Probably
we
can
get
some
insights
from
the
that
is
one
second
thing
is
also
I'm.
Looking
at
to
integrate
its
opendaylight,
then
the
simpler
one
was
essentially
to
use
the
G
RPC
endpoint
and
invoke
from
opened
a
rate,
but
for
sure
there
are
different
models.
I
think
we
can
discuss
a
very,
very
important
area.
B
G
Yeah
I
think
if
you
had
something
as
simple
as
just
putting
like
you
know
an
Apache
nginx
web
server
and
a
kubernetes
cluster
and
then
having
a
virtual
firewall
sitting
in
an
open,
stacker,
VMware
cluster
somewhere
and
just
writing
a
simple
service
across
that
I
mean
very
similar
to
like
the
the
bridge
domain
example
that
Fredrik
gave
last
year.
I
I
think
that
that
would
make
the
light
bulb
like
illuminate
for
a
lot
of
people.
G
It
definitely
was
a
key
part
of
it
right,
I
mean
so
I
do
a
lot
of
stuff
with
bridge
domains
and
OpenStack,
specifically
because
on
the
commercial
side,
every
customer
needs
a
completely
isolated
instantiation
for
their
network,
and
you
just
spinning
up
a
bridge
domain
with
a
couple
lines
of
go.
Definitely
resonated
with
me,
but
ultimately
you
know
you're
not
selling
to
me
you're
selling,
to
the
people
that
I
work
for,
and
they
just
want
to
see
that
they
can
take
I
mean
we've
got
tens
of
thousands
of
nodes.
G
You
know
managed
by
both
VMware
and
OpenStack
and
showing
these
you
know
exact
that
you
can
take
all
that
infrastructure
and
continue
to
innovate.
Alongside
of
that
on
the
application
side,
I
think
is
a
huge
value.
Add
because
I
mean
the
truth
of
the
matter
is
is
no
matter.
You
know
how
many
like
packets
per
second
cease.
It
shows
us
there's
no
vendors
right
now
that
are
selling
me
a
virtual
firewall,
that's
worthwhile
and
I'm.
G
B
Because
we're
gonna
have
to
have
some
way
to
bridge
from
what's
existing
right
now
to
what
it's
coming
right.
That
absolutely
has
to
happen,
and
even
when
I
talk
to
the
most
ardent
supporters
in
actual
IT
organizations,
is
you
know
the
kinds
of
people
who
tell
me
they
don't
expect
to
do
anything
new?
It's
not
kubernetes
a
year
from
now
those
exact
same
people
will
tell
me
that
they
have
vm
is
running
at
OpenStack
or
VMware.
That
will
continue
to
run
like
that
for
decades.
Right
and
so
that's
I
think
it'll
be
very
important.
B
B
G
F
G
On
the
VMware
side,
and
then
I
mean
just
being
someone
who
lives
eats
and
breathes,
OpenStack
I
can
tell
you
that
they're
perpetually
trying
to
find
relevance
in
the
age
of
kubernetes,
and
you
know
with
the
airship
and
the
Ukraine.
Oh
people
trying
to
figure
out
how
they're
going
to
stitch
these
little
I
don't
even
know
what
to
call
their
blueprints
but
I
think
giving
them
a
tool
like
this
is
something
that
they
would
latch
on
to
and
we
could
probably
co-op
some
development
fingers
to
smash
on
keyboards.
B
Before
and
I
think
it's
to
get
there
before
cube
con,
oh,
and
so
you
know
you
sort
of
followed
through
and
I
gave,
links
to
much
the
plate.
The
various
places
we're
gonna
have
to
do
a
project
proposal.
Probably
we
want
to
come
through
as
a
sandbox
project,
but
for
that
we
need
the
sponsorship
of
two
committee
members,
technical
Oversight
Committee.
Now
they
just
seeded
in
brand
new
talk.
So
there
are
a
bunch
of
new
folks
there.
What's
probably
gonna
be
helpful
is
if
we
go
look
at
the
list
of
folks
on
the
talk.
B
B
B
So
also
I
just
wanted
to
sort
of
get
that
out.
There
there's
still
work
to
be
done,
putting
together
the
proposal,
etc,
etc,
but
I
wanted
to
sort
of
start
the
wheels
turning
in
people's
heads,
particularly
since
you
know,
we've
got
a
pretty
broad
social
network
here
within
the
group,
and
so,
if
you
know
folks,
you
could
reach
out
to
that
would
be
excellent.
A.
B
B
B
Which,
I
think
is
probably
a
generally
it
also.
You
know
it
also
makes
available
sort
of
CNCs
resources
in
general.
You
know
so
I
think.
Overall,
it's
a
good
thing
for
us
to
associate
with
CNCs.
That's
we're
really
quite
cloud
native,
and
you
know
so.
Basically
you
know.
If
you
look
at
the
entry
requirements,
you
know
effectively,
we
have
to
have
to
stock
sponsors.
We
got
to
present
to
the
talk.
We've
got
to
adhere
to
the
CNC
FIP
policy,
which
would
include
transferring
at
a
trademarks
on
our
collateral
I.
B
Don't
think
we
have
any
trademarks
under
collateral,
so
we're
okay,
but
this
isn't
generally
good
right.
It
makes
people
more
comfortable
consuming
you
know
and
then
effectively
we
would
have
to
prominently
indicate
that
we
are
in
fact
a
CNC
F.
You
know
project
so
I
think,
overall,
it
would
be
good.
You
know
it
also
opens
up
the
opportunity
to
do
things
more
formally
with
the
CNC
f.
You
know,
I
know,
for
example,
various
projects
have
done
sort
of
co-located
events
by
virtue
of
being
a
CNC
F
project
and
a
bunch
of
other
good
things.
D
A
B
Let
me
over
all
the
the
CN
CF
guys
are
actually
pretty
awesome
about
the
support
that
they
provide
to
write
their
projects.
I've
been
on
other
calls,
for
example,
I
was
on
a
call
at
one
point
with
the
I
think
this
was
the
ongoing
community
and
Chris
Anna
check
was
on
the
call,
and
the
Amba
community
was
mulling
about
needing
to
do
some
things
with
various
virtual
infrastructure
at
a
public
cloud
and
I.
Think
Christiana
chicks
comment
was
as
long
as
you
give
me
the
necessary
information
to
pay
the
bill.