►
From YouTube: Network Service Mesh WG Meeting - 2018-10-23
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
D
B
B
A
B
A
A
B
E
B
B
E
B
B
B
B
Okay,
that's
going
to
get
started
so
we
have
in
the
events
we
have
cubic
on
Seattle,
coming
up
from
December
10th
through
13th.
Inside
of
that
we
have
a
intro
to
network
service,
mesh
talk
and
a
deep
dive
talk
and
we
are
in
the
process
of
making
a
demo
of
which
we
will
talk
more
about
soon.
We
are
also
looking
for
people
to
contribute
through
the
form
of
a
podcast,
blog
or
phyto.
B
B
A
Go
through
and
try
and
provide
for
doing
the
explained
mechanism
for
the
BTB
data
plane,
so
I,
you
know
I
went
so
far
as
to
sort
of
go
through
and
say:
okay,
you
know
you
I
do
to
roll
everything
out,
starting
from
okay,
what
the
hell
is
VX
land
and
then
rolling
through
your
various
pointers
to
okay.
Well,
how
do
I
do
this
in
VPP?
Okay?
How
do
I
do
it
up
to
CLI,
so
you
can
poke
at
it?
How
do
I
do
it
with
the
API?
A
A
B
Gonna
suggest
that
exact
same
thing,
so,
let's
do
it
and
as
an
action
item,
anyone
who
wants
to
have
their
github
account
added
and
the
ability
to
take
to
be
given
ownership
of
of
items.
Put
your
github
ID
down
on
this
document
and
I
will
go
through
after
the
meeting
and
add
everyone
who
does
so
put.
B
B
B
C
B
A
B
A
B
Okay,
X
Factor
CNF,
so
I
I
continue
to
shop
this
the
surrounded
different
groups
and
different
people,
so
we
can
get
some
more
some
more
people
looking
at
it.
I'm
going
I
am
in
the
process
of
thinking
through
some
of
the
some
of
the
things
on
the
first,
the
first
12,
because
I
the
first
world
need
to
be
rewritten
to
be
more
cloud
cloud
native,
Network
specific.
B
B
B
Okay,
MSM
enhancement
proposals,
support
with
the
CN
CF
CNF
project,
so
that
continues
to
be
a
ongoing
thing
that
we
are
that
we
are
doing
so.
This
is
this.
A
large
part
of
this
is
being
discussed
through
the
through
the
demo,
so
we'll
defer
this
a
test
to
a
sidecar.
Two
containers
is
p'tee
con
by
any
chance.
A
B
A
Sfe
is
quite
to
the
point
where
it
needs
to
be
merged,
but
I
do
know
that
it's
been
a
really
good
point
of
discussion
because
it
sort
of
lays
out
the
the
network
managers
used
to
talk
to
each
other,
and
so
there's
been
a
lot
of
back-and-forth
in
discussion
on
ok.
What
should
this
look
like,
and
that
kind
of
thing
in
so
more
voices
here
would
definitely
be
helpful.
B
B
A
A
We
aren't
actually
plugging
the
hardware,
quite
yet
in
general,
and
so
you
have
to
do
all
kinds
of
things
to
get
the
underlying
server
right
and
so
for
the
moment,
I
sort
of
took
the
huge
pages
requirement
away
by
disabling
the
DP
TK
plugin
we're
not
using
it,
and
this
means
that
we
can
run
the
new
VPP
data
playing.
You
know
pretty
much
everywhere.
A
I've
got
a
patch
further
down
that
runs
it
in
vagrant
I'm
working
on
a
patch
that
as
soon
as
I
sort
of
iron,
it
out
will
hopefully
get
get
it
running
in
the
ocean
test
at
the
circle
CI
and
in
Travis
CI.
So
it
it
makes
the
world
easier
in
the
short
term.
Eventually,
when
people
get
serious
about
deploying
these
things,
performance
they're
going
to
need
to
turn
on
huge
pages
because
it
does
make
the
world
go
faster,
but
that
doesn't
impact
the
actual
code
being
developed.
F
So
ed,
when
mm
IFR
packing
passing
back
and
forth,
they
don't
use
any
huge
page
backing
so.
F
A
A
Basically,
the
only
places
that
VPP
needs
huge
pages
you're.
Putting
aside
that
you
can
make
some
things
faster
with
them
is
situations
where
huge
pages
are
legitimately
required.
For
example,
DP
DK
insists
upon
them
and
in
those
cases
you
gotta
have
them,
because
otherwise
DP
DK
crashes,
with
a
very,
very
descriptive
error
message
for
the
way.
So
you
know
exactly
what
went
wrong
if
you
did
a
crash
crash,
clear.
B
A
It's
useful
for
CI
and
it's
also
used
for
useful
for
development
because
you
don't
have
to
go
running
extra
large
if
you're
running,
for
example,
VMs
on
your
laptop
for
your
cover
identity
for
developments,
you
don't
have
to
go
running
extra
large
VMs.
In
order
to
do
that,
because
you've
dedicated
gigabytes
and
gigabytes
to
huge
pages.
E
B
A
Yep
yeah,
that's
exactly
is
you
know,
super
super
easy.
If
you're
doing
development,
you
know-
and
it
includes
little
niceties
like
you
know-
make
docker
say
if
that
will
save
the
images
into
a
place
where
they
be
a
vagrant,
will
import
them
or
where
you
can
import
them
from
a
running
vagrant,
simply
so
that
you
can
sort
of
iterate
quickly
and
there's
also
a
readme
that
sort
of
talks
you
through
all
the
sort
of
incantation
for
use.
B
A
E
A
Know
if
you're
working,
for
example
on
things
in
the
data
playing
you
can
know
immediately
if
you've
broken
the
data
playing.
When
you
push
your
patch,
which
hopefully
I
know
that
makes
me
sleep
better
at
night,
and
it
also
means
as
a
reviewer.
You
could
also
know
that
you
know
at
least
it
deploys
successfully.
B
Okay,
cool,
so
who
knows
I?
Don't
we're
gonna
need
a
lot
of
time
to
discuss
the
main
main
agenda
items
so
we'll
head
back
to
the
main
to
the
main
agenda,
and
we
will
jump
straight
into
the
cube
con
demo
for
network
service
question.
B
A
It's
a
very
productive
things
there
and
then
the
following
slide.
They
sort
of
got
into
questions
about
okay.
What?
If
we
were
to
draw
pictures
of
the
architecture,
so
they
drew
a
picture
of
sort
of
where
CNI
stands
today
and
then
you
lead
to
another
picture.
The
next
slide
with
CNI
looks
like
with
networks,
sir,
but
what
the
system
looks
like
with
network
service
mesh
running
and
then
they
had
some
thoughts
on
the
user
interface
in
slide.
A
Nine
in
particular,
I'm
very
interested
in
the
sky
dive
option
because
they
already
does
a
ton
of
good
topology
visualization.
We
would
just
need
to
get
someone
who
would
be
interested
in
going
and
writing
a
sky
dive
probe.
So
that's
another
interesting
thing
that
folks
are
looking
to
pick
something
up
and
part
of
writing.
That
would
also
be
working
together
to
help
figure
out
what
makes
sense,
as
API
is
out
of
network
service
mesh
for
communicating
topology
information
to
subsystems,
cool
and
then
slide
10.
A
G
G
Yeah
I
think
trying
to
use
the
same
server.
So
as
long
as
they're
meeting
else
requirements
sounds
good
and
we
already
have
several
projects
that
we
could
invite,
whoever
so
packet
net
project,
specifically
under
the
CN
CF
org.
We
could
invite
people
into
those
and
then
you'd
be
able
to
do
deployments.
This
is
on
the
CNCs
lab,
and
so,
if
you
go
to
CN
CF
io
under
their
community,
it's
the
packet
net
lab
and
we're
using
access
to
that.
G
B
B
B
Yeah
and
this
this
will
absolutely
help
us
push
it
forward
as
well.
So
it
makes
it
very
easy
for
people
to
understand-
and
this
is
very
important
for
me,
like
start
using
sky
dive,
which
may
or
may
not
be
useful
for
people
in
production.
I
know
that
the
team
is
aiming
if
they
have
it
down
there
already
I
know
that
the
team
is
aiming
to
become
production,
quality
that
it
certainly
like
being
able
to
even
generate
these
metrics
is
say,
is
a
absolute
requirement
for
us
in
a
production
in
the
production
environment.
So.
B
B
B
Okay,
so
X
Factor,
see
n
updates,
so
I
think
it
was
a
correct
me
if
I
get
the
name
wrong.
Chris
Mets
is
offered
to
help
with
the
with
the
overall
design
of
the
of
the
page
itself
and
he's
going
to
have
something
on
this
weekend,
a
new
design
for
for
the
website.
So
definitely
looking
forward
to
see
to
see
what
that
looks
like
and
I
will
continue,
putting
content
down
and
I
invite
anyone
else
who
wishes
to
help.
B
We
find
a
document
or
add
more
or
discuss
what
is
right
or
wrong,
or
anything
like
that
or
how
do
to
join
them.
So
there
should
be
a
very
low
entry
to
barrier
to
join
in
if
this
is
something
you
want
to
help
with
in
the
long
run,
something
one
of
the
things
that
I
am
definitely
gonna
ask
for
help
from
from
this
community
is
once
we
get
something.
It's
a
bit
more
settled.
B
It's
going
to
be
help
with
shopping
it
around
and
getting
people
to
anymore
people
look
at
it
so
expect
that
as
a
future
thing,
I
don't
think
Serkis
going
to
move
on
to
the
next
item.
I.
Don't
think
sir
gaze
on
the
call
today
so
I
believe
you
had
a
another
work
thing
that
preempted
him
do
you
have
any
comments
about
the
data
plan
work
that
he's
doing.
A
Stuff
has
been
merged
and
I
think
Illya
is
working
on
the
meh
enhancements
for
that
I
had
a
bunch
of
patches
that
we
talked
about
earlier
that
are
basically,
and
you
know,
trying
to
move
that
more.
The
mainstream
of
our
CI
and
I
think
there's
some
work
on
going
on
trying
to
integrate
that
with
the
network
service
manager
as
well.
B
A
B
A
Yeah
I
mean
overall
it's
working
quite
well.
The
the
only
thing
I
think
is:
there's
a
known
core
OS
issue
that
hopefully
should
be
fixed
in
packets
shortly
they've
been
wonderful
about
that.
So
I
think
we're
on
the
road
towards
being
able
to
make
it
switch
off
we're
just
not
quite
there.
Yet.
Okay.
B
G
Think
I
may
be
the
only
person
on
the
call
out
of
that
list.
So
I'll
speak
up
and,
let's
see
so,
we
got
several
different
parts
happening.
Michael
is
actually
he
had
to
step
away,
he's
not
available
until
November
and
bocek's
bringing
on
someone
else
that
he
works
with
on
the
performance
side
to
keep
moving
that
forward.
G
G
As
far
as
performance,
then
we're
also
working
on
the
OpenStack
deployment
side
to
do
comparisons
for
the
vnf
side
of
things
and
that's
been
the
design
for
what
we
want
on
the
clusters
and
then
and
we're
working
with
the
OpenStack
expert
actually
later
today,
so
that
they
can
start
implementing
the
different
parts
that
we
need
pour
the
comparison
and
then
we've
been
focused
on
the
automation
portions
of
the
project.
So
we
have,
one
of
those
parts
will
be
I.
G
Think
something
useful
for
the
innocent
project
is
turning
on
the
huge
page
stuff
doing
configuration
for
layer
two
on
packet
net,
which
includes
both
host
configuration
as
well
as
talking
to
the
packet
dotnet
API
and
configuring
layer.
Two
it's
not
available
in
any
of
the
libraries
or
packet
dotnet
sealife.
So
we
actually
have
to
talk
directly
to
the
API,
which
is
fine.
But
it's
what
we're
working
on
on
that
side
and
then
I
think
be
useful.
G
G
I
think
the
only
issues
are
around.
Maybe
some
performance
numbers
ed
is
aware
of
with
regards
to
what
we're
seeing
on
packet,
dotnet
versus
FTIL.
So
right
now
we're
tuning
so
I.
Don't
think
blockers
we'll
see
how
it
goes
with
the
new
person
that
came
on,
and
we
basically
got
stopped
in
the
middle
of
this
tuning
to
validate
that
so
I,
don't
I,
don't
think
anything
today.
B
A
B
Cool
and
let's,
let's
delete
the
one
from
rocky
but
add
an
action
item
to
describes
that
we
are
not
affected
by
kubernetes
Network
policy
so
effectively.
I
think
what
romkey
was
coming
up
to
talk
about.
The
next
agenda
item
is
that
they're
made
I
think
there
was
a
misunderstanding
in
terms
of
network
service
fashion
and
in
its
use
of
networking
in
communities
and
that
you,
like
the
network
service
messages
orthogonal
in
terms
of
its
networking,
and
so
you
put
a
policy
on
that.
B
There
is
a
there
is
a
hole
that
will
eventually
have
to
fill
in
terms
of
this
policy
look
like
and
that
can
be
we
have.
We
can
address
major
portions
of
that
through
to
our
custom,
resource
definitions
and
discoverability,
and
the
network
service
endpoint
itself
as
well,
but
there's
also
there
may
be
other
policies
that
we
may
need
to
to
address
within
network
service
mesh
once
once
we
get
the
basic
stem.
B
B
Yeah
and
so
I
can
see
potential
integration
points,
but
it's
not
they're
not
like
strong
requirements,
so
so
I
think
we
should
put
together
a
small
document
or
blurb
that
we
can
afford
people
to
when
they
ask
this
question,
because
this
question,
a
rock
you
brought
up,
is
going
to
be
brought
up
again
like
what
you're
going
to
see
this
question
multiple
times.
So
we
want
to
have
a
clearer
idea
of
this
and
when
people
ask
us-
and
we
point
it
to
it-
and
they
have
misconceptions
that
are
right
for
that
will
evolve.
B
Okay,
one
less
item
that
I
did
not
add
in
but
I'll
talk
about
it.
I
have
been
working
on
a
network
service
endpoint
to
reef,
to
redefine
and
tighten
up
the
api's,
so
no
major
changes
from
what
you
from
what
we've
discussed,
but
what
we've
discussed
we're
where
the
endpoints
are
need
to
be
updated
to
the
to
the
ideas
that
we've
had
that
we
currently
have
so
I'm.
B
B
B
B
B
Okay,
great
and
finally,
make
sure
you
add
yourself
to
the
github
issue.
It's
already
been
created.
If
you
want
to
have
issues
assigned
to
you
and
we
will
start
adding
names
and
with
that.
Thank
you,
everyone
for
your
time,
and
we
will
see
you
again
same
time
next
week,
all
right,
bye,
everybody
take.