►
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
Intro + Deep Dive BoF: Telecom User Group and Cloud Native Network Functions (CNF) Testbed - Cheryl Hung, Dan Kohn, CNCF & Taylor Carpenter, Vulk Coop
This birds-of-a-feather (BOF) session will discuss how telcos are evolving their Virtual Network Functions (VNFs) into Cloud-native Network Functions (CNFs) running on Kubernetes. CNCF is spinning up a new Telecom User Group for service providers and their vendors to discuss the adoption of cloud native technologies, and this session will also be a kickoff for that group.
https://sched.co/OBhN
A
Daddy,
aha
washes
CN,
CF
comes
to
the
Shanghai
C
home
from
gem
from
Shaitan
I'm,
the
CN
CF
director
of
ecosystem
chou-heung.
This
is
a
meeting
for
the
vendors
and
operators
within
the
telecoms
user
group.
So,
to
begin
with,
I
would
like
to
ask
who
is
here
from
an
operator.
So
please
raise
your
hands
if
your
phone
telecom
operator,
okay
and
how
many
people
from
the
vendor?
A
A
B
C
You
Cheryl
meet
me
how
I
am
I'm
not
gonna,
do
any
more
Mandarin
than
that.
We're
very
happy
to
have
the
kickoff
of
the
telecom
user
group.
Here
in
China,
we
had
the
first
meeting
a
month
ago
in
Barcelona
at
cube
con
Europe
our
goal.
We're
now
doing
two
phone
calls
a
month
as
our
show,
and
we
have
a
document
and
other
work
going
forward.
We're
also
very
open
to
doing
Chinese
language
meetings.
C
C
So
I
will
just
mention
the
next
in-person
meeting.
We're
going
to
have
will
be
at
the
open
networking
summit
Europe,
which
will
be
in
Antwerp
in
September,
and
we're
going
to
have
a
half-day
meeting
of
the
telecom
user
group.
To
try
and
move
things
forward
will
then
also
be
meeting
in
San
Diego
at
Q,
Khan
cloud
native
Con
in
November
I
hope.
Many
of
you
can
attend
so
we're
launching
this
group.
It
is
a
similar
to
our
end
user
community.
C
C
The
CNF
testbed,
which
Taylor
is
going
to
talk
about
in
a
minute
open
policy
agent,
which
has
a
lot
of
interest,
observability
tools,
Prometheus,
fluid
D,
Jaeger
and
now
the
newest
one
open
telemetry
service
meshes
like
linker,
D
and
sto,
and
I
will
point
out
that
sto
is
not.
Unlike
the
other
things
listed
here,
a
CMC
F
hosted
project
network
service
mesh,
which
is
a
very
exciting
but
still
very
immature
project.
C
It's
a
new
sandbox
project
in
ciencia
operators
as
a
way
of
packaging
up
functionality,
are
the
kubernetes,
IOT
edge
working
group
and
then
finally,
there's
a
link
here
to
using
kubernetes
as
an
inventory
manager
to
keep
track
of
all
the
equipment
on
a
network.
So
the
kind
of
analysis
that
we
want
to
do
the
kind
of
work
that
we
want
to
do
here,
as
I
said,
is
not
coding,
but
things
like
gap
analysis
of
talking
about
what
are
the
solutions
that
are
available
today
and
what
are
they
missing?
C
There's
a
particularly
three
areas
around
connecting
via
naps
and
CNS,
motifs,
danam
and
network
service
mesh
talking
about
best
practices
for
cloud
native,
4cn,
apps,
and
then
one
of
the
goals
here
is
that
if
we
are
beginning
to
work
on
a
white
paper,
if
that
white
paper
would
not
just
be
a
description
of
different
approaches,
but
that
we
would
actually
implement
those
different
approaches
in
the
CNF
test.
Bed
and
tailor
will
talk
more
about
that
and
then
be
able
to
make
real
evaluations
of
the
comparison.
C
You've
now
seen
an
evolution
over
the
last
couple
years
and
with
the
current
version
of
own
app
that
you
can
run
it
on
top
of
either
OpenStack
or
kubernetes
kubernetes
lets
you
run
on
top
of
either
bare
metal
or
any
cloud,
and
the
evolution
of
where
we
think
this
is
going
is
to
have
CN
FS
to
have
your
also
your
open
source
system,
your
business
support
systems
all
running
on
top
of
kubernetes.
That
kubernetes
can
be
this
Universal
abstraction
layer,
they're
still
going
to
be
a
number
of
legacy.
Vn
apps
virtual
network
functions.
C
Those
can
run
on
top
of
cube
bird
on
vert
lit
on
OpenStack,
and
then
you
can
still
have
the
own
app
Orchestrator
helping
these
connect
together
and
the
kubernetes
can
run
on
top
of
either
bare
metal
or
any
public
cloud
or
a
hybrid
combination.
So
this
is
just
a
phrase
that
I
like
describing
a
key
aspect
of
the
kubernetes
architecture.
The
entire
system
can
be
described
as
an
unbounded
number
of
independent,
asynchronous
control,
loops
reading
and
writing
to
the
schema
ties.
C
C
Okay,
I
think
I
just
have
a
couple
more
slides
and
then
I'm
gonna
hand
this
off
to
Taylor.
So
there's
one
more
kind
of
high
level
thought
that
I
would
communicate
here,
which
is
imagine
a
physical
firewall
device
that
might
have
been
installed
ten
years
ago
in
a
telco
and
operators
network,
and
then
maybe
five
years
ago
it
was
ported
to
become
a
vnf.
C
A
virtual
network
function
running
in
a
virtual
machine,
but
really
nothing
else
was
changed
about
it.
So
now,
if
you
want
to
take
that
firewall
and
make
it
a
CNF,
it's
no
longer
possible
to
keep
custom
kernel
packet,
patch
patches
or
to
use
custom
kernel
modules
and
this
it.
The
technology
needs
to
be
compatible
with
any
kernel
version,
310
or
higher.
That
could
run
docker,
but
that
very
minimal
number
of
changes.
C
What
you
could
call
a
lift
and
shift
approach
of
your
firewall
doesn't
actually
get
you
that
many
benefits,
and
so
the
particularly
you
could
call
it
a
so
that
that
if
we
call
this
a
bronze,
CNF
and
there's
a
argument
about
this-
that
maybe
we
should
call
it
a
higher
nor
a
copper
or
something
very,
very
bad
metal,
but
that
you
still
have
you
could
have
a
completely
proprietary
managed
an
interface.
Oh,
you
might
even
need,
like
a
Windows
PC
running
a
special
application.
C
In
order
to
configure
it,
it
can
have
stateful
storage
using
a
completely
proprietary,
opaque
format.
It
may
have
no
support
for
horizontal
scalability
to
be
able
to
go
across
multiple
machines,
no
support
for
vertical
scale,
ability
to
be
able
to
go
to
larger,
smaller
machines,
no
support
for
automatic
configuration
and
then
a
completely
proprietary
installer.
So
this
is
kind
of
the
worst
case
for
a
CNF
and
I
mean
it's
still
CNF
in
that
it
is
a
container,
and
it
is
something
that
can
be
orchestrated
by
Club
native
by
kubernetes.
C
It's
not
clear
that
this
generates
very
much
value
for
in
the
real
world.
Then
you
have
the
other
extreme,
we'll
call
this
a
gold
CNF,
and
the
idea
is
that
this
is
a
set
of
best
practices
about
how
you
can
implement
the
network
functions,
so
it
should
be
compatible
with
any
certified
kubernetes
implementation.
We
now
have
I
believe
89
of
those
it
should
be
stateless
so
that
it
can
scale
horizontally
storage
state
in
a
database
or
another
service.
It
should
be
able
to
run
unprivileged
to
be
more
secure.
C
I
mentioned
the
scaling,
support,
configuration
and
life
cycle,
so
it
should
be
upgradable.
It
should
be
configurable
using
standard
mechanisms.
Observability
is
key
in
terms
of
monitoring
and
tracing
and
logging,
supporting
all
of
these
projects
install
both
upgradable
and
then
to
the
degree
it
needs
custom
hardware
or
to
take
advantage
of
it.
There's
a
standard
ways
that
it
should
be
accessing
that.
So
we
would
like
to
engage
with
you.
C
That's
a
very
rough
draft
of
what
a
bronze
or
the
gold
could
be,
and
we
do
want
to
try
and
define
that
more
we'd
like
to
engage
with
you
and
one
of
the
key
ideas
would
be
to
define
a
silver
CNF
to
define
something
in
the
middle.
That's
good
enough.
That
is
adding
value,
even
if
it's
not
perfect
in
every
way.
But
you
could
imagine
in
a
couple
years
the
China,
Mobile
and
and
other
carriers
might
start
specifying
that
you
need
a
silver
or
CNF
as
part
of
an
RFP.
C
C
There
is
a
ton
of
progress
that
has
been
made
here
that
people
are
doing
PCI
level,
1
and
HIPAA,
and
ISO
27001
using
kubernetes
that
micro
VMs
are
a
useful
technology
that
people
are
investing
in,
but
it's
not
at
all
clear
that
that
those
are
necessary
or
useful
in
a
telecom
environment
and
with
that
I
will
hand
it
off
to
Taylor,
and
maybe
if
we
can
do
like
5
or
10
minutes
and
then
leave
5
or
10
for
questions.
Thank
you.
B
E
Leading
the
scenic
testbed
initiative:
this
is
one
of
the
initiatives
related
to
the
tough
on
user
group,
and
this
would
be
the
project.
That's
also
trying
to
complement
existing
projects.
So
there's
a
lot
of
other
initiatives
to
build
testing
platforms
and
software
is
a
different
level,
and-
and
this
is
there
to
complement
this,
so
this
is
on
github
and
the
idea
is
to
provide
examples
in
a
code
base.
That's
completely
open-source
that
anyone
can
download
and
see
what
may
be
an
existing
use
case
that
you,
using
as
a
token
operator.
E
You
have
use
cases
where
you
may
have
SRV
network
functions
and
firewalls
and
other
things,
and
you
want
to
show
how
they
work
together
and
you
want
to
collaborate
with
others.
So
this
is
an
area
where
you
can
download
the
software
and
test,
and
we
do
that
by
using
a
packet.
That's
there
in
the
bottom,
packets,
a
provider
that
has
bare-metal
servers
and
if
you
have
a
API
key
and
from
packet,
you
can
download
the
software
from
github
and
you'll
be
able
to
test
those.
E
E
What
this
is
as
IP
routers,
so
very
simple,
and
that
would
be
something
that
you
would
put
in
a
service
chain,
a
set
of
network
functions
and
we
send
the
traffic
through
those
and
they
loop
back
out
and
we
check
the
throughput
the
packets
per
second
and
then
we
show
the
same
sort
of
thing
on
kubernetes,
we're
also
on
the
bottom
area,
we're
showing
what
you
can
do,
which
is
currently
just
available
with
containers.
You
can
do
these
mem
if'
interfaces,
so
we
can
directly
connect,
and
this
is
on
one
physical
node,
one
physical
machine.
E
You
can
directly
connect
the
network
functions
via
memory
interface
and
and
then
we
look
at
what
the
throughput
is
on
that.
So
this
is
some
of
the
testing
that
we've
done,
and
this
is
something
that
could
be
applied
to
any
type
of
network
function
on
our
machines.
We
also
support
multi
node
machine
to
machine,
but
we're
doing
different
type
of
example,
test,
and
this
is
just
some
more
stats
on
the
comparisons
and
what
we're
doing
is
testing
performance.
E
Things
like
this
to
verify
that
you
can
you're
going
to
get
the
performance
that
you
would
have
expected
when
you're
running
it
as
a
VM,
a
virtual
network
function
that
we
can
meet,
meet
or
exceed
those
and
then
build
out
the
more
complex
use
cases
so
going
back
here
for
anyone
that
wants
to
participate
similar
to
the
the
telecom
user
group
has
white
papers
and
documents
that
you
can
contribute
to
for
understanding.
The
test
bed
is
about
contributing
code
and
real
examples
that
you
can
go
in
and
understand.
E
So,
if
you
create
an
API
key
unpack
it
you
can
download
the
code
and
you'll
be
able
to
recreate
it
on
packet.
The
code
is
portable,
we're
trying
to
make
it
where,
if
you
want
it,
you
could
take
pieces
out
and
test
on
your
own
lab
equipment
at
different
levels,
whether
that's
deploying
the
kubernetes
cluster
or
setting
up
the
network
or
deploying
the
actual
cns
most
of
the
CNS
or
originally
based
on
some
yep
code,
we're
starting
to
get
more
contributions.
E
We're
looking
at
to
show
a
hybrid
type
of
configuration
where
you
may
have
an
open
stack
cluster
running
of
the
BNF
from
a
vendor
that
hasn't.
Maybe
they
haven't
even
gone
to
a
bronze
as
Dan
was
saying,
and
you
want,
you
need
to
be
able
to
connect
to
that,
we're
going
to
be
showing
a
kubernetes
cluster
running
and
clients
connecting
over
a
tunnel
to
the
vnf
and
some
other
stuff
like
that,
so
that
you
have
that
migration
as
you're
moving
forward
on
your
platforms.
D
E
E
Some
difference
we
go
into
some
of
the
challenges.
This
goes
into
stuff,
like
the
gaps
that
Dan
was
talking
about.
What
do
we
need
in
the
platform's?
What's
missing?
What's
there
there's
a
lot
of
solutions
and
trying
to
add
pieces
around
the
device,
plugins
and
other
stuff,
so
we
want
contributions
both
in
code
to
test
new
examples,
as
well
as
to
the
different
white
papers
that
are
in
progress.
You
think,
that's
probably
about
good,
maybe
we
can
switch
to.
E
Doesn't
have
to
use
packet?
Yes,
so
right
now,
this!
This
is
kind
of
an
overview
of
the
stages
that
we
have,
but
this
is
to
setup.
If
you
wanted
everything
set
up
what
we
have
in
packet,
so
you've
downloaded
you're,
creating
a
top-level
configuration
number
of
notes,
machines
and
your
clusters,
and
then
you
can
run
some
software
that
we
have
that
all
provision,
the
machines
and
go
on
you
can.
Actually,
if
you
were
interested
in
using
your
own
kubernetes
cluster,
you
could
just
skip
down
and
jump
right
into
the
other
parts.
C
A
B
Hey
I'm
Marisol
Szabo
from
Ponte
own
tech.
Actually
we
do
some
some
CNF
deployments
for
some
of
our
customers.
I
do
not
have
any
specific
question,
but
maybe
some
some
additional
information.
If
anybody's
interested
in
CNF
there
is
an
open
source
project
called
arigato
which
actually
packages
FDI
of
EPP
into
a
docker
container
to
get
a
written
management
agent,
which
provides
some
some
northbound
API
outside
cloud
native
way
of
of
configuration,
configuring,
the
VDP,
that's
ready
for
anybody
who
is
interested
and
wants
to
play
with
some
CNS.
B
You
can
check
it
out
when
we
get
to
the
I/o.
Also
there
is.
There
is
one
more
CNI
which
I
would
mention
that
is
called
County
VPP,
which
runs
also
FDI,
o
VPP,
as
we
switch
between
the
containers
that
can
help
you
as
well.
If
you
need
to
connect
multiple
interfaces
towards
the
ports
and
also
it
supports
some
simple
service
change
between
the
bots.
C
Just
quickly
on
the
legato
I
think
many
of
you
have
seen
the
CN
CF
cloud
native,
interactive
landscape
and
it's
available
at
landscape.
Cn
CF
do
and
you
can
get
to
it
from
your
phone
and
I'll
just
mention
that
under
the
networking
section
here
we
have
legato,
which
was
the
project.
You
mentioned
a
control,
plane
management,
agent,
Burke,
FBI,
OS,
VPP
and
then
also
I
had
earlier
mentioned,
danam,
which
is
a
project
out
of
Nokia
that
provides
access,
connects
to
essentially
between
CN,
F's
and
vnfs.
C
E
So
the
if
it
wasn't
completely
clear
one
of
the
things
with
scene
of
test,
but
is
we're
not
saying
here-
is
one
piece
of
software
that
you
should
use
everywhere,
similar
to
CNCs
in
general.
It's
saying:
there's
many
options
and
here's
what
they
do
so
that
you
can
pick
the
one
that
works,
so
the
implementation
that
we
have
the
main
examples
that
we
have
right
now.
E
This
is
what
it's
covering,
but
we
could
use
additional
software,
so
we
actually
do
use
VPP
as
a
V
switch,
and
this
kind
of
goes
over
that,
but
we
have
here's.
Vpp
right
here
is
a
V
switch
on
open.
Second,
we
use
the
VPP
OpenStack
plug-in
that
connects
the
V,
the
talking
to
VPP
with
Neutron
and
then
on
kubernetes
we
have
the
equivalent
of
a
V
switch
running
both
as
a
hose
or
you
can
run
it
as
a
CNF,
and
that
runs
just
in
a
container
on
the
system.
E
E
Then
we're
interested
in
trying
to
build
that
out
and
Pontiff
VPP
is
definitely
something
that
we
were
looking
at
early
on,
but
it's
a
platform.
What
we're
not
trying
to
do
is
build
a
product.
We
want
to
help
the
projects
that
are
building
something
like
network
service.
Mesh
has
very
extensive
features.
What
we're
trying
to
do
is
highlight
the
pieces
or
a
use
case
for
discussions
and
that's
the
driver.
E
Unlike
the
telecom
user
group
is
white
papers
and
documents
to
have
discussions
on
how
best
we
should
go
about
it,
but
we'd
love
to
have
contributions
from
the
other
groups
and,
if
you're
doing
something,
that's
maybe
a
complex
setup
and
you're
wanting
to
highlight
parts
of
it
and
how
it
would
work
then
be
happy
to
work
with
you
on
that.
Could.
C
Yeah,
so
this
is
inconvenient
I
guess
you
gotta
click
through
it's
gonna
be
11
p.m.
here
in
China,
but
we
would
love
to
have
your
participation
if
you're
willing
to
stay
up
late
to
join
us,
and
then
we
also
do
post
all
these
calls
to
welcome
to
YouTube
if
you
have
a
VPN
afterwards
to
access
it.
But
these
are
all
open
meetings.
We
would
love
to
have
your
engagement
participation
and
then
we
also
have
a
slack
channel
and
the
white
papers.