►
From YouTube: Build Your Own Private 5G Network on Kubernetes - Frank Zdarsky, Red Hat & Raymond Knopp, Eurecom
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
Build Your Own Private 5G Network on Kubernetes - Frank Zdarsky, Red Hat & Raymond Knopp, Eurecom
Private 5G networks are dedicated cellular networks, confined to user premises and tailored to a specific use case. In smart factories, for instance, they may soon enable remote control of robots, augmented reality-enhanced maintenance, and other use cases for which ultra-low latency, high bandwidth, and reliable radio connectivity to local edge computing services is a must. Did you know you can build your own private 5G network purely from open source software and off the shelf hardware? This session will introduce OpenAirInterface, an open source 5G radio and core network implementation, and how to pick and set up hardware for it. Participants will learn how the latest Kubernetes technologies like Multus, SR/IOV CNI, real-time workers, device plugins, etc. need to come together to support these exigent Containerized Network Functions on Kubernetes and to manage them using Operators.
https://sched.co/UaWZ
A
You
can
it's
really
nice
to
see
so
many
makers
and
builders
in
this
room.
So
let's
build
some
networks
here,
thanks
by
the
way
Randy
for
the
introduction,
my
name's
Frank
Starsky
I
mean
right
at
City
office
responsible
for
edge
computing.
My
background,
though,
is
in
telecommunication.
I've
been
working
in
mobile
networks
at
a
large
European
taco
provider,
and
this
is
the
reason
why
I'm
so
happy
to
be
here
on
stage
with
Raymond
and
have
done
this
project
together.
Please
Raymond,
okay,.
B
So
what's
all
the
fuss
about
5g
I
mean
you
saw,
you
saw
the
keynote
this
morning,
or
many
even
may
have
seen
the
keynote
this
morning
looks
pretty
fancy
a
pretty
pretty
expensive
equipment
up
there.
Really.
What
it's
about
is
ultra
flexible
radio
access
and
the
term
that
the
industry
used
is
actually
called
5g
new
radio,
it's
actually
a
very
poorly
chosen
term,
because
it's
actually
not
that
new.
The
radio
is
not
new,
but
that's
what
that's
what
they
that's,
what
they
called
it.
The
objectives
of
5g
really
are
to
boost
bandwidth.
B
That's
one
thing,
but
also
the
the
more
spectrally
efficient
systems
I
mean
the
speciality
places
like
in
the
United.
States
spectrum
is
very
hard
to
come
by
these
days.
I
don't
know
if
all
of
you
realize
that,
but
5g
tries
to
address
that
to
make
the
systems
more
more
effective.
From
the
perspective
of
using
the
spectrum,
the
spectrum
is
very
costly
for
the
operators,
and
you
know
there
are
different
ways
of
doing
that,
and
basically
one
of
them.
B
One
of
the
ways
is
it's
a
bit
technical,
it's
called
the
bandwidth
part
and
basically,
what
that
is.
Is
you
take
this
big
chunk
of
spectrum?
The
demo
that
you
saw
this
morning
was
using
a
hundred
megahertz
of
spectrum
compared
to
5g
to
4g
that's
five
times
as
much
okay
and
what
they've
decided
to
do
was
come
up
with
a
notion
of
the
bandwidth
part
where
you
segregate
pieces
of
the
spectrum
for
different
types
of
services.
So
what
that
means
is,
when
you
start
to
do
concepts
like
slicing
high
up
in
the
in
the
network.
B
The
the
other
big
thing
about
5g
is
the
the
evolution
of
the
course.
So
the
core
networks
and
what
they've
decided
to
do
is
actually
keep
the
compatibility
between
the
4G
and
Phi
3
cores
with
respect
to
the
radio
access.
The
5g
networks,
almost
everywhere
in
the
world
today,
are
deployed
still
using
the
same
core
as
the
4G
networks.
It
may
be
implemented
using
new
paradigms
in
in
the
cloud,
but
it's
fundamentally
the
same
set
of
procedures
that
will
change
as
time
goes
on
now.
B
The
way
they
do,
that
is
something
called
dual
connectivity
and
I'll
explain
that
a
little
bit
and
also
they
use
of
the
term
non
standalone
operation
non
standalone,
because
the
5g
network
depends
on
the
4G
network.
Again
in
that
demo
that
you
saw
this
morning,
there
was
actually
a
4G
network
there
too.
Now
that's
it's
fundamental.
B
Now,
the
other
big
thing
and
we'll
talk
less
about
that
here
is
because
actually
I'm
not
really
very
much
of
an
expert
on
this
part.
The
the
5g
core
itself
is
now
designed
for
a
cloud
native
architecture.
Ok,
the
4G
core
is
now
being
implemented
using
a
cloud
native
actors,
but
it
wasn't
designed
from
the
start
and
so
they've
put
in
certain
aspects
in
the
actual
design
of
the
the
core
network
itself,
so
that
it
lends
itself
better
to
a
cloud
native
implementation.
B
B
B
Maybe
you
don't
realize
it,
but
some
of
the
the
credit
card
machines
that
you
have
in
the
restaurant
they're
already
using
a
3gpp
radio,
it's
a
2g
solution,
but
or
3G
solution,
but
now
it's
going
towards
the
40
and
5g
case
as
well,
but
these
I
would
say
are
the
very
traditional
applications.
3Gpp
is
now
looking
for
non-traditional
applications
like
industrial,
augmented
reality,
remote
control,
fancy
application
is
drones,
but
a
real
application
is
farming
vehicles.
Okay,
we
want
to
remotely
control
farming
vehicles
due
to
to
optimize
agriculture.
B
That's
a
real
application
of
high
G
and
also
the
support
for
new
vehicular
services.
There's
something
new
radial
v2x,
which
is
very
popular
here
in
the
United
States
and
actually
promoted
by
Qualcomm
here
in
San
Diego.
It's
really
pushing
this.
This
idea
of
advanced
vehicular
services
in
the
in
the
5g
network,
so
this
is
all
going
to
come,
this
kind
of
stuff
and
basically
they
would.
What
they
wanted
to
do
is
address
requirements
that
are
not
satisfied
by
Y
feet,
Wi-Fi
or
4G
based
solutions.
B
Okay,
so,
as
you
will
see
in
what
I'm
going
to
talk
about,
the
network
density
is
going
to
explode,
so
they're
going
to
be
base
stations
all
over
the
place
and
so
you're
going
to
need
new
new
forms
of
scalability
and
you're
going
to
because
of
all
of
these
new
types
of
applications.
You're
gonna
need
resilience
for
critical
communications
as
well,
so
that's
a
very
new
thing
in
in
the
3gpp
world
or
in
the
in
the
commodity
telecommunications
world.
This
notion
of
critical
communications
and
providing
resilience
in
the
networks.
B
B
So
this
is
actually
an
industrial
use
case
of
what
we're
going
to
talk
about
today
and
basically
you'll
have
the
evolution
of
private
LTE
exploiting
some
of
these
new
features
that
for
the
for
these,
for
these
verticals,
okay
and
ultimately,
what
you
want
is
a
customized,
IT
and
radial
solution.
So
the
DISA
combination
of
the
operating
system,
the
server's,
the
radio
and
the
edge
computing
to
actually
for
these
for
these
types
of
applications.
Okay.
So
what
what
is
here
is
just
a
couple
of
slides.
B
What
a
5g
network
actually
looks
like
now
up
here
on
the
top.
What
and
what
I'm
trying
to
explain
is
where
you
know
where,
where
this
is
all
coming
from,
if
we
want
a
really
brought
broadband
system,
which
has
also
extremely
low
latency,
you
heard
that
this
morning
in
the
in
the
keynote,
somehow
that
implies
more
bandwidth,
both
for
the
broadband
part
and
the
latency
part.
If
you
want
a
low
latency,
you
need
a
large
bandwidth,
and
you
know
if
you
have
a
large
bandwidth.
B
That
means
you
have
to
go
high
in
frequency,
because
in
low
frequencies,
there's
no
power
bandwidth
left
in
the
United
States.
There's
nothing
left
here
in
Europe,
there's
not
much
more,
and
if
you
up
in
higher
frequencies
physics
is
going
to
limit
you
to
have
very
short
range
communications,
so
the
type
of
equipment
that
you
saw
on
stage
today
might
have
a
range
of
tens
to.
Maybe
100
meters
tops.
Okay,
they'll
push
it
as
time
as
technology
evolves,
but
today
that's
that's
about
what
it
is.
B
So
these
are
not
big
cells
they're
going
to
be
all
over
the
place.
So
if
you
look
at
the
little
picture
that
I
drew
down
here,
the
5g
network
is
built
upon
the
existing
infrastructure.
So
you
have
your
cell
towers,
which
are
covering
a
couple
of
miles
radius
or
maybe
10
miles
radius,
even
depending
on
the
frequency
band.
B
So
you
basically
always
in
coverage
with
the
4G
and
you
get
the
super
high-speed.
Ultra
wideband,
what
was
what
the
term
that
was
used
this
morning
5g
when
you're
near
the
the
what
they're
called
the
G
note
B?
That's
the
name
of
the
5g
base
station,
the
G
note
B
all
right
now.
This
will
evolve.
Also
the
the
4G
component
will
become
5g,
but
it
will
retain
the
same
features
of
the
4G
large
range
coverage.
Okay,
now
the
network
itself
I
won't
spend
too
much
time
on
this.
B
But
basically
you
have
two
pieces
in
the
network.
You
have
the
so-called
core
network
and
you
have
the
radio
access
network
today
in
the
in
the
keynote,
the
core
network.
Well,
one
of
the
core
networks
was
actually
running
in
Montreal
in
Canada
or
the
or
the
control
plane
of
the
core
network
was
running
in
Canada
and
the
the
user
plane
and
the
radio
was
actually
running
here
in
San
Diego.
So
those
two
pieces
were
geographically
separated
at
least
partially
geographically
separated
in
France.
B
B
So
that's
that's
really
what
the
what
the
the
network
looks
like
and
you
can
see
their
interconnections,
but
between
almost
all
of
these
elements
the
bases
are
interconnected
between
each
others
so
that
they
can
communicate
when
you're
doing
a
handover
and
the
core
network
entities
are
interconnected
with
the
base
station,
so
the
traffic
can
flow
in
and
out
of
the
radio
network.
Okay-
and
there
are
also
kinds
of
fancy
interfaces
that
3gpp
specifies
okay.
So
this
this
slide
here
is
really
about
open
air
interface.
B
So
this
is
the
software
that
we've
been
promoting
over
the
last
couple
of
years
and
also
some
of
the
friends
that
you
know
that
are
doing
similar
kinds
of
software,
not
always
in
a
community-based
manner,
but
similar
types
of
software,
and
basically
it's
all
built
on
the
premise
that
it's
feasible
to
put
a
fully
compliant
4G
or
5g
base
station
and
the
core
network
in
a
commodity,
x86
box.
Okay,
that's
fairly
recent
I
would
say
in
the
last
10
years
we
were
trying
to
do
it
a
long
time
ago,
but
it
wasn't
wasn't.
B
You
know
it
wasn't
something
that
was
close
to
commercial
reality.
Let's
put
it
that
way
and
today,
actually,
if
you
look
even
major
vendors
adopt
this
approach,
I
was
very
recently
looking
at
some
of
the
equipment
from
one
of
our
partners.
Nokia.
If
you
look
at
their
gino,
be
servers
and
you
could
basically
just
take
the
Dell
server,
take
off
the
Dell
logo
and
snap
on
the
Nokia
logo
on
the
server.
B
Now
for
the
run-of-the-mill
users,
like
us,
people
that
are
doing
proof-of-concept
today
there
is
software
out
there.
You
can
buy
software
like
a
company
like
a
Mario
soft,
which
happens
to
be
a
French
company
too,
that
produces
closed
and
commercial
software
for
doing
3gpp,
4G
and
now
5g.
Also,
you
have
our
software
that
I'll
talk
about
a
bit
open
air
interface,
which
is
open
source.
It
so
happens
that
it's
also
friendly
to
the
3gpp
standardization
process,
which
is
a
complicated
thing
in
the
open
source
world.
B
You
also
have
some
code
now
coming
from
the
o
ran
initiative,
which
is
an
initiative
of
the
bunch
of
operators
that
are
getting
together
to
try
and
influence
cellular
standards
standardization
and
bring
some
open
source
into
the
world
of
cellular
communications.
These
are
both
three.
What
we
can
call
3gpp
friendly
software
examples
and
there's
another
one
which
is
really
a
traditional
open
source
GPL
based
approach
like
by
a
an
Irish
company
called
SRS
or
software
radio
systems.
B
There's
another
one
called
open,
LTE
I,
don't
know
what
the
status
is
that
these
are
both
more
traditional
open
source
projects,
but
not
necessarily
community-based.
It's,
at
least
from
the
perspective
of
SRS.
They
try
and
commercialize
their
software.
So
they
can't
really
let
a
community
feed.
You
know
feed
into
the
into
the
codebase
itself.
So
with
all
of
this,
your
there's
now
an
emergence
of
radio
hackers
out
there
and
development
and
user
communities
that
are
experimenting
with
3G
3gpp
software
implementations.
B
And
that's
that's
really
where
we
stand
here
today
and
you
know
in
a
nutshell:
we
have
a
sort
of
a
democratization
of
the
radio
access
through
and
through
open-source
software
and
even
in
some
cases,
open-source
hardware
or
at
least
open
hardware.
So
that's
really
the
the
scope
of
this.
Now.
What
we
tried
to
do
was
put
together
a
community
that
has
some
form
of
governance
and
that
actually
encourages
collaboration
with
the
big
players
inside
of
3gpp.
Okay,
that's
and
that's
what
the
open
air
interface
software
Alliance
is.
B
How
do
we
put
together
a
community
that
includes
the
big
players
and
telecommunications
today
and
still
retains
the
methodologies
and
the
spirit
of
an
open-source
community
and
what
you
can
see
here?
Some
of
our
members
that
are
actually
feeding
this
and
you
can
see
some
of
the
you
know,
Nokia
Fujitsu,
they're,
they're
welcome!
Is
there
two
very
recently
other
companies
like
interdigital?
These
are
all
big
intellectual
property
holders
inside
a
3gpp.
B
We
have
lots
of
members
like
red,
hat
and
lots
of
universities
from
across
the
world,
including
several
here
in
the
United
States
that
are
contributing
now
quite
heavily
to
this,
and
we
actually
accept
donations
that
are
there
to
maintain
an
engineering
support
team
for
doing
see,
ICD
for
doing
community
management
and
building
and
also
for
managing
the
relations
with
all
of
these
big
guys
in
in
telecommunications.
All
right.
So
that's
what
that's!
What
open
air
interface
is
from
the
governance
point
of
view.
B
B
These
are
big
European
projects
in
5g
that
are
funding
us
very,
very
nicely,
and
also
our
partner
Retta,
we're
very,
very
fortunate
if
the
work
was
right
at
and
we
have
access
to
a
lot
of
the
software
tools
from
from
Red
Hat
that
that
are
that
that
Frank
is
going
to
talk
to
to
talk
about
in
a
couple
of
minutes
here.
So,
in
terms
of
hardware,
in
the
demo
that
you
saw
today,
the
equipment
that
was
running
in
France
is
actually
this.
These
little
radio
units
here
these
are
completely
open
design,
not
necessarily
by
us.
B
We,
we
put
them
together
using
open,
open
products
inside
there's,
a
radio
device
called
the
USRP
b200
it's
worth
about
five
hundred
dollars,
that
comes
from
from
National
Instruments
and
originally
from
a
Southern
California
company
called
ettus
research,
and
you
know
this.
This
can
give
us
up
to
50
megahertz
bandwidth.
You
can
do
almost
a
low-end
5g
system
with
this.
B
With
this
device,
we
put
a
little
bit
of
custom
hardware
in
there
for
the
for
the
RF
electronics,
so
that
you
could
actually
transmit
at
a
significant
range
and
we
put
a
crowdfunded
x86
board
in
there.
It's
actually
called
the
the
up
board,
which
is
a
very
low-end
PC,
which
is
there
to
allow
us
to
experiment
with
the
the
networking
between
the
radio
units
and
the
cloud
alright.
So
this
radio
unit
is
connected
by
one
gigabit
power
over
ethernet
to
the
cloud
at
your
to
the
data
center
at
your
home.
B
Alright,
so
it's
a
very
you
know
it's
a
very
cheap
device.
Anybody
could
build
it.
There's
there's
no
secret
in
building
this
device.
You
could
also
build
more
more
expensive
hardware.
This
is
this
is
the
hardware
that
we
have
for
proper
5g
outdoor
deployment
a
little
bit
different
than
the
the
5g
that
you
saw
on
on
stage
today.
B
This
is
the
5g
variant
that
you
will
have
primarily
in
Europe
and
to
a
certain
extent
here
in
the
United
States,
for
the
Sprint
networks
will
look
a
bit
like
we'll
have
equipment
a
little
bit
like
this,
a
little
bit
larger
range
and
in
lower
frequency
bands
than
the
one
that
you
saw
today.
But
this
is
much
more
costly,
but
it's
still
built
on
off-the-shelf
components
here.
So
we
have
the
the
high-end
USR
peas,
which
are
giving
us
hundred
megahertz
of
bandwidth.
B
This
is
gonna
cost
tens
of
thousands
of
dollars
to
put
this
together,
but
nonetheless,
you
can
put
it
together
for
a
for
a
proof
of
concept
design
and
you
could
even
find
RF
components.
We
found
a
small
company
in
Hong
Kong
that
provided
us,
the
RF
front
ends
for
this
very,
very
cheap
design,
and
you
know
very,
very
efficient
design
for
for
our
needs.
So
this
was
completely
built
up
of
off-the-shelf
a
bit
expensive
but
completely
configurable.
B
You
could
also
find
production
level
hardware,
which
is
not
from
the
big
guys
like
an
okay
accent,
Huawei,
okay,
very
similar,
but
but
but
not-
and
why.
This
is
interesting
because
some
of
these
some
of
these
smaller
equipment
vendors
they
provide
equipment
that
can
actually
be
seen
as
more
or
less
commodity
and
white
box
in
the
sense
that
you
could
connect
it
also
to
a
data
center
right.
The
networking
that's
in
these
devices,
they're
they're
radio
devices,
but
it
is
generic
enough.
B
You
you,
you
run
it
off
two
times:
twenty-five
Gigabit
Ethernet,
right
to
the
right
to
the
servers
or
right
to
the
switch
in
the
in
the
server
room,
and
so
I
listed
a
couple
of
companies
here
that
that
can
provide
this
type
of
equipment,
and
this
will
give
you
something
that
is
much
closer
to
what
is
deployed
in
the
field
today
by
major
operators.
Okay,
it's
less
generic
and
flexible
than
when
I
showed
you
before,
but
it's
it's
really
ready
for
deployment
in
very
specific
bands.
B
A
Thanks
Raymond,
so
we
have
this
collaboration
with
URI
come
to
actually
help
them.
You
know,
provide
real-time,
capable
systems,
help
with
the
CI
CD
flow
in
order
to
accelerate
the
development
and
delivery
of
open
air
and
the
face.
We've
been
collaborating,
containerization
and,
of
course,
recently
or
currently,
there's
work
in
progress
in
order
to
basically
run
all
the
software
development,
as
well
as
a
lot
of
the
experiments
with
yokum's
partner
in
the
open
air
interface
Alliance
on
cue
Benitez
right.
So
just
to
give
you
a
high-level
overview
of
the
deployment
architecture.
A
A
That
typically,
are
doing
both
the
data
plane
processing
as
well
as
signaling
or
control
plane,
processing,
but
they're.
Not
they
don't
have
the
requirements
to
actually
run
real-time.
On
the
other
hand,
you
have
the
e
note
B,
so
the
40
and
the
G
not
be
the
5g
distributed
units,
and
these
this
software
has
to
run
on
real-time,
capable
systems
right
and
then
to
the
far
left
you
see.
Actually
these
us
are
peas,
which
are
so
very
find
radios
that
are
not
on
the
cluster
but
are
connected
via
the
network.
A
Another
thing
I
would
like
to
get
to
to
notice.
Is
that
there's
a
lot
of
different
networking
involved
because,
as
women
showed
earlier,
you
know
all
of
these
components
have
to
heavily
communicate
with
each
other.
They
rely
on
multiple
network
connectivity,
and
you
see
that
both
of
these
qualities
and
there
are
more
qualities
that
are
needed
actually
to
run
the
software
on
various.
They
are
not
not
like
your
traditional
IT
applications
right,
so
some
of
the
extensions
were
necessary.
A
In
particular,
you
need
to
be
able
to
provide
real-time
support,
whereas
real
time,
meaning
real
time,
means
determinism
right,
deterministic
performance
and
that's
an
engineering
trade-off.
You
can
get
deterministic
performance
by
training
of
some
other
qualities
like
energy
efficiency
or
trading
of
throughput,
for
example,
but
it's
always
a
trade-off
right
in
order
to
tune
a
system
for
real
time.
You
really
have
to
touch
every
single
layer.
You
start
really
at
the
bottom,
with
a
machine.
You
have
to
tune
your
buyers
first
right.
A
It's
very
hardware
dependent,
but
you
have
to
tune
this
because
whatever
non-determinism,
the
hardware
gives
you
you
cannot
compensate
for
at
the
software
level
anymore,
going
one
way
up
like
steps.
Next,
stop
is
the
host
OS
configuration
right,
so
there
are
ways,
of
course,
to
set
Linux
two
to
schedule
with
FIFO
discipline
and
the
real
high
priority.
A
If
you
apply
the
real
time
use
case
or-
and
it
does
a
lot
of
the
motivation
for
you
and
then,
of
course,
you
can
do
further
tuning
based
on
your
workload
and
as
always
in
these
real
time
critical
systems
at
every
step
that
you
configure,
you
have
to
validate
that
you
actually
were
effective
in
this
in
this
tuning
next
day.
If
you
don't
want
to
do
this
manually,
so
what
you
want
to
do
is
that
cuban
is
does
it
for
you
and
so
in
an
open
shift,
which
is
the
relative
krq
mentis
distribution.
A
What
happens
is
that
we
use
the
Machine
API
to
declaratively
configure,
not
only
which
machines
with
which
operating
system,
but
also
how
to
configure
them,
and
so
here
you
see,
for
example,
a
way
to
configure
a
machine
for
real
time
by
specifying
machine
configurations
that
place
specific
system
D
units
on
to
the
system
and
do
some
of
the
tuning
that
I
showed
on
the
previous
slide.
And
then
you
create
a
mapping
called
that
actually
says
if
your
worker
is
supposed
to
be
a
real
time.
A
Worker
role
then
map
it
to
these
configuration
sets
right
and
when
you
do
this,
every
time
you
replace
the
machines
automatically
getting
reconfigured
and
which
now
the
next
step
is,
of
course,
also
on
the
cubelet
level.
You
have
to
ensure
determinism,
and
so
what
you
need
to
do
is
to
to
ensure
that
your
pods
have
certain
guaranteed
q:s
resources,
both
on
the
CPU
side,
as
well
as
on
the
memory
side,
both
really
go
together.
They're.
A
A
Another
difference
you
saw
earlier
is
on
the
networking
side,
especially
in
the
under
telco
workloads.
These
were
close
need,
multiple
interfaces,
because
you
want
to
separate
control
traffic
from
data
plane,
traffic,
so
traffic
that
carries
user
data
from
your
management
data
in
the
cluster
and
actually
the
tool
to
do
that
is
the
multi
CNI.
So
here
you
see,
for
example,
how
we
set
up
the
networks
for
the
G
note
B
software
running
on
the
real
time
workers.
You
have,
of
course,
the
default
faster
interface.
A
You
have
additional
interfaces
for
the
control
traffic
going
to
the
mobility
management
entity,
the
user,
plane
traffic
I'm
going
towards
the
packet
gateway
and
then
eventually
the
the
the
Internet.
And
then
you
have
what's
called
the
front
hall,
which
is
as
dedicated
network
high-speed
network
connecting
you
to
the
radios,
and
if
you
want
to
do
this
kind
of
setup,
what
you
need
is
to
define
so-called
network
attachments.
A
Now
we've
done
this
worked
together
to
actually
take
the
opener
interface
software
containerize
that
solve
all
the
networking
and
security
issues
around
that
and
then
putting
them
onto
the
queue
benitez
cluster.
If
you
want
to
build
this,
there's
a
lot
of
this
configuration,
it
has
to
go
on.
I
didn't
even
mention
all
of
them.
Yes,
PTP
time
synchronization
with
high
precision
there,
yes,
FPGA
support,
etc.
This
is
pretty
complex
to
set
up.
Fortunately,
there
is
a
related
open-source
community
called
a
crane.
A
You
will
find
the
link
in
the
references
at
the
end
of
the
slide,
and
then
we
currently
don't
provide
pre-built
images.
Here
are
the
commands
that
you
basically
need
to
build.
The
images
push
them
to
your
cluster
and
then
the
next
step
you
need
to
do
is
adapt
your
deployment
configuration
because
there
are
certain
things
that
are
specific
to
your
machine
setup.
So,
for
example,
how
do
you
find
the
software-defined
radios
etc?
B
B
What
we've
done
here,
if
you
recall
earlier
there,
was
a
picture
of
a
rooftop
installation
we've
taken
some
of
the
radio
of
coming
out
of
the
roof,
so
I'll
put
it
in
the
lab
so
that
we
could
control
it
here
for
this
for
this
little
demo
and
it
just
it's
just
showing
one
component
in
the
network,
how
you,
how
you
run
it
using
using
kubernetes
and
and
showing
also
the
standard
compliance.
We
have
a.
We
have
a
signal:
analyzer,
that's
next
to
the
next
to
the
setup,
which
is
showing
the
the
integrity
of
the
signal.
B
B
A
B
B
That's
in
that's
in
our
server
room
and
you
can
see
behind
there
those
turquoise
cables,
those
those
are
40
if
4
times,
10
Gigabit
Ethernet
going
to
the
going
to
the
server
room,
which
is
about
a
hundred
meters
away,
alright,
so
we're
actually
connecting
this
is
the
equivalent
of
the
the
radio
units
that
you
did
that
were
on
the
roof
and
that
you
saw
in
Frank's
Frank's
diagram
on
the
left
here.
This
is
a.
This
is
a
laptop
that
we're
using
to
act
as
the
terminal.
B
A
Console
that
gives
you
like
the
ability
to
drill
down
on
anything
that's
going
on,
in
particular,
of
course,
the
resource
consumption
in
this
case-
and
you
know,
if
your
deployments,
your
configuration
of
your
config
maps,
even
the
machines
machine
configurations
that
you
saw
earlier
and
at
the
end
of
the
day,
you
basically
start
this
command.
That
I
was
showing,
on
the
previous
slide,
to
kick
off
the
deployment
of
the.
In
this
case,
the
core
is
already
running
and
they're,
basically
adding
the
5g
base
station
here.
B
Okay,
so
you
can
see
here
the
normally
the
green
light
should
go
on.
That
means
it's
transmitting
on
one
of
the
antenna:
ports
we're
not
using
all
the
antenna
ports
here,
you
should
see
come
on
there
there
it
is
there.
It
is
okay.
Now
it's
gonna
go
over
to
the
left
and
you're
gonna
see
a
device
you
didn't
see
at
the
beginning,
because
we
split
this
into
pieces
within
the
second
time
we
filmed
it.
We
didn't
have
the
device
anymore.
It
was
unknown
for
a
couple
of
weeks.
B
So
what
this
is
showing
basically
is
the
is
the
signal
integrity
of
the
the
signal
that's
being
generated
by
the
the
the
cloud
deployment
of
5g
and
the
fact
that
you
actually
see
very
geometric
figures.
There
means
that
we
are
we
we
are
complying
to
the
standard,
and
this
actually
is
a
laptop,
that's
running
a
similar
device
down
on
the
bottom
doing
the
same
thing
in
a
software
implementation
of
the
terminal
right,
so
we're
just
showing
the
two
things
one
was
with
a
commercial
tester
and
the
other
one
was
with
open
air
interface
terminal.