►
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
Birds of a Feather: Radio Access Network LTE/CBRS Edge - John Studarus, Packet Host
Discussion on deploying and running Kubernetes clusters across micro datacenters supporting edge applications. In specific, we'll be talking about how Packet Host has architected its Kubernetes deployment to support these cell tower collocated clusters. We'll talk briefly about this new ecosystem of ultra-low latency networking via LTE and CBRS (Citizen Broadband Radio Service). We're interested in sharing our experience and then letting the conversation flow to hear how others are adapting to use this new ecosystem.
https://sched.co/MPXx
A
Okay,
Ola
I've
been
told,
that's
not
bienvenidos,
it's
been
be
good,
that's
how
you
say
it
in
Catalan
iam,
so
welcome
everyone.
My
name
is
John
studious.
Today,
we're
gonna
be
talking
a
little
bit
of
edge
infrastructure
and
experiences
deploying
out
software
across
some
brand-new
spectrum
available
in
the
United
States
and
some
edge
locations.
We've
been
not
turning
on
line
here.
Oh
to
start.
Here
we
have
some
interesting
graphics.
All
about
all
the
technologies
are
raking
up,
5g
and
all
the
players
in
there
with
the
CNC
F
leading
the
way.
A
Okay,
so
I
call
this
presentation.
The
edges
open
part
of
it
is
the
fact
that
there
is
so
much
open
source
and
how
much
Tekken
technology
involved
in
all
this
we're
gonna
see
us
across
the
way
how
the
edge
computing
is
opening
up
to
commoners.
It's
no
longer
the
walled
garden,
we're
seeing
anymore
and
making
it
available
for
businesses
to
use
most
of
all,
it's
open
for
improvement,
so
my
experiences
so
far
at
least
on
the
software
side
is
not
everything.
A
Is
there
that
we
need
so
a
lot
of
the
algorithms
that
are
available
in
kubernetes
and
OpenStack?
Don't
quite
cut
it
yet,
and
so
I've
got
some
ideas
and
this
is
a
buff.
So
it's
supposed
to
be
an
informal
and
get
participation
from
the
audience,
so
I'll
save
some
time
at
the
end.
I
also
want
to
hear
people's
feedback
on
this
and
how
to
get
everyone
else
involved
in
all.
This
quick
overview
is
a
slide
I
sold
from
OpenStack
summit
I.
A
Think
a
couple
years
ago
where
China
was
talking
about
what
edge
computing
is
all
about.
For
those
of
you
that
aren't
aware
data
centers,
you
know
we
got
our
centralized
data
centers
then,
as
we
move
closer
to
the
user,
we
have
switching
facilities
used
by
telcos
and
then,
as
we
get
closer
and
closer
to
get
cell
towers
and
as
we
get
closer,
that's
end-users
on
those
mobile
devices
or
self-driving
cars
latency
goes
down
and
that's
what
we
need
for
these
new
killer
applications
that
are
being
driven.
A
So
when
we
talked
to
people
and
even
at
this
conference
here,
what
is
the
definition
of
edge
computing?
So
we
have
some
people
envision
that
it's
on
sell
devices
that
it's
your
ultra
small
versions
of
kubernetes
that
were
installing
on
edge
devices.
Is
it
mobile
towers,
many
towers
that
are
being
driven
around
on
on
trucks
or
is
it
traditional
cellphone
towers
with
computing
infrastructure
moved
in
we'll
talk
about
that
more
in
a
little
second
and
getting
on
more
about?
Why
do
we
need?
Why
do
we
need
to
move
stuff
to
the
edge?
A
Well
prime
example:
one
of
the
killer
applications
is
self-driving
cars.
We're
gonna,
have
self-driving
cars
out
there
they're
going
to
be
able
to
talk
to
each
other.
They
need
to
be
ultra
low
latency
so
that
the
data
between
all
the
cars
gets
passed
back
and
forth.
So
the
cars
can
respond
quickly
enough
and
not
all
the
data.
A
Not
only
compute
power
is
going
to
be
available
inside
the
car,
so
we're
gonna
have
to
have
compute
somewhere
close
to
the
cars
alongside
the
road
easy
for
the
cars
to
connect
to
and
and
for
the
cars
to
communicate
to
across
them.
So
this
is
a
presentation
from
the
IFS.
It's
a
edge
computing
conference
done
every
December
at
the
same
time
as
AWS,
so
their
definition
of
Wedge
computing
is
it's
where
the
digital
and
physical
worlds
interact,
so
where
physical
locale
is
important
so
really
as
close
to
the
energy
user
as
possible.
A
Okay,
so
let's
talk
a
little
bit
about
what
we're
seeing
in
these
edge
locations
and
how
they're
becoming
open
source
and
open
technology.
So
the
locations
that
we're
gonna
see
a
little
bit
more
about
ARM
processors.
We're
seeing
them
rolled
out
in
edge
locations
arm
is
a
open
source
processor
design
originally
designed
for
cell
phones,
they're
great
for
edge
locations
because
they're
low
power,
high
core
counts.
Low
power
is
critical
edge
locations
because
we're
dealing
with
limited
amount
of
power
a
limited
space.
A
So
we
need
to
have
as
much
compact
computing
power
as
possible
and
these
locations
we're
turning
on
and
the
lower
power
that
you
can
do,
the
higher
density.
You
can
do
less
air
conditioning
you
need
and
oftentimes
these
edge
locations
might
not
have
very
much
in
the
way
of
air
conditioning
so
seeing
a
lot
of
foreign
processors
we're
seeing
a
lot
of
standardization
in
the
in
the
actual
cages
and
the
racks
and
the
cabling
about
all
this
equipment
is
being
put
into
since
these
edge
locations
are
becoming
open
and
they're.
A
No
longer
these
controlled
environments
by
telcos.
We
need
to
have
make
it
available
for
anyone
that
wants
to
come
in
to
and
ship
some
infrastructure
and
want
to
make
it
easier
for
them
to
just
ship
in
and
roll
in
a
single
you
equipment,
piece
of
equipment.
It's
not
gonna
be
staffed
by
people.
So
it's
gonna,
be
you
know,
FedEx
guy
rolling
out
once
a
month
to
swap
in
and
out
equipment,
so
it
needs
to
be
easily
swapped
in
and
out
so
we're
seeing
a
bunch
of
open
standards
available.
A
So
the
open
maintained,
foundation
and
they're
involved
in
standardizing
a
lot
of
the
cabling
a
lot
of
the
the
rack
formats
and
then
the
Open
Compute
Project
start
by
Facebook,
because
standardizing
a
lot
of
VM
server
frameworks,
the
motherboards
and
the
networking
infrastructure.
So
then
the
containers
that
are
being
deployed.
So
this
is
a
container
conference
interesting
enough.
A
The
ik
we're
using
shipping
containers
and
those
are
being
deployed
at
these
cell
phone
towers,
so
vapor
IO
they've
been
busy
building
out
these
containers
and
also
providing
the
backhaul
connectivity
from
the
friendly
cell
phone
cell
tower
locations
and
back
out
to
the
Internet.
So
these
are
half
sized
shipping
containers,
full-size
shipping
containers,
loaded
up
with
equipment
and
then
connected
into
the
cell
phone
towers.
So
let's
talk
about
the
spectrum,
that's
being
connected
onto
it.
A
Traditionally,
at
least
in
the
States
spectrum
has
been
auctioned
off
by
the
federal
governments
and
made
available
to
the
telcos,
so
it
ends
up
being
a
very
controlled
environments,
the
telcos
on
the
spectrum
and
it's
at
being
expensive.
You
know
both
for
the
telcos
to
purchase
the
spectrum
and
then
also
as
for
an
end-user
to
gain
access
to
that
spectrum
in
the
States,
things
have
changed
around
a
little
bit.
There's
some
new
spectrum
becoming
available,
CB
RS.
It's
called
sis
and
broadband
radio
service
to
3.5
gigahertz
spectrum.
That's
it's
very
similar
to
5g.
A
However,
it's
not
souls
to
the
telcos.
It's
freely
available.
You
can
buy
CB
RS
enabled
devices
that
will
use
the
spectrum.
You
can
use
it
with
lower-class
antennas.
If
you
want
to
use
it
within
a
office
setting,
you
can
get
with
larger
antennas
for
use
to
connect
to
a
cell
phone
tower,
just
by
using
cvrs
you're
not
going
to
get
Internet
connectivity
right.
It's
just
going
to
be
something
you
can
use
much
like
a
Wi-Fi.
A
It's
create
your
own
environment,
but
there's
nothing
to
say
you
couldn't
connect
it
on
to
a
cell
phone
tower
and
have
your
own
connectivity
out.
Maybe
if
you're
using
CRS,
you
don't
even
need
Internet
connectivity.
These
are
just
connecting
to
a
one
of
these
shipping
containers.
That's
at
a
cell
phone
tower
with
the
CBR
s
antenna
attached
onto
it
and
then
the
car
or
whatever
device
that
you
have
is
using
CPRS
connecting
back
to
the
antenna
and
using
the
compute
power
at
that
cell
phone
tower.
A
So
there's
no
need
to
even
backhaul
it
back
across
an
internet
connection,
so
CBR
s.
Currently
there
are
locations
enabled
at
Foxboro
Stadium,
which
is
the
home
of
the
Patriots
in
the
United
States
football
team,
so
football
stadium,
Las
Vegas
along
the
Miracle
Mile
Mall,
is
huge
mall
within
so
retail
use,
use
cases
there
Chicago
and
downtown
Chicago
there
number
of
installations
about
12
other
locations
turning
on
and
the
spectrum
should
be
available,
hopefully
by
nationwide
by
the
end
of
this
year
in
the
United
States.
A
So
people
ask
me:
is
this
fact
I'm
going
to
be
available
overseas?
There
are
other
governments
going
to
release
it
and
from
what
I've
heard
is
they're
wanting
to
see
how
well
it
goes
in
the
States
and
then
hopefully
they'll
adopt
the
same
model
overseas,
so
the
spectrum
could
be
become
available
overseas
as
well.
A
So
this
is
a
team
play
making
serious
and
a
CRS
available
along
with
the
cell
phone
towers,
so
we've
got
the
cell
phone
tower.
Operators
they're
the
ones
that
own
the
physical
land
at
the
cell
phone
towers
they're
the
ones
that
the
antennas
traditionally
this
used
to
be
the
telcos
and
now
they've
sold
that
sold
that
land
and
the
antennas
to
these
real
estate
investment
trusts.
So
crown
Castle,
American
Tower
SBA
their
major
players
in
in
the
US
and
Central
and
South
America,
so
they're
the
ones
that
own
the
space.
A
Then
we
see
companies
like
federated
Wireless,
they're,
the
ones
that
are
the
overseers
of
the
spectrum
and
they're
the
ones
that
certify
devices
they
use
CB
RS
and
they
also
have
sensors
that
monitor
the
spectrum.
And
then
we
have
companies
like
a
packet
they're,
a
barrel
service
provider.
There
are
the
ones
that
provide
physical
infrastructure
that
you
can
rent
by
the
hour
that
are
co-located
within
these
cell
phone
towers.
So
it's
a
combination
of
all
of
these
that
make
the
spectrum
available.
A
You
can
buy
the
devices
from
you
know,
traditional
channels
that
have
CBS
and
enabled
rents
the
infrastructure
from
a
company
to
packets,
select
which
edge
locations
you
want,
tie
it
into
the
CB
RS
antenna
at
that
edge,
location
and
bingo.
You've
got
yourself
some
edge
computing
at
a
container
environment.
That's
connected
back
to
your
your
devices.
So
we'll
talk
a
little
bit
more
about
my
experiences
going
out
and
deploying
software
across
those.
So
here
a
couple
of
pictures
of
the
first
ones
being
installed.
A
If
you
are
interested
in
learning
a
little
more
and
actually
coming
in
and
testing-
and
you
have
software-
that's
sort
of
I'm,
a
software
guy,
so
I
spend
a
lot
of
my
time.
Software
testing
I
am
looking
for
other
companies
that
are
interested
in
testing
out
and
using
some
of
these
other
locations.
So
there
is
a
hedge
Alliance
program,
I
feel
free
to
talk
to
me
about
that.
We
can
get
you
access
to.
Some
of
these
I
really
want
to
get
more
experience.
This
is
a
you
know.
A
A
Bop
session
I
want
to
get
more
feedback
from
people
and
get
people
involved
in
and
trying
out
these
locations.
So
here
just
a
couple:
more
slides,
I'm,
showing
a
location
with
the
antenna
with
Olien.
You
know
with
all
the
antennas
on
installed
on
it
and
just
sort
of
the
equipped
that's
inside
fit
okay.
So
enough
of
the
hardware.
Let's
talk
a
little
bit
about
the
software,
that's
making
this
all
happen.
So,
on
the
software
side,
we've
been
doing
some
work.
Building
out
cube
spray.
A
We
put
in
some
releases
into
cubes
right,
so
we
can
automate
out
the
installation
and
turning
on
off
the
hardware,
so
cube
spray.
For
those
you
don't
know,
it's
a
manner
of
deploying
kubernetes,
so
we've
been
deploying
kubernetes
across
these
installations
using
cube
spray,
it
automates
the
full
process
of
provisioning
the
hardware
and
then
turning
on
the
clustering.
So
the
full
clusters
available
we'll
be
doing
a
little
bit
with
work
Brook
with
work
work
with
rook
as
well.
A
Turning
on
a
persistent
storage
on
there
and
that's
been
working,
great
we've
also
been
working
with
Starling
X
on
the
OpenStack
side,
working
with
them
to
get
some
of
their
software
installed
and
running
on
these
ancient
locations
and
then
doing
some
initial
talks
with
a
couple
of
other
players
on
the
web.
The
Linux
Foundation,
one
of
the
software
systems
who
we've
been
playing
with
is
a
open
source
video
conferencing
system.
This
is
interesting
for
us.
We're
looking
for
what
those
killer
applications
are
that
are
ultra-low
latency
and
latency
sensitive.
A
So
jitsi
was
one
I
just
ran
across
it.
It's
an
open
source
system.
We
can
run
it
in
containers.
We
can
run
on
VM,
we
can
run
on
bare
metal,
so
a
bunch
of
different
installation
mechanisms,
and
the
idea
is
how
what's
the
best
way
to
go
and
deploy
something
like
jitsi
to
support
end
users.
One
of
the
features
jitsi
has:
is
this
cascading
solution?
So
if
you
start
up
multiple
server
sites,
instances
of
the
software
it'll
appear
the
server's
together
tabac
hall,
the
video
streams
back
and
forth
across
the
locations.
A
So
then
it
comes
down
to
me
as
an
operator
of
this
software.
Where
do
I
want
to
turn
all
these
instances
do
I
want
to
turn
you
know
which
edge
locations.
Sorry
I
want
to
turn
on.
Where
are
my
end
users,
and
this
is
actually
one
of
the
key
problems
we're
finding
is.
Where
do
we
turn
on
these
which
edge
locations?
Do
we
turn
it
on?
You
know
the
kubernetes
scheduler
doesn't
keep
track
of
where
your
end
users
are,
and
it
doesn't
keep
try
of
where
to
deploy
the
software
to
keep
latency
really
low.
A
The
kubernetes
scheduler
is
interested
in
deploying
and
we're
computing
resources
are
available
and
we
want
to
deploy
it
where
the
end
users
are
and
where
and
how
we
can
keep
our
latency
ultra
level.
So
this
is
just
a
little
bit
about
the
performance
gains
you
get
by
using
that
server
cascading
that
the
features.
So
we
really
wanted
to
get
the
application
running
at
the
best
location
possible.
So
then
it
probably
becomes
a
little
more
tricky.
What
do
you
do
when
the
user
is
driving
down
the
freeway
right
so
he's
using
one
edge
location?
A
Now
it's
driving
down
Highway
five.
You
know
a
couple
minutes
later
he's
no
longer
at
that
cellphone
towers
at
the
next
cell
phone
tower.
What
does
our
edge
application
do?
We
need
to
have
some
sort
of
schedule.
It's
going
to
be
smart
enough
to
say
it's
driving
down.
Highway
five
I
need
to
turn
on
a
container
with
the
application
running
a
couple
of
miles
down
the
freeway.
A
couple
miles
after
that
down
the
freeway
and
maybe
turn
the
old
one
off
how
many
users
are
there
within
a
region?
A
Okay,
you've
got
ten
users
within
this
region
in
downtown
New
York.
Let's
go
ahead
and
turn
on
an
instance
at
that.look
location
and
by
instance
it
could
be
a
container.
It
could
be
a
VM.
It
could
be
a
bare-metal
instance
of
the
software,
but
the
middle
where
the
cluster
management
software
needs
to
go
through
and
keep
track
lab
and
the
problems
just
going
to
get
more
difficult.
A
There
could
be
hundreds
and
thousands
of
these
edge
locations
turned
on
they're
all
going
to
be
connected
to
different
types
of
wireless
spectrum,
some
of
them
going
to
be
connected,
CBR
a
some
other
kind
of
connection
5g
from
different
providers.
Your
end-users
are
going
to
have
different
types
of
spectrum
available
on
their
edge
device.
A
Maybe
they're
not
going
to
have
CBR
s,
maybe
they're
going
to
be
connected
at
different
telco,
so
we're
going
to
need
to
have
some
sort
of
mechanism
of
tracking
those
users
where
the
users
are
what
connectivity
they
have,
what
our
latency
requirements
of
those
end
users
and
feed
that
back
into
the
system,
so
that
the
middleware
layer
it
goes
through
and
deploys
the
application.
So
from
a
software
side,
our
experience
is
going
through.
A
These
are
sort
of
things
that
we
would
like
to
see
improve
from
the
edge
systems,
so
some
site
service
registry,
when
software
gets
deployed
where
it
is
which
edge
location
is,
is
it
at
what
type
of
spectrum
is
it
connected
to?
Where
are
the
users?
These
are
the
things
that
I
mentioned
before
and
then
tied
that
into
the
application.
What
does
what
are
the
application
requirements?
How
low
latency
does
the
application
require?
So
the
scheduler
knows
when
it
goes
and
it
deploys
the
software?
A
How
many
instances
if
it
doesn't
need
and
how
close
an
edge
location
doesn't
need
to
get
that
user
to
and
when
to
scale
it
up
and
when
to
scale
it
out.
So
when
to
scale
up
additional
containers
within
one
edge
location
and
when
to
go
ahead
and
turn
on
multiple
education,
central
locations
there
so,
like
I
said
we
do
have
a
packet
does
have
an
edge
Alliance
program.
A
So
if
you
want
to
get
access
to
some
of
this
infrastructure
and
just
in
trying
it
out
if
you're
working
on
some
sort
of
clustering
software
or
enhancements
to
kubernetes
or
some
type
of
application,
we're
happy
to
work
with
you
and
get
you
supports
and
get
some
feedback.
So
we
can
work
together
on
improving
this,
so
hopefully,
all
together,
we
can
help
build
a
better
internet
here.
So,
like
I
said,
this
is
a
bird
a
feather
session
so
supposed
to
be
informal.
B
A
Right
so
I've
been
reaching
out
to
them
and
offering
them
access
to
this
infrastructure.
Oh
sorry,
we
repeat
the
question.
The
question
was:
what
there
are
multiple
different
edge
plays
in
edge
technologies
in
play.
Right
we've
got
Starling
X,
we've
got
a
Carina,
we
got
some
k3
s,
we've
got,
you
know
a
bunch
of
different
ones.
So
all
right
are
you
asking
which
one
he's
gonna
win.
C
A
A
A
Now
now
now,
we've
just
been
deploying
stricter
the
regular
deployment
through
using
cube
spray
to
go
out.
We
haven't
been,
we
haven't,
played
with
any
type
of
network
function,
virtualization
or
any
type
of
telco
applications.
Yeah
so
I
know
a
lot
of
the
edge
plays
are
focused
in
on
telcos
and
managing
telco
traffic.
We
have
haven't
been
involved
in
that
we've
just
been
looking
at
use.
Cases
like
CDN
acceleration
and
moving
more
traditional
compute
towards
those
edge
locations.
A
A
Okay,
so
the
question
is:
do
you
have
access
to
the
actual
spectrum
to
allocate
how
many
channels
you
know
are
allocated
and
how
much
bandwidth
you
have
as
a
normal
user
know,
you
would
come
in
and
you
would
connect
into
whatever
channel
your
devices
are,
there's
nothing
to
say
that
you
couldn't
go
to
federated
wireless
and
if
you
have
your
own
devices
and
because
they're
the
they're,
the
controllers
the
government
has.
There
are
three
players
in
the
United,
States
and
they're
one
of
the
ones
that
are
controlling
that
spectrum
and
working
with
the
device
manufactures.
A
A
C
A
Right,
so
the
question
is:
how
do
I
use
the
edge
Alliance
program,
so
the
illegible
Alliance
program
will
get
you
access
to
is
devices
that
have
been
deployed
so
wireless
devices
that
are
using
the
cbio
spectrum
at
the
edge
at
that
cellphone
tower
and
it
actually.
It
would
also
give
you
API
access
to
deploy
bare
metal
infrastructure
at
those
locations
which
then
you
could
deploy
whatever
type
of
cluster
or
cloud
you
want
to,
or
just
your
application
directly
out
to
the
bare
metal
so
you'd
have
access
to
bare
metal
infrastructure.