►
Description
Jonny and Andrew are joined by the OpenShift team to talk about multi-arch support in OpenShift 4.13.
If you want to know about deploying OpenShift across multiple architectures including arm then this is the stream for you!
Want to learn more?
Try Red Hat OpenShift: https://red.ht/3uVxkya
Follow us:
Andrew Sullivan https://twitter.com/practicalandrew
Jonathan Rickard https://twitter.com/jrocktx1
OpenShift https://twitter.com/openshift
Subscribe to Red Hat's YouTube channel: https://www.youtube.com/redhat/?sub_c...
About Ask an OpenShift Admin:
Do you have questions? We want to answer them! Every Wednesday at 11am ET, hosts Andrew Sullivan and Jonathan Rickard sit down to answer your questions about Red Hat OpenShift.
About OpenShift:
Red Hat OpenShift is an open source container application platform based on the Kubernetes container orchestrator for enterprise application development and deployment.
OpenShift on redhat.com https://red.ht/3rYz5ZS
OpenShift on GitHub https://github.com/openshift/
About OpenShift Commons:
OpenShift Commons is a gathering of end users of OpenShift, an enterprise-grade platform for container application development, deployment, and management. The gathering focuses on emerging technology and case studies from end users who share their use cases, insights into their workloads, and lessons learned.
https://commons.openshift.org
Join us on Slack: https://openshiftcommons.slack.com/
#RedHat #OpenShift #Kubernetes #ContainerPlatform
A
A
Thank
you
hello,
good
morning,
good
afternoon,
good
evening,
wherever
you
are
in
the
world,
this
is
the
ask
an
openshift
admin
episode.
103..
You
may
notice
that
my
co-host
Andrew
Sullivan
is
missing
and
he
is
at
home
taking
care
of
family
business
and
he
will
be
back
soon
so
Andrew
we
miss
you
happy
belated
Father's,
Day
to
all
the
dads
out
in
the
world.
Happy
belated
Father's
day.
I
hope
you
got
a
nice
day
to
go
golf
barbecue
get
away
from
the
kids.
A
You
know
celebrate
it
naturally,
so
today
we're
going
to
talk
about
multi-architecture,
support
and
openshift,
and
it's
it's
something.
That's
been
coming
up
quite
a
bit
like
we
always
hear
about
deploying
on
x86
and
deploying
on
arm
and
deploying
on
IBM
and
stuff,
like
that.
So
like
now,
we've
got,
we've
got
the
Nerds
that
are
actually
going
to
be
able
to
talk
to
us
about
it
and
actually
give
us
the
details
on
it.
So
that's
going
to
be
a
lot
of
fun,
so
Duncan
you
want
to
go
ahead
and
introduce
yourself
I.
B
B
C
D
A
A
So
hopefully
you
have
questions
about
Military
Support,
but
if
you
have
just
questions
in
general
about
openshift
or
other
things,
go
ahead
and
feel
free
to
ask
we'll,
try
and
answer
them
as
best
we
can
so
before
we
get
started,
though,
let's
go
ahead
and
talk
about
our
top
of
Mind
topics,
and
let
me
see
if
I
can
find
these
little
buttons
over
here.
Yes,
all
right,
so
there's
been
a
lot.
That's
been
going
on
since
we've
been
on
so
I
think.
A
Summit
recap
we'll
do
a
recap
on
openshift,
Commons
and
yeah,
we'll
focus
on
all
the
new
things
that
are
coming
out
and
some
of
the
some
of
the
other
exciting
things
that
are
that
are
happening
around
red
hat
and
the
open
source
community.
So
the
first
thing
is,
it
was
announced
at
Summit
and
it's
not
really
new,
but
it's
new
for
us
I
think
it's
called
the
red
hat
service
interconnect
and
it's
it's
based
on
Discover
and
so
at
its
most
basic
level.
A
It's
app
to
app
network
configuration
that
a
user
can
provision.
So
it's
at
a
layer,
seven
and
it's
it's
something.
I'm
really
excited
about
like
where
my
team
is
looking
forward
to
doing
a
validated
pattern
based
on
Scupper,
hopefully
in
Q4,
so
towards
the
end
of
this
year.
So
it's
something
that
like
really
really
excited
about
and
I
can't
wait
to
use
it.
A
Red
Hat
data
Hub
is
the
red
hat,
Downstream
and
or
excuse
me,
the
red
hat
Upstream
product
for
the
Janus
IDP,
which
is
a
midstream
deployment
of
Backstage,
which
backstage
is
an
internal
developer
platform
that
Spotify
open
sourced
a
couple
years
ago
and
if
you're
not
familiar
with
backstage,
take
a
peek.
It's
pretty
cool.
A
It's
a
developer
platform
is
a
single
point
of
a
single
pane
of
glass.
Where
your
developers
can
come
in
and
use
Golden
path
templates,
they
can
do
essentially
they
can
deploy
microservices
and
deploy
Things
based
off
of
processes
and
images
and
templates
that
your
company
or
organization
has
already
approved.
So
instead
of
having
to
deploy
something
and
then
get
approval,
it
really
helps
enable
the
developer
to
do
their
thing
faster
and
again,
I'll
paste
the
link
for
Janice
idp.io,
and
that
is
again
our
that's
the
Midstream
product
right.
A
The
the
integration
makes
like
application
set
deployment
and
then
that
overall,
governance
of
your
application
deployments,
multi-cluster
management
that
much
easier
so
looking
forward
to
kicking
the
tires
on
that
and
hopefully
getting
some
more
people
from
the
ACM
team
to
come
on
the
stream
and
talk
about
that
all
right
and
then.
Finally,
this
is
a
little
bit
old,
because
you
know
again,
it's
been
since
May.
The
git
lab
operator
is
openshift
certified.
So
that's
a
big
deal
right.
We've
been
they've
been
working
on
that
for
a
long
time.
A
I
can
remember
two
and
a
half
years
ago
when
the
git
lab
operator
was
first
released
and
we
were
working
with
the
git
live
team
to
get
that
going
and
we
were
testing
it
on.
Like
different
versions,
so
yeah
I'm
glad
to
see
that
it's
finally
out
there
gitlab's
a
great
product
and
a
great
platform.
So
if
you're,
if
you're
interested
in
GitHub
to
get
lab
operator
should
be
out
in
the
the
the
library
the
operator
Library,
so
that's
it
for
me,
oh
what
I'm
sorry
I'm
lying!
A
That's
it
for
me
on
the
top
of
Mind
topics,
but
the
next.
So
next
week
we
have
Gerald,
Nunn
and
team
coming
on
to
talk
about
open
ship
and
devops
or
devops
and
openshift
so
really
excited
to
have
those
guys
coming
on
and
then
we're
working
with
a
couple
of
Partners
to
come
on
in
July,
as
well
as
trying
to
get
some
of
the
teams
internally
to
come.
Talk
about
things
like
service
mesh
and
Scupper,
and
you
know
some
other
things.
A
B
Thanks
Johnny
I
do
you
know:
I
was
you're
talking
about
summer.
I
was
actually
lucky
to
go
enough
to
go
to
my
first
summit
this
year
and
it
was
really
great
to
I.
Don't
know
about
anyone
else
who
managed
to
go
is
starting
to
go
back
to
conferences,
but
it
was
Nate
really
good
to
get
some
face-to-face
time
with
customers
and
partners,
and
everything
like
that
and
you're
talking
about
the
Keynotes-
and
this
is
a
nice
segue
into
what
we're
about
to
talk
about
today.
B
Actually
one
of
our
customers
that
kind
of
kicked
off
one
of
the
projects
we're
going
to
talk
about
today
was
in
the
keynote
and
gave
us
a
call
out.
So
I
was
about
the
only
one
in
the
audience
with
my
arms
in
the
air
cheering
very
loudly,
but
it
was
really
good
to
see
your
product
get
mentioned
so
so
yeah
I
think
we're
here.
B
You
know
if
you
wanna
we're
gonna
start
with
a
few
slides
and
give
you
an
impression
of
where
we
are
and
what
we
did
and
why
we
did
it
and
then
Jeff
has
pulled
the
Short
Straw
he's
going
to
get
to
go.
Do
a
nice
live
demo
for
us.
Also,
fingers
crossed
I've
been
sacrificing
some
deities
in
the
background,
just
to
make
sure
everything
all
works.
So
I
probably
skipped
that
now
and
then
we'll
take
your
questions.
B
B
If
you
look
at
it,
we've
supported
multiple
architectures
for
a
long
time
now
you
know,
Andy
has
the
scars
of
bringing
openshift
onto
IBM
power
and
IBM
z,
and
so
you
know,
since
the
4-6
release
actually
slightly
before
that,
we've
had
support
for
openshift
on
IBM
power
and
IBM
Z
environments.
So
you
can
talk
about
that
area
of
things
and
that
we
have
multiple
architecture
support
today,
both
Jeff
and
Andy
worked
on
bringing
our
arm
developer
preview,
which
was
based
on
the
409
release
to
into
being
so
we've
got
now.
B
If
you
look
at
it,
x86
arm
power
ANSI,
the
openshift
will
all
run
on,
but
up
until
recently,
you
were
only
able
to
do
that
on
a
single
cluster.
So
your
cluster
had
to
be
you
know.
Homogeneous
is
one
of
the
terms
you
might
see
thrown
around,
but
it
had
to
be
all
around
all
around
one
type,
and
why
did
we
go
down
that
route?
B
Well,
for
reasons
one
is,
you
know
we
were
seeing
growing
at
adoption
of
openshift
anyways,
and
you
know
in
recent
years
we've
seen
that
pick
up
on
Power
and
IBM
Z
environments,
so
there
that
you
know
on
the
Mainframe
side,
they've
gone
through
the
Journey
that
many
of
us
on
the
x86
side
have
already
been
been
through
right.
You
know
Cloud,
native
and
and
cicd
and
bringing
all
those
environments
together,
so
we're
seeing
kind
of
that.
Picking
up
on
that
side.
B
The
other
reason
that
we
support
a
multiple
architectures
and
what
we're
here
about
to
talk
about
today
is
the
rise
of
arm
right.
I've
got
a
kind
of
another
slide
on
that.
So
we'll
talk
about
that
and
then
you,
the
customers,
your
need
for
flexibility
was
becoming
more
and
more
clear
to
us.
We
needed
to
address
that
so
that's
kind
of
one
side
of
the
house
and
that
that's
we're
not
going
to
talk
too
much
more
about
that,
because
we,
you
know
that
story
is
evolving
now.
B
But
what
about
the
other
side
of
the
house-
and
this
is
the
idea
of
a
single
cluster
with
mixed
architectures
inside
it,
and
this
is
the
story
that
we're
here
to
tell
you
about
today
where
we're,
what
we're
going
to
do
where
we've
been
and
the
reason
that
we
looked
at
this
was
again.
You
know
customers
driving
us
to
it,
but
there
were
some
key
questions
that
we
needed
to
ask.
So
we
needed
to
understand
what
the
main
use
cases
were.
What
were
the
benefits
to
you?
B
B
I
do
want
to
do
a
kind
of
highlight
for
the
team,
because
we
never
kind
of
we
don't
often
come
out
and
blow
our
own
trumpet
as
the
expression
we
use
over
here.
But
this
is
the
slide
from
our
413
release,
so
the
team
has
been
really
busy
on
on
the
arms
side.
Support
for
Armand
Azure
was
pretty
important
for
us
for
those
of
you
who
have
got
power
in
Z
systems
in
your
environment,
lots
of
work
going
on
there
all
the
time.
B
B
So
I
would
normally
at
this
point,
if
I
had
an
audience,
ask
people
to
put
their
hands
up
to
say
who
you
has
heard
of
before
who's.
Actually,
now
it's
more
about
who's
got
an
arm
system
in
their
in
their
room
and
at
some
I
asked
that
very
question
and
I
would
say:
90
of
people
put
their
hands
up
and
that's
either
Apple
Mac
users
or
people
that
have
gone
out.
B
You
know
there's
more
and
more
things
out
there
or
the
raspberry
pies
that
are
all
over
the
place
and
that
you
know
around
the
409
times
frame.
We
were
kind
of
looking
at
why
consider,
consider
using
arm
and
why
we
would
do
it,
and
why
is
the
industry
looking
at
Arms?
So
the
big
thing
you
know,
there's
multiple
reasons,
but
the
big
thing
I'm
seeing
is
just
this
incredible
performance
per
war.
B
You
know,
that's
why
it's
in
all
the
you
know
your
handheld
devices
that
you've
seen.
That's
why
it's
extremely
interesting
to
telcos
and
the
edge,
but
also
likewise
we're
seeing
more
and
more
companies
actually
pushed
to
Eco
standards
and
being
able
to
reduce
down.
You
know
their
cost
and
their
CO2
footprint
by
using
systems
like
this
in
the
right
places
is
very,
very
strong
argument
to
look
at
it
and
the
other
reason
why
we
looked
at
it
was
just
on
the
cloud
adoption
side.
B
B
So
you
have
the
apples
of
this
world
and
the
aws's
of
this
world
that
are
designing
their
own
arm
chips,
and
you
know
making
them
work
really
really
well
or
you
have
the
azures
and
the
gcps
of
this
world,
where
they're
working
with
companies
like
ampere
to
bring
their
systems
in
and
use
them
in
their
environment.
Essentially,
that's
a
long-winded
way
of
seeing
that
every
cloud
has
it,
and
now
customers
have
the
choice
to
do
it.
There's
an
another
add-on
there
that
there's
certainly
an
interesting
price
performance
piece
here.
B
So
when
AWS
first
came
out
with
a
graviton
offering
when
you
were
looking
like
for
like
for
instances,
there
was
roughly
a
40
saving
on
what
you
would
get
compared
to
an
x86
instance
of
an
equivalent
one.
The
differences
are
shrunk.
Recently
we
did
a
presentation
with
Microsoft
someone
from
the
Azure
team
and
they
talked
about
a
10
to
20
performance,
even
so
sorry
price
saving.
B
Now,
if
you
look
at
a
lot
of
companies
who
are
suddenly
finding
how
Lodge
their
Cloud
bills-
or
if
you
can
say
to
them,
you
know
I'm
going
to
save
you
even
two
percent
off
your
yearly
bill
or
your
price
and
cost.
That's
a
very,
very
interesting
thing.
Why
else
were
you
considering
on
well?
I've
already
talked
about
Edge
adoption,
it's
everywhere
right
and
it's
going
to
continue
to
be
everywhere.
B
So
all
really
strong
reasons
to
look
at
arm.
But
why
openshift
with
arm?
Well,
it's
all
about
choice
right.
The
thing
that
you
should
be
seeing,
Time
After
Time
with
openshift,
is
that
you
will
get
that
same
openshift,
feeling
experience
on
any
platform
now
that
you
want
on
any
Cloud
that
you
want
and
we
wanted
arm
to
be
part
of
that
ecosystem.
B
So
what
do
we
do
today?
What
do
we
support?
You
know,
I,
haven't
let
Jeff
and
Andy
have
five
minutes
without
asking
for
extra
features,
so
we're
running
we're
in
good
shape.
Now
we've
got,
you
know
our
full
stack
automation
on
our
pre-existing
infrastructure,
support
for
AWS
Azure
and
bare
metal
and
watch
for
Google
come
in
pretty
soon
now.
Is
that
as
well?
We
want
to
do
that
as
soon
as
possible,
so
we're
on
all
the
mail
platforms
that
we
need
to
be.
B
If
you
want
to
run
it-
and
this
is
really
just
you
know
benefiting
from
that
perfect
storm-
that
I
talked
about
on
the
slide
before
I.
Don't
know
how
many
of
you
resolved
as
I
am
probably
not
a
lot,
but
it's
definitely
when
was
it
1985
I
think
the
first
arm
chip
came
out
and
you
can
probably
Trace
back
to
the
that
next
year
was
going
to
be
the
year
of
arm,
was
kind
of
the
line
just
like
this
is
the
year
of
Linux
on
the
desktop
right.
Well,.
B
So
we're
actually
there
now
we've
got,
we've
got
the
hardware.
We've
got
the
servers,
we've
got
the
middleware,
we've
got
the
isvs
and
the
applications
on
top.
So
this
probably
is
the
is
the
you
know.
This
is
going
to
be
the
year
of
arm
and
we're
seeing
that
happening,
but
we
do
have
some
work
to
do
on
the
layered
product
support.
So
we've
got
the
main
things
there
and
you
know
going
with
what
Johnny
said
earlier
about
the
ACM
announcement.
B
Red
Hat
Advanced
cluster
manager
does
support
fully
and
that's
the
great
thing
for
us.
So
you've
got
something
to
manage
your
environment
and
your
and
your
openshift
instances
and
arm
and
we've
also
got
the
main
things
that
we
think
are
interesting,
so
pipelines
and
gitups
are
there
and
obviously
you
need
some
logins.
So
those
pieces
are
done,
but
we
haven't
got
everything
yet
we're
aware
of
that.
We're
beaver
and
Away
to
get
the
other
pieces
there
and
you'll
see
that
fill
out
as
the
roadmap
goes
on
so
okay.
B
So
that's
my
arm
pitch
as
you
can
tell
I've
drunk
the
Kool-Aid
I
had
the
fortune
to
speak
to
John
Masters
when
he
used
to
work
for
red
heart
and
how
he
definitely
did
a
Jedi
mind
trick
on
me.
So
I'm
all
been
on
arm.
But
there's
a
new
moment
coming
now,
and
the
413
release
was
the
new
moment
for
openshift
and
that's
the
idea
of
heterogeneous
clusters
or
mixed
architecture
clusters.
B
You'll
see
the
the
things
kind
of
being
interchanged,
and
this
is
was
the
right
hand,
side
of
the
slide
that
we
had
on
very
early
on,
and
you
know
with
Jeff
and
Andy
and
the
rest
of
the
team
and
a
help
of
a
few
customers.
We
went
away
and
we
wanted
to
look
at
the
main
use
cases,
and
these
were
the
three
main
ones
though.
B
B
B
Just
very
easily
add
things
in
and
if
you're
on
the
cloud,
that's
incredibly
easy
right
now,
they've
got
arm
instances,
so
you
want
an
arm
instance
in
your
cluster
Jeff's,
going
to
show
you
how
easy
it
is
to
add
one
in
now
with
openshift,
and
that
facilitates
this
gradual
rollout,
and
it
gives
you
some
interesting
opportunities
there.
You
know
you
can
bring
those
changes.
Services
that
you
want
to
change
over
first
see
how
they
go.
B
You
can
even
do
kind
of
benchmarking
check
how
they
are,
and
this
is
very
useful
to
some
of
our
customers,
because
even
though
it
doesn't
take
too
long
to
spin
up
an
openshift
cluster,
it
might
take
you
a
while
to
set
up
your
openshift
environment
or
the
your
application
environment.
The
way
you
want
it
there's
a
customer
that
we
work
with
now
that
it
typically
takes
them
two
weeks
to
kind
of
get
their
environment
where
they
want
it
to
be
so,
you
know
have
been
able
to
just
pull
some
arm
nodes
into
an
existing
environment.
B
We
aren't
no
that's
one
of
the
areas
that
we
want
to
talk
to
right
now.
Our
Focus
has
been
and
and
Jeff
will
show
this,
but
our
Focus
has
been
on
just
making
this
thing.
Well,
one
of
these
things,
multi-art
multi-architecture,
compute,
okay,
very,
very
easy
to
run
and
pull
in
so
we're
kind
of
just
kind
of
starting
on
that
ecosystem.
But
definitely
it's
an
interesting
idea
to
try
to
try
that
out.
B
C
B
An
interesting
topic
to
get
into
so
the
middle
one.
I
love
doing
this
really
really
bad
joke
that
no
one
loves
that.
But
this
is
a
for
those
of
you
like
me,
with
kids
of
a
certain
age.
This
is
straight
out
of
a
Disney
story
called
beauty
of
the
Beast.
This
is
a
tale
as
old
as
time
and
I
hope.
Some
of
you
now
that
I
have
young
kids
of
a
certain
age,
just
singing
that
song
in
your
head
over
and
over,
but
Legacy
applications.
So
this
has
been
gone
forever.
B
B
Maybe
they
just
don't
work
on
the
environment
that
you
want
them
to
work
on,
or
maybe
there
are
some
big,
very,
very
critical
service
that
you
need
to
keep
going
right
now
before
you
maybe
break
it
up
into
containerized
services,
and
you
just
want
to
leave
it
running
so
you
have
these
Legacy
applications
classic
applications.
Some
people
talk
talk
about
and
you
want
to
keep
them
in
the
environment
and
have
everything
else
running
so
somewhere
else.
B
B
You
know,
people
are
controlling
your
environments,
this
last
one
is
aim,
squalia,
I,
guess
Cloud
providers
or,
if
you've
got
an
internal
Cloud
that
you
use-
and
the
idea
here
is
that
you
want
to
you,
know,
make
the
best
use
of
your
more
expensive
systems
or
more
more
capable
systems
as
pure
compute
nodes,
and
maybe
you
want
your
control
plane
to
be
of
a
a
more
cost
effective
system
type.
So
the
one
that
I
actually
use
here
is
is
not
Arm
based
but
or
IBM.
B
Z
friends,
you
know,
you've
got
these
Mainframe
systems
that
are
highly
available.
You
know
massive
scalability
IO
of
the
charts,
but
you
just
really
want
your
main
applications
to
run
on
them.
You
don't
necessarily
want
to
use
that
a
state
to
run
your
control
plane,
so
you
could
stick
an
x86
or
an
arm
piece
in
there
and
again
to
go
back
to
the
call
like
I
made
earlier.
B
If
you've
got,
if
you,
you
know
a
providing
a
cloud
service
and
say
you
make
your
control
plane
based
of
arm
well,
if
you're
running
on
Azure,
you
know
10
to
20
savings,
there
that's
kind
of
instant
off
your
overheads.
There
yeah
we
we
actually
Johnny.
We
did
an
internal
kind
of
it's
back
back
of
the
napkin
coyote
calculation
and
we
think
you
know
just
because
you're
just
doing
the
control
plane,
but
we
reckon
about
a
12
saving.
B
D
A
So
yeah
I
mean
even
Even
in
our
internal
group
right.
We
we
are
eating
podcasts
like
crazy
and
like
it's
like
the
conversations
around
like
well
we're
spinning
up
these
control
plane,
nodes
they're
on
x86
yeah,
it's
just
the
standard
way
of
deploying
right,
so
you
have
an
m5xl
and
AWS
and
like
it's
not
cheap
over
time
and
so
repositioning
something
or
repurposing
into
an
arm
bit.
That's
that's
a
significant
cost
savings
right
like
that's,
that's
a
legit
thing!
That's
awesome!
Yeah.
B
That's
definitely,
you
know
anyway,
we'll
we'll
we
can.
We
can
have
another
topic
about
that,
to
put
that
a
pin
in
that
one
as
well
so
so
we've
got
two
technologies
to
introduce
you
today.
The
first
one
is
multi-architecture
compute
and
this
kind
of
does
what
it
says
on
the
tin
or
maybe
it
doesn't
see
it
on
the
tin.
B
But
the
idea
here
is
you
start
with
your
x86
cluster
and
you
can
add
in
arms,
arm
nodes
into
that
cluster
and
then
you're
good
to
go,
but
don't
just
take
my
node
take
my
node
for
it.
There's
a
nice
Freudian
slip.
Don't
take
my
word
for
it.
This
is
the
point
where
Jeff
is
going
to
show
us
a
fantastic
demo
of
everything.
So
let
me
just.
A
B
Yeah
I'm,
actually
I
can
that's
a
nice
segue
into
my
very
last
slide,
but
we'll,
let's
do
it
now
so
I
think
in
five
to
ten
years,
I
think
the
idea
of
hybrid
Cloud
that
we've
been
talking
about
so
I
guess
I'll
over
share
on
on
the
show.
So
when
I
was
I've
been
at
Red
Hat
now
I
think
I'm
coming
up
for
five
years
and
when
I
was
first
interviewing
for
red
heart
at
the
time
the
move
was,
you
know,
Cloud's
the
answer.
B
The
answer
is
actually
going
to
be
hybrid
cloud
in
the
middle
and
that's
where
we
play
really
well
and
if
you
look
at
Technologies
like
this
and
Technologies
like
ACM
and
ACS,
and
the
things
that
we're
doing
with
Quay
and
the
things
that
we're
doing
with
odf,
they
fit
really
really
well
into
this
hybrid
Cloud
idea.
So
I
think
that's
where
openshift
is
going
to
be
in
five
to
ten
years,
we're
going
to
be
sitting
there.
B
Everyone
will
be
using
it
and
if
you
want
to
run
your
application
in
Azure,
it
will
just
nicely
happen
for
you
and
the
network
connections
will
be
there
to
make
sure
it
happens.
This
storage
will
be
clever
enough
to
make
sure
the
data
is
in
the
right
place
to
to
make
sure
you
do
it.
If
your
application
is
well
suited
to
running
on
arm,
then
it'll
run
on
arm
and
it,
and
you
know
we're
just
seeing
that
we'll
just
see
that
footprint
grow.
There's
things
like
red
hat
device,
Edge,
that's
allowing
you
to
run.
B
You
know
kind
of
run.
This
minimized
open
shift
right
at
the
edge
on
very
small
systems,
so
we'll
be
interfacing
into
that.
So
I
think
we'll
just
go
to
this.
You
know
totally
analogous
Anonymous
environment
that
you
can
just
use
anything
anywhere
and
then
there's
you
know,
there's
other
Technologies.
We
talked
about
at
Summit
like
cxl
that
we'll
integrate
in
which
the
five
second,
what
is
cxl
is
basically
you
can
build
your
own
system
right.
B
B
So
that's
a
very
long
way
of
saying
we'll
end
up
in
a
very
flexible
World
and
what
will
become
important,
I
think
in
those
five
to
ten
years
times
is
extracting
away
the
complexity
so
that
the
applications
kind
of
you
know
you
just
throw
them
in
and
it
just
runs.
I
don't
know.
If
that
answers
the
question,
though,
but
that.
A
Was
mine,
yeah,
no,
really
well
really
well,
I.
Think,
like
one
thing
that
you
you
kid
in
on
that,
like
you
know,
I
completely
agree
with
this
like
yo.
If
you
look
back
to
like
virtualization
right,
we
went
from
physical
servers
to
virtualized
servers,
then
to
containers
and
then
we've
kind
of
it.
It
has
totally
been
a
pendulum
right
like
the
technology
like
it's,
it
hasn't
necessarily
changed.
It's
just
definitely
gotten
better
right,
like
I.
A
A
A
A
A
The
way
that
all
of
the
installers
work
is
that
they
essentially
use
terraform
on
the
back
end
they'll
use
your
credentials
and
they
will
provision
out
the
cloud
resources
if
you
want
to
integrate
like
if
you're,
if
you're
talking
about
above
the
infrastructure,
layer
and
you're
talking
about
applications,
there
are
ways
that
you
can
use
tokens
and
stuff
like
that.
To
there's
a
an
operator.
B
B
It
will
look
exactly
the
same
Andy
and
Jeff,
and
the
teams
have
done
lots
of
work
to
make
sure
that
you
know
we
build
it
in
the
same
way,
we
use
the
same
cicd
tools.
We
go
through
the
same
QE
pieces,
so
that
experience
should
be
exactly
the
same
where
you
start
to
see
deviations.
Actually
let
me
go
back
so
actually
let
me
not
because
it's
a
building
slide
and
where
you
see
deviations
is
kind
of
on
the
layered
products.
B
Now,
if
you
look
at
the
IBN
Z
side
and
the
IBM
power
side,
my
teammates
and
that
group
will
shout
at
me,
but
we're
kind
of
done
there,
we've
got
everything
that
we
need.
The
only
major
piece,
that's
missing
is
something
called
queer
and
that
should
come
pretty
soon.
So
you'll
have
have
all
the
pieces
that
you
need
and
and
Maria
don't
worry
about.
Taking
notes
here.
Andy
and
Jeff
helped
me
build
a
KCs
which
go
actually
goes
and
details.
B
B
I
think
ACS
is
coming,
acms
done,
odfs
coming
as
well,
so
we're
kind
of
building
that
out
so
you'll
see
when
we
bring
a
new
architecture
online,
we'll
concentrate
on
the
core
pieces
first
and
then
we'll
go
and
scroll
away
and
and
get
everything
else
in.
But
the
idea
is
what
these
platforms
morph
into
is
what
we
call
First
Class
architectures
and
what
we
mean
by
that
is.
They
will
launch
and
be
available
the
same
day.
B
So
when
openshift
launches,
all
the
architectures
will
be
available
and
we
will
strive
very
hard
to
get
everything
on
there
and
the
deviations
that
you
will
see
going
forward
are
just
the
main
new
projects
that
come
along
just
because
of
the
I
can
hesitate
to
say
this,
but
I'll
say
anyways
the
DNA
in
the
company
and
the
engineering
will
tend
to
do
things
on
x86.
First,
that's
our
biggest
platform.
By
far
far
at
the
moment,
so
you'll
see
things
there.
B
First
I'm
going
to
talk
about
something
called
hosted,
control
planes
once
Jeff's
done
his
demo
side.
That's
one
of
the
things
that
we,
you
know
we're
concentrating
on
the
x86
and
Cloud
site
first
and
then
the
other
architectures
like
arm
like
power
engine
come
I'll,
come
a
little
bit
later,
as
far
as
which
one
you
would
choose.
The
whole
idea
behind
us
doing.
This
is
that
you
choose
the
right
one
for
your
application
and
your
services.
B
So
I
was
talking
to
a
large
financial
institution
and
they
were
explaining
how
they,
when
they
bring
a
new
service
online.
It's
almost
like
they've
got
this
intricate
flowchart
that
they
have
to
go
through
of
a
bunch
of
questions.
You
know
is
if
one
go
to
three
and
that
decides
which
Cloud
they
run
in
what
architecture
they
run
on,
and
everything
like
that
and
what
we're
doing,
and
what
Andy
and
Jeff
are
helping
me
do.
Is
that
we're
making
an
environment
where
the
what
architecture
you
run
on
gets
eliminated
from
that
decision
tree.
B
Oh
I've
got
this
high-end
database
application
that
hooks
in
that
I
need
to
get
data
into
really
really
quickly
and
latency
is
very
important
and
I
need
it
to
be
highly
available.
I
might
need
that
to
run
on
my
zsys
IBM
Z
platform.
You
can
do
that
right,
so
so
we'd
like
to
be
in
a
position
where,
essentially
my
short
answer
to
your
question
and
Johnny
knows
I.
Never
do
short
answers
would
be
anything
you
want.
B
B
So
that's
an
interesting
one.
You
know,
given
the
experience
that
I've
got
I
I,
don't
think
you
can
ever
choose
between
the
two,
because
there's
different
disruptions
happen
at
different
times.
So
there's
always
this
drive
around
cost
I
think
that's
always
behind
a
lot
of
things.
But
then
you
get
these
descriptive.
Technologies
come
around
every
now
and
again
where
it
all
goes
out
the
window.
B
B
So
that's
the
technology
driving
things
for
for
my
mind
and
then
you
know,
you've
got
the
arm
side
of
things
where,
in
on
the
cloud
space
you
know,
I
wasn't
joking
before
I
I
didn't
want
to
say
it
because
I
I
always
misremember
things
but
I'm
sure
you're
talking
to
a
customer
where
they
were
saying
their
Cloud
yearly.
Cloud
bill
was
half
a
billion
dollars
right.
So
if
you
can
wipe
just
one
percent
off
that
never
mind
the
10
saving
that
we're
possibly
talking
about,
then
you're
gonna
get
driven
by
cost.
A
Yeah,
it's
it's
incredible.
Like
I
know
just
coming
from
my
customers
like
it
wasn't,
it
was
in
the
millions,
but
it's
definitely
in
the
hundreds
of
thousands
and
that's
a
lot
for
organizations
to
eat.
Right.
I
mean
like
especially
when
you're
when
you're
a
a
profit-based
organization
and
you're
trying
to
make
money
and
you're
just
gouging
out
to
these
public
clouds.
Yeah
it's
it's
important
right
and
so
every
little
bit
counts
I
mean
that's.
Why
people
use
spot
instances.
A
C
Well,
let
me
talk
to
this
slide
here.
Real
first
I'll
give
everyone
a
quick
overview
of
what
goes
on
here
and
then
I'll
then
I'll
show
you
basically
the
result
of
what
happens
so
kind
of
starting
from
the
top.
Here
you
know
you
can
go
ahead
and
create
an
open
shift
cluster
with
IPI
on
Amazon
and
basically,
when
you
do
that,
what
happens?
Is
you
have
a
completely
100
x86
cluster?
C
The
payload
and
everything
is
1A,
is
x86
and
basically,
when
I
say,
payload
I
mean
the
openshift
payload
that
has
the
release
version
and
then
all
the
components
of
openshift
NCD,
Prometheus
Etc,
all
that
stuff
is
all
built
x86
and
that
payload
and
so
to
enable
multi-arch.
What
you
need
to
do
is
basically
upgrade
the
cluster
to
the
multi-payload
and
all
the
multi-payload
is.
C
C
So
essentially,
what
that
enables
us
to
do
is
once
we
upgrade
to
that
payload
I
can
deploy
a
node
via
a
machine
set
or,
if
you're,
on
bare
metal.
It's
basically
a
very
manual
thing
where
you
deploy
our
costs
and
then
say:
hey
here:
here's
the
ignition
config
from
the
cluster
and
it'll
just
join
the
cluster
and
the
pods
that
spin
up
on
those
new
nodes
that
are
other
architectures.
Look
to
that
manifest
list
and
pull
down
the
right
architecture
to
run
themselves
on.
So
there
isn't
anything
funky
happening
in
like
the
openshift
API.
C
C
C
But
there's
a
caveat
with
that,
so
when
you
do
that
it
kind
of
get
into
this
a
little
bit
later.
But
when
you
do
that
you
have
a
note
of
a
second
architecture.
The
scheduler
really
is
an
architecture
aware,
so
it
really
falls
back
onto
the
user
to
taint
and
tolerate
that
node.
Accordingly,
so
they're
x86
workloads
don't
get
scheduled
to
that
thing.
You
almost
treat
it
like
a
GPU
machine.
At
this
point.
C
Right
appreciate
it
so
yeah,
so
this
guy
started
its
life
as
an
x86
cluster
in
Amazon,
Us
East,
one
and
I'll
kind
of
go
down
here
to
Cluster
settings
and
you
can
see
when
I
did
the
multi
upgrade
and
it's
basically
just
an
OC
admin
upgrade
command
multi-arch.
You
can
see
it
in
the
the
docs
there
and
when
I
do
run
that
command
it
basically
does
a
full
cluster
upgrade.
C
C
So
once
you're
at
that
point-
and
you
have
the
multi-payload-
you
literally-
can
go
over
to
machine
sets
and
I
cheated
a
little
bit
here.
I
just
took
an
existing
machine
set
and
changed
the
Amazon
instance
type
and
the
Ami
and
that's
all
you
really
need
to
do
I'll
go
ahead
and
do
it
here
live
on
just
another
machine
set
just
so
you
can
see.
C
What's
going
on
so
the
instance
type
m6gs
I
think
there's
like
M6
GM
7g
in
Amazon
c6g,
but
anyway,
the
G
ones
are
the
ones
that
are
armed
and
Ami.
You'll
need
the
Ami
for
the
region
as
well,
and
this
is
pretty
well
spelled
out,
I
think
in
the
documentation
that
you
put
in
there
Johnny.
So
if
I
just
go
to
another
machine
set
and
just
edit
the
yaml
file,
it's
it's
literally
changing
the
Ami
and
the
Machine
type,
for
instance,
typed
m6g.
C
C
C
Yes,
but
it's
not
supported
I,
think
there's
some
other
things
in
play.
There
there's
there's
certain
components
of
openshift
that
I'm
drawing
a
blank
on
that
use
the
control,
plane,
architecture,
okay
for
things.
So
if
you,
if
you
started
with
x86,
it
went
to
arm
I'm,
not
sure
what
Behavior
you
would
get.
I
think
image
streams
kind
of
come
to
mind.
B
B
If
you,
if
we
think
about
the
example
that
we
talked
about
earlier
about
this
cost-effective
control
plan,
then
you
know
helping
our
customers
move
to
that
space
is
something
we're
very
interested
in
and
there's
something
I
always
get
mixed
up
on
that
we
have.
That
allows
you
to
kind
of
change
the
control,
plane
and
migrate
it
to
more
powerful
systems
and
everything
so
we're
looking
at
using
that
mechanism
to
see
if
we
can
jump
architectures
at
the
same
time.
So
it's
a
really
it's
a
really
good.
B
A
It's
oh,
my
God
amazing
I
mean
just
just
from
an
infrastructure
person
right
because
I've
done,
the
like,
oh
oops,
I
can't
delete
that
machine,
but
I
did
how
do
I
recover
this
and
it's
such
a
pain,
sometimes
especially
if
you're
in
like
a
cloud
environment,
Because,
You're,
Expecting,
The,
Machine,
set
or
you're
used
to
compute
nodes
machine
sets
and
you're
right.
Oh,
it's
just
it'll
rebuild
and
then
you're
like
this
isn't
happening
and
so
like
it,
it
gets
a
little
dicey
and
then
Etsy
D
recovery
is
not
always
yeah.
A
D
C
C
One
thing
maybe
I
should
point
out
here
on
one
of
my
arm
configurations
or
a
machine
set
configurations
for
arm.
You
can
now
I'm
sure
most
people
know
this,
but
if
you're
doing
this
for
the
first
time
in
a
cluster,
that's
all
x86
add
a
taint
into
your
machine
set
on
your
arm
nodes.
Just
so
things
don't
get
scheduled
there
accidentally,
because
the
schedule
like
I
was
saying
is
not
architecture
aware.
It
doesn't
have
a
mechanism
to
basically
pull
a
pull
secret
for
your
containers,
inspect
it
and
to
know
what
to
do.
Gotcha.
A
B
So
let
me
be
super
clear
on
that,
because
you,
you
said
it
doesn't
matter
which
platform
so
for
arm
and
x86.
The
subscription
is
exactly
the
same.
So
you
you
just
buy
the
same
subscription
for
IBM,
Z
and
IBM
power.
They
are
separate
subscriptions
that
you
need
to
buy
from
so
in
the
environment
that
we
support
today
that
Jeff's
just
going
to
go
into
where
we're
just
basically
saying
multi-art
compute
supports
arm
as
a
day-to
thing,
then
you're
good
to
go.
B
A
Okay,
no,
that's
cool!
That's
fair!
So
we
got
a
couple
questions
that
came
in
and
let's
see
the
first
one
is:
is
there
a
oh
okay,
actually
I'm
going
to
read
the
wrong
person?
Here's
the
first
one!
Is
there
an
example
of
the
decision
tree
available
for
public
or
even
Partners?
As
far
as
like,
like
the
conversation
you're
having
with
Mira
earlier
about
like
when
you
would
use
IBM
Z
over
yeah.
B
We
don't
we
don't
have
one
Johnny,
because
I
our
Academy
a
thing
is
run
anything
anywhere.
The
one
the
decision
tree
that
I
was
talking
about
was
actually
the
customer's
decision
tree.
So
we
were
looking
at
the
points
in
it.
So
it's
not
something
we
even
got
to
see
it
was
kind
of
proprietary
to
them.
They
were
just
talking
about
certain
points
that
we
want
deliver
to
limit.
B
A
Awesome
and
then
Steven
had
asked.
Is
there
a
use
case
or
workflow
for
spinning
up
cheaper
arm
nodes
for
test
clusters
that
are
more
that
are
more
ephemeral
than
your
main
or
prod
clusters?
Basically,
from
like
the
ACM
get
Ops
perspective
right
like
is,
there
are
y'all
developing
like
any
kind
of
like
I.
Don't
know
Helm
chart
well
I,
guess
it
wouldn't
really
be
on
you
guys.
Anyways.
Do
you
know
of
anything
going
on
like
that?
Like
yeah.
B
B
This
technology
is
often
described
it
as
a
very,
very
short
pair
of
scissors.
So
it's
extremely
good
for
cutting
things,
but
you
need
to
be
careful
and
just
walk
with
it
and
carry
it
the
right
way.
So
we
don't
do
anything
around
scheduling,
for
example,
so
we
don't
put
tints
and
tolerations
in
there.
That's
an
exercise
for
the
end
user.
At
the
moment,
we
want
to
see
how
we
can
improve
those
types
of
experiences
and
kind
of
the
sky's
the
limit
here.
B
So
you
could
look
at
how
can
we
do
clever
things
around
scheduling
to
make
that
easier?
But
what
about
cost
models
right
to
you?
The
question
that
came
in
is:
can
we
do
something
clever,
working
with
ACM,
where
you
know
what
cost
models
are,
and
maybe
you
have
the
controlled
plane,
migration
tool
or
whatever
we
call
it
as
well,
so
that
helps
you
through.
B
So
when
you
know
those
are
kind
of
things
that
are
quite
a
way
out
that
we're
now
in
the
mode
of
okay,
what
can
we
do
next,
so
I
think
again
the
short
answer
to
that
question
is
we
don't
have
anything,
that's
looking
at
ephemeral
on
things
and
and
doing
it
right
now,
but
we're
looking
to
do
that
and
and
this
kind
of
lines
into
the
other
comment
around
there's,
not
a
lot
of
content.
You
know
I'm.
B
A
B
Idea
of
serverless
and
my
poor
understanding
is
you
just
give
some
functions
to
run
and
they
run
magically
and
everything
happens,
so
I
think
the
benefit
would
be
for
the
cloud
provider.
In
the
you
know,
those
Services
could
run
on
okay,
let's
say
it
more
cost
optimized
arm
boxes
if
they
needed
to
invisibly
to
the
end
person
using
it.
B
It's
probably
do
doable
the
only
thing
we're
putting
a
red
line
through
right
now
is
mixing
architectures
on
the
control
plane.
So
we
don't
want
to
go
down
that
rabbit
hole,
that's
kind
of
Pandora's
Box
territory,
so
we
don't
want
to
do
that.
Andy
was
nodding,
though
so
I'm
gonna
believe
that
he
thinks
it's
possible.
I.
D
Mean
it's
possible
I
I,
probably
wouldn't
recommend
mixing
architecture
and
control
plane,
though,
just
because
you'll
you'll
get
a
different
performance
out
of
each
of
the
hosts
and
it
could
like
skew
the
way
things
get
get
handled
and
I.
Think
I.
Think
you're
opening
yourself
up
to
some
some
pretty
interesting
issues
that
that
you
probably
just
want
to
avoid
by
keeping
it
the
same.
D
D
D
If
you
don't
have
to
care
where
they're
running
and
I
think
service,
mesh
and
serverless
having
having
those
applications
running
on
top
of
of
would
make
a
lot
of
sense
once
we
have
support
for
that,
but
there
are
always
instances
where
you
would
want
to
run
things
on
on
x86
or
other
architectures
like
like
Army
and
PNC,
so
yeah
I
think,
like
Duncan
said,
the
key
thing
is
like
being
able
to
choose,
but
I
think
when
it
comes
to
control
plane.
You
want
to
pick
one
and
stick
with
it.
A
C
You
have
basically
an
upgrade
a
multi-upgrade
Cincinnati
server
behind
the
scenes
that
does
everything
in
parallel
with
x86
you're
not
going
to
get
like
a
multi-arch
cluster
or
at
least
behind
the
single
Arch
clusters,
they're
all
in
parallel
kind
of
talking
a
little
bit
about
the
image
stream
stuff.
Basically,
the
old
way
image
streams
and
even
build
configs
too.
C
A
C
D
Going
to
go
ahead
to
the
next
slide,
it
was
kind
of
a
key
thing,
adding
the
Manifest
the
support
to
the
image
registry.
For
us
that
was
it
that
was
a
pretty
big
deal
yeah
it's
before
it
made
it
really
difficult
to
store
images
inside
of
of
openshift
without
having
them
like
totally
separate
images
like
image,
one,
the
an
image
one
x86
so.
C
B
Jeff
Sunday
the
thing
I
liked
about
it
is
OCD
book
didn't
work.
So
when
you
had
an
image
registry,
that
was
a
single
architecture
and
you
run
OC
debug
on
an
arm
node.
It
just
didn't
want
to
know
so.
The
team
did
a
lot
of
work
to
make
sure
those
things
kind
of
and
again
it's
about.
Sometimes
it's.
You
know
where
what
is
it?
I
can't
remember,
Convoy
or
something
like
that?
Wasn't
it
where
they
talked
about
rubber
duck
and
he
was
like
very
smooth
on
the
surface,
but
underneath
we're
paddling
like
mad.
B
D
C
Yeah
so
just
kind
of
review
some
limitations-
we
talked
about
a
couple
of
these
here
but
like
I,
was
saying
the
the
kubernetes
scheduler
is
not
multi-arch
aware.
So
it's
really
up
to
the
user
to
set
notification
change,
tolerations
labels
or,
however,
you
decide
to
do
it,
treat
it
like
GPU
nodes
or
whatever
fits
your
use
case
there.
C
Basically,
there
it
kind
of
goes
back
to
the
same
issue
with
the
scheduler,
the
auto
scaler.
It
really
doesn't
know
what
architecture
things
are
running
so
again,
you
have
to
basically
label
and
take
and
tolerate
for
it
to
work
properly.
It's
same
with
auto
scaling
from
zero
to
like
there's
There's
issues
there.
We
know
about
them
and
we're
working
on
them,
but
basically
it's
not
automatic.
There's
user
user
intervention
required
there
and
the
last
thing
I'll
talk
about
here
is
the
openshift
build
classic.
C
So
basically
we
have
image
streams
and
they're
in
the
registry,
but
you
don't
really
have
a
build
config
that
would
know
what
to
do
with
those
things.
So
you
can't
just
say:
hey
build
this
I
can't
say
build,
say
this
sample
on
both
architectures.
You
would
end
up
with
two
extremes,
basically
from
the
same
Source
image,
and
then
you
would
have
to
figure
out
how
to
stitch
the
output
of
those
things
back
together
for
it
to
work
foreign.
B
Okay,
excellent,
so
if
it's
all
right
Johnny,
if
you've
got
no
more
questions,
there's
one
more
thing:
I
wanted
to
touch
on
yeah
we're
good
to
go
excellent.
So
the
stuff
we've
talked
about
so
far.
You
know
openshift
running
on
arm
the
multi-art
compute
that
that
that's
owned
by
the
team
that
Jeff
Andy
and
I
live
in.
But
there
is
another
technology
that
we
have
to
talk
about
and
I
don't
know
if
you've
had
a
session
on
this
already
Johnny.
But
you
know,
maybe
everybody
hasn't
seen
it.
A
B
B
That
is
what
the
team
have
done,
is
containerize
the
control
plane,
and
that
has
some
amazing
deliberately
designed
in
benefits
and
I'm
going
to
randomly
chose
choose
the
kind
of
ones
that
I've
got
on
the
slide
here,
but
the
first
one
is:
you
can
lower
your
capex
and
op
X
costs
and
the
way
that
you
do.
That
is
because
the
controls
playing
essentially
now
runs
in
a
pod.
You
can
have
multiple
control
planes
side
by
side.
So,
if
you're
looking,
you
know,
let's
mix
the
architectures
in.
B
If
you're,
looking
at
a
IBM
Z
system
in
your
standard,
config
you're
talking
about
an
lpar
for
each
of
the
control,
plane
nodes
and
then
an
L
power
for
each
compute
node
that
you
have
it
soon
gobbles
up
the
capacity
of
your
of
your
box.
But
those
systems
have
a
lot
of
you
know
redundancy
built
in
so
you
can
actually
take
hosted
control
planes
as
an
idea
when
it
comes
out
on
IBM's
Z
system
support
and
have
multiple
control
planes
side
by
side.
B
I.E,
multiple
clusters
there,
so
you
can
do
you
know
again
again.
Another
tail
is
all
this
time:
consolidation
down
and
save
yourself,
soft
money,
there's
also
in
hosted
control
plans.
The
way
it's
designed
is
almost
the
separation
of
church
and
state
between
the
control
plane
and
the
the
compute
nodes
themselves,
and
that
gives
you
some
interesting
models
around
Network
and
Trust
segmentation.
So,
for
example,
you
know
the
a
developer
role.
That's
living
on
the
compute
nodes
has
no
view
into
the
control
plane
at
all.
B
I
don't
know
if
you're
all
like
me,
but
when
I'm
in
the
cloud
and
I
press
a
button
to
get
a
new
thing,
I
want
that
new
thing
straight
away
and
I
want
it
as
quickly
as
possible
and
hosted
control.
Planes
is
going
to
do
that
really
quickly
for
you.
So
this
is
like
a
really
big
change.
If
you've
got
an
internal
Cloud
offering
or
you're,
you
know
doing
some
cloud
provider
things.
This
is
going
to
mean
a
major
change,
a
change,
a
great
beneficial
change
for
you,
but
why
am
I
talking
about
it?
B
Well,
there's
a
couple
of
future
things
coming
along
that
make
it
really
interesting
and
the
first
one
that
kind
of
leads
into
that
lends
itself
to
the
question
about
you
know
five
or
ten
years,
whereas
openshift
going
to
be
I
mentioned
hybrid.
Well,
there's
the
opportunity
here
that
you
can
have
your
control,
your
compute
compute
nodes,
your
worker
nodes,
actually
on
different
clouds.
So
there's
an
idea
that
you
could
have
some
in
AWS
and
some
in
Azure.
B
So
if
we
flip
that
around
that's
right
in
the
use
case
that
I
talked
about,
so
you
could
have
an
arm
control
plane,
making
use
of
the
cost
efficient
side
of
it
and
then
have
your
x86
out
there.
So
it's
a
great
technology
I
see
Johnny
as
always
as
miles
ahead
and
has
posted
a
link
to
the
discussion,
a
full
discussion
about
it,
but
definitely
have
a
look
at
it.
B
A
Well,
I
I
think
that
goes
back
to
Steven's
question
earlier
too
right
about
like
how
a
use
case
for
managing
like
Dev
environments.
That
may
not
be
you
know,
like
production
workload,
Trail
lights,
let
me
spin
something
up
quickly
at
low
cost,
do
my
thing
and
then
fly
away
right
Yep!
This
is
an
awesome
way
of
doing
that
right
and
then
having
the
multi-arch
support,
where
it's
like.
Okay,
go
spin
up
a
single
node
arm,
or
maybe
a
two
or
three
node
arm
cluster.
Do
your
thing
you
know
get
commit
push
fire
away.
A
B
No
that
no
it's
I,
think
you
make
a
great
point
for
me.
It's
and
you
know,
we've
heard
this
mention
for
a
while.
You
know
the
things
about
pets
versus
cattle
and
we
seem
to
be
pushing
that
boundary.
I
mean
I.
Remember
the
the
old
three
point
x
open
shift
days
right.
He
installed
that
cluster
and
you
never
touched
that.
Damn
thing
again
right.
B
Hard
to
get
going
and
what
we
saw
was,
we
saw
very
few
very
large
clusters.
Typically
and
then
you
had
things
move
stuff
along
so
with
openshift
four,
we
re-architected,
so
we
made
the
install
and
upgrade
experience
easier
and
we
saw
that
reflected
in
Telemetry.
So
we
saw
more
and
more
customer
clusters
kind
of
appearing
at
customers
because
they
were
much
easier
to
create
and
they
were
much
easier
to
maintain
and
you
know
hosted
control.
Planes
is
going
to
take
that
to
extreme.
You
need
a
cluster
in
your
environment.
B
You
just
spin
up
a
new
one
and
I,
don't
know
where
you
go,
which
very
nicely.
Thank
you.
Johnny
brings
me
on
to
my
last
slide,
which
is
you
know.
Hybrid
cloud
is
multi-cluster
multi-architecture,
this
environment,
that
we
talked
about
where
you're
going
to
have
that
Hub
and
you
want
to
run
in
a
public
Cloud,
a
private,
a
cloud
Bare,
Mount
or
secure
Cloud.
You
want
to
run
on
arm.
You
want
to
run
an
x86.
B
B
A
C
A
Yeah
I
I'm
I'm
super
excited
about
this
right
because
you
know,
in
my
experience
it's
been,
it's
been
a
combination
of
like
a
private
Cloud.
You
know
either
it
be
openstack
or
you
know
some
region
in
like
the
the
gov
Cloud,
something
like
that
or
they've
got
some
bare
metal
somewhere
or
now
like
with
the
edge
conversation
poking
up
like
that's
we're.
Seeing
a
lot
of
this
and
I
had
a
customer
that
was
really
interested
in
like
the
multi-cloud
story
and
like
they
then
like
Raspberry
Pi
does
single
board
computer.
A
All
this
stuff
was
like
it
became.
It
became
the
chat
GPT.
If
it's
time
you
know
like
it
was
just
like
everybody
was
talking
about
the
singleboard
computer
and
so
like
that's
where
that
conversation
started
going
so
like
seeing
that
all
these
things
are
finally
coming
out
and
we
can
have
multi-cloud
like
legitimate,
multi-cloud,
bare
metal
and
then
things
like
Scupper
that,
like
eventually,
will
be
able
to
actually
talk
to
all
of
these
things
or
our
applications
will
be
able
to
talk
to
each
other
in
all
these
different
places.
A
B
A
Yeah,
but
just
the
Simplicity
of
it
right,
I
mean,
like
I,
think
I
think
that
there's
a
lot
lost
in
that
translation
when
somebody
goes
and
they're
like,
oh
I,
just
do
a
thing:
I
I
just
updated
a
machine
set
with
a
with
an
Ami
ID
and
a
node
type,
and
it
did
it
all
that
complexity
that
went
into
building
that
simple
click
right
or
that
simple
change
that
gets
lost
and
I
all
the
engineering
effort.
That's
got
in
to
make
all
this
stuff
easy.
It's
Bravo!
You
know
it's
really
amazing.
B
So
that
is
it
Johnny
I,
don't
know
if
there's
any
more
questions,
I
didn't
see
any
squeeze
past,
but
you've
been
good
at
Eagle
Eye
to
point
them
out,
but
I
didn't
want
to
keep
people
too
long
either.
I
know
it's
it's
very
tiring.
Looking
at
my
face
for
too
long,
in
fact,
my
wife
tells
me
that
regularly
so.
A
Yeah,
we
don't
have
any
more
questions.
I
do
want
to
you,
know
Jeff
or
Andy.
If
you
guys
want
a
chance
to
close
out
with
with
some
comments,
then
what
I'll
do
is
I'll
I
want
to
publicize
some
things
that
we've
got
going
on
on
the
on
the
other
side
of
this,
but
go
ahead.
If
you
want
to,
you
know,
make
some
closing
statements.
A
B
C
B
Andy
kind
of
hinted
out
before,
as
well
as
solving
some
problems,
we're
making
some
problems
at
the
same
time.
So,
as
you
become
very
flexible
about
where
you
go,
we
need
to
make
sure
that
openshift
still
has
all
the
Enterprise
qualities
that
it
needs.
So
we
need
to
test
those
environments,
and
you
know
if
we
had
more
time.
One
of
the
topics
we
could
get
into
was
a
discussion
we
had
internally
so
right
now
we
you
know,
we've
got
four
architectures
for
that.
B
We
can
run
openshift
on,
and
people
quite
rightly
started
asking
the
question:
do
we
have
to
test
everything
on
everything?
Every
architecture-
and
you
know,
there's
interesting
discussions.
Have
there
and
that's
kind
of
lends
itself
into
what
I've
shown
on
this.
This
final
slide
as
well,
and
that
you
know
once
we
start
opening
up
things
to
be
that
configurable,
we
need
to
make
sure
it
all
works.
So
this
is
a
vision
of
kind
of
where
we're
heading
to
at
some
point,
and
we've
got
a
lot
of
pieces
now.
B
You
know
ACM
ACS,
a
multi-cluster
aware
so
that
kind
of
fits
really
well
into
this,
but
the
whole
Span
in
the
different
kind
of
cloud
environments.
That's
still
something
that
you
know
you
could
kind
of
do
it
with
stretch
clusters,
clusters
and
things,
but
it's
something
we're
still
kind
of
filling
out.
Yeah
yeah.
A
D
Yeah
I
think
I'd
encourage
everyone
to
just
give
it
a
go.
If
you,
if
you
have
an
opportunity,
especially
on
on
AWS
it
is,
it
is
pretty
straightforward.
There's
no
one
I
think
one
of
the
coolest
things
about
it
is
there
there
is
no
real
magic
there.
You
know
we
haven't.
We
haven't
tried
to
make
things
like
overly
complicated,
it's
all,
just
as
as
straightforward
as
we
can
and
I
know,
Duncan
mentioned
it,
but
what
that
does
result
in.
D
D
You
know,
make
sure
try
some
some
workloads
try
building
stuff
on
on
from
multi,
multi-compute
cluster
and
and
if
there's
any
feedback
you
have
off
to
China.
Definitely
let
us
know.
We've
spoken
to
a
couple,
a
couple:
customers
that
have
started
using
and
have
had
some
really
good
feedback,
but
we're
always
looking
to
to
make
it
simpler
and
to
make
it
easier
for
for
people
to
use
and
kind
of
automate
up
those
bits
that
seem
a
little
bit
manual
and
tricky
but
yeah.
C
Maddie,
it
went
really
well
appreciate
it.
Yeah
I,
just
kind
of
extend
what
Andy
said
and
what
Duncan's
been
saying.
We
I
think
just
having
more
use
cases
and
and
more
people
doing.
This
will
help
us
maybe
steer
the
ship
in
the
right
direction.
I
think
right
now
we're
at
a
very
good
starting
point
right,
but
there
are
some
manual
things
that
I
don't
think
it's
clear
on
how
we
want
to
address
yet
scheduling
being
one
of
them
right
and
how
to
do
like
multi-arts
builds
and
a
cluster
things
like
that.
C
A
B
Right
did
you
wanna?
Do
you
want
to
add
anything?
No
I
just
wanted
to
thank
everyone
for
their
time
and
the
questions
and
the
chat.
It's
it's
been
really
good.
Just
emphasize
what
the
others
have
said
reach
out.
Let
us
know
how
you're
getting
on
try
it
out.
Tell
us
all
the
cool
things
that
we
haven't
thought
about,
that
need
do
and
that
that
would
be
great
yeah.
A
Definitely
definitely
and
to
jump
to
your
point
like
it's
baby
steps
right
like
we,
we
got
this
out
there,
we've
got:
we've
got
a
capability
now
now
we've
got
to
figure
out
like
okay.
How
do
I
tune
it?
How
do
I
help
it
grow
right
and
then
so
it's
just
getting
getting
these
little
cakes
worked
out
and
then
like
with
the
build
configs
and
stuff
like
that,
so
it'll
be
nice.
Yep.
A
A
So
if
you're
interested
in
openshift-
and
you
want
to
find
out
more
information-
and
you
know
you
just
want
to
ask
a
question
in
a
in
a
public
forum
where
there's
a
lot
of
people
looking
at
it-
then
the
openshift
commons
slack
channel
is
here
so
it's
openshiftcommons.sight.com,
so
you
go
ahead
and
register
I'm
on
that.
My
team
is
on
that
I'm,
not
sure
if
solely
is,
but
between
openshift
users
and
and
openshift
Commons,
we'll
be
able
to
help
get
your
your
stuff
answered
so
guys.
Thank
you.
A
A
Then
we
have
the
break
for
Independence
Day
here
in
the
states
and
then
I
think
on
the
12th
I'm,
not
exactly
we're
waiting
to
confirm.
But
we've
got
a
couple.
Things
lined
up
just
like
what
we're
talking
about
earlier,
so
we're
trying
to
get
service
mesh
we're
trying
to
get
Scupper
we're
working
with
a
couple
of
Partners,
as
well
as
the
red
hat,
open
data
science
team.