►
From YouTube: Panel Discussion: Going Multi-Cloud for Realz. Stories From Media, Retail, & Infrastructure
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
Panel Discussion: Going Multi-Cloud for Realz. Stories From Media, Retail, & Infrastructure - Lisa-Marie Namphy, Portworx; Eric Han, Portworx; Umair Mufti, DreamWorks Animation; Allan Naim, Google; and Joseph Sandoval, Adobe
Multi-cloud is a long-term strategy for how enterprises source infrastructure that also informs a practical approach for running across disjoint, hybrid environments. Being able to move deliver entire applications into any infrastructure requires both a mind-shift in operations and new set of tooling. It can be hard for teams to decide how to approach multi-cloud workflows, and operationalize, and push their infrastructure providers as organizations move to a multi-cloud world. This panel brings together practitioners and infrastructure providers who have built this multi-cloud world and will offer their perspective. As a panel, we will cover: --How multi-cloud changes the development process --What gains can teams expect in the early days of multi-cloud (and how to achieve them) --What kinds of workloads are being moved --What are the best successes and worst horror stories
https://sched.co/MPY9
A
Ok,
it's
2
o'clock
come
on
in
lots
of
seats.
Lots
up
front,
don't
be
afraid.
I
understand
that
on
the
app
it
said
that
this
talk
was
in
room
D
1,
so
that
also
possibly
means
that
something
else
was
advertised
as
being
in
this
room.
So
if
you're
not
in
here
for
the
multi-cloud
for
reals
talk,
we've
already
shut
the
door.
You
can't
leave
so
you're
in
for
a
treat
we've.
We've
got
a
really
really
fun.
Awesome
group
of
experts
assembled
for
you,
so
we
won't
disappoint.
A
You
will,
depending
on
what
the
talk
you
were
coming
into
here
was.
If
it
was
some
sort
of
you
know,
lion
taming
act,
we'll
try
to
work
that
in
but
no
guarantees,
but
anyway,
we'll
go
ahead
and
get
started
so
keep
on
coming
up.
Don't
worry,
ok,
I
will
let
my
panelists
introduce
themselves,
but
I.
Just
will
start
out
by
saying.
I
am
really
excited
to
be
on
stage
with
this
particular
group
of
people.
A
I
always
love
choosing
panels
with
an
incredibly
diverse,
skillset,
diverse
way
of
thinking,
diverse
jobs,
and
so
we've
got
a
little
bit
of
everything
up
here
for
you
and
and
also
really
really
good
people.
So
it's
a
pleasure
to
share
a
stage
with
Mayor,
Joseph,
Eric
and
Alan.
This
is
us
on
Twitter
I
want
to
encourage
everybody
to
keep
the
conversation
going.
If
you
don't
get
a
question
in
for
us
today,
I'll
try
to
leave
a
little
room
at
the
end
for
questions,
but
I
don't
see
microphones,
so
I'm,
not
sure
how
that
would
work.
A
But
if
you
don't
get
your
question
today,
you
can
ask
us
at
any
time
we
will
answer
it
since
I,
don't
see
microphones
if
you
I
have
turned
my
DMS
on
courageously
don't
spam
me.
But
if
you
want
to
ask
a
question,
you
can
DM
me
on
Twitter.
They
should
be
open
and
available
to
everyone
and
I'll
go
ahead
and
do
my
best
to
get
your
question
in
and
if
you
want
to
be
anonymous,
you
can
not
mention
who
you
are
or
tell
me
not
to
mention
who
you
are.
A
B
A
I
usually
like
to
introduce
my
panelists
because
they're
all
so
modest
and
they
never
say
they
never
brag
about
themselves.
I've
worked
with
many
of
them
for
many
years
and
they're
so
talented
relevant
to
this
crowd.
Eric
Hahn
was
the
first
product
manager
on
kubernetes
when
he
was
at
Google
he's
the
co-creator
of
gke
so
just
had
to
brag
about
him
a
little
bit
and
I
have
the
pleasure
of
working
with
him
daily,
which
is
awesome
for
me.
A
Okay,
so
I'm,
oh
I,
didn't
understand
so
I'm
Lisa
I
was
asked
to
be
part
of
this
panel,
because
I
run
the
world's
largest,
the
ncf
user
group
and
I've
been
orchestrating
and
architecting
communities
and
user
groups
for
many
many
years,
and
so
I've
done
gosh
over
25
meetups
on
kubernetes
and
still
at
surface
mentioned,
so
many
topics
that
are
relevant
to
our
ecosystem,
I'm,
very
passionate
about
this
community
and
this
technology.
So
it's
a
great
pleasure
to
be
here
and
to
meet
so
many
of
you
in
person.
Our
meetup
is
also
online.
A
So
usually
so
you
can
check
us
out
if
you
don't
happen
to
live
in
the
San
Francisco
area
and
can
come
visit.
Us
live
okay,
so
I'm
gonna
start
with
a
question
for
America,
because
America
DreamWorks-
and
this
was
the
intro
of
a
Forbes
article
and
actually
before
before
we
asked
a
mirror.
This
article
talked
about
DreamWorks
going
to
multi-cloud,
but
before
that,
as
you
can
see,
multi-cloud
is
one
of
those
words
buzzwords.
A
You
can
say
that
may
be
replacing
or
being
added
to
a
lot
of
these
other
cloudy
words
like
public/private,
hybrid
cloud,
there's
multi,
cluster,
there's,
hybrid
multi
cloud,
maybe
so
Eric.
We
should
take
a
minute
and
talk
about
how
we're
gonna
view
multi-cloud
for
the
purpose
of
our
discussion
today,
yeah.
E
A
Okay,
so
I
will
thank
you
Eric,
so
now
I'm
asking
written
I'm.
Sorry
I
had
put
the
wrong
date
on
here
at
auto-filled
into
today's
date,
but
this
article
was
actually
in
Forbes
last
year.
I'll
fix
that
before
we
resubmit
these
slides,
and
so
maybe,
if
you
can
talk
about
a
little
bit,
dreamworks
plans
to
get
multi-cloud
and
really,
where
you're,
seeing
the
value
and
why
you
would
want
to
do
this.
Yeah.
C
Sure
so
for
us
multi-cloud,
really
just
it's
all
about
getting
the
finding
the
best
talent
around
the
world
for
us
in
the
media
industry.
That
really
means
artists.
We
make
cartoons
we're
a
movie
company,
so
we
need
to
go
out
and
find
the
better
best
artists
wherever
they
live.
We
have.
No,
you
know
we're
under
no
and
under
no
illusion
that
all
the
best
artists
are
in
Los
Angeles,
where
we're
located.
So
we
really
wanted
to
go
out
and
and
find
the
artists
where
they
live,
that
might
be
in
Omaha
that
might
be
in
London.
C
Wherever
happens,
to
be
the
thing
that
complicates
matters
for
us
all,
the
tooling
that
our
artists
use
is
all
custom
tooling,
all
based
on
a
service-oriented
architecture.
So
we
throw
up
services
currently
in
our
own
data
center
and
and
in
the
cloud,
so
what
we
need
to
do
and
we,
if
we
were
to
bring
out
a
new
artist
located
again
in
Omaha
or
London.
It's
basically
stand
up
an
entire
service
stack
as
close
as
possible
to
where
that
artist
lives,
so
that
that
artists,
user
experience
can
be
as
performed
as
possible.
C
It
wouldn't
make
sense
for
all
the
servers
to
come
back
and
hit
data
centers
in
Los
Angeles,
and
so
that
cloud
might
be
Google.
That
cloud
might
be
Amazon
if
it's
in
Los
Angeles.
That
cloud
might
even
be
our
own
data
center,
but
all
those
all
the
all
those
artists
working
around
the
world
only
to
collaborate
on
the
same
movie
and
have
the
same
experience
using
the
same
data
using
the
same
tooling,
and
so
we
kind
of
have
to
sort
of
virtualize
that
entire
pipeline,
so
they're
all
working
on
the
same
same
set
of
data.
B
So
for
the
Adobe
advertising
cloud
kind
of
you
know
similar
motivations
for
us
in
regards
to
like
customer
reach
and
part
of
it
is
like
it's
almost
like
accidental
and
in
a
certain
sense
we
have
a
platform
that
is
highly
transactional.
We
have
low
latency
requirements,
so
we
do
have
like
six
data
centers
around
the
world,
but
we
want
to
use
public
cloud
as
a
way
to
reach
not
so
much
in
the
sense
of
where
we
have
like
edge
use
cases.
B
But
when
we
have
like
emerging
markets,
it's
it's
so
much
more
easier
for
us
to
be
able
to
spend
something
up.
You
know
the
region
that
will
serve
those
customers
and
we
can
bet
out
whether
that
that
region
is
gonna
support
us
in
there
and
the
thing
about.
That,
though,
is
you
know,
there's
challenges
with
that,
because
and
that's
what's
happening
to
to
my
team
over.
We
started
with
the
infrastructure
as
a
service
and
those
data
centers,
so
we
run
in
a
cloud-like
manner.
B
There
then
we're
spinning
up
resources
and
AWS,
and
then
we
also
have
assets
that
are
in
Azure,
and
so
you
can
imagine
like
it's
just
it
gets
complex.
I
have
a
very
lean
organization,
and
so
you
can
imagine
the
complexities,
the
challenges
you
know
like
what
what
is
running,
where
at
any
one
time
you
know.
Obviously
we
have
concerns
with
like
security
and
observability
of
like
what's
happening
at
any
moment
of
time
and
in
being
able
to
scale,
and
so
with
the
emergence
of
kubernetes.
B
For
us,
you
know,
being
that
it
has
now
become
the
platform
for
our
developers
and
for
us
to
be
able
to
deliver
our
services.
What
I
want
now
for
multi-cloud
is
the
same
thing.
I
want
things
to
be
run
in
a
decorative
manner.
I
want
to
be
able
to
run.
You
know,
run
workloads
on
an
abstraction
that
allows
me
to
get
the
benefits
of
these
cloud
providers
and
be
able
to
reach
our
customers
wherever
they
may
be.
At
okay,.
A
D
Kubernetes
is
making
multiclad
a
reality.
You
know
we
hear
from
customers.
Typically,
they
don't
want
lock-in
right.
You
want
to
be
able
to
run
wherever
they
want.
If
you
want
the
flexibility
to
move
fast,
but
not
lock
into
particular
cloud,
and
we've
heard
that
loud
and
clear-
and
you
know
one
of
the
reasons
why
we're
really
pushing
kubernetes
so
that
customers
can
run
it
not
only
on
Google
cloud
but
run
it
anyway
on
other
clouds,
and
it's
not
just
about
getting
kubernetes
up
and
running
it's
what
happens
after
it's
up
and
running.
D
D
So
our
idea
with
multi
cloud
is:
let
the
customer
pick
what
cloud
makes
most
choice,
best
choice
for
those
applications
and
and
do
it
in
a
way
where
it's
consistent
from
both
the
developer
and
operations
experience
and
that's
really
how
we're
driving
and
thinking
about
multi
cloud.
Finally,
on
the
isp
side
of
the
house,
which
is
different
from
the
enterprise,
you
know
these
software
vendors
basically
want
to
build
their
software
once
like
right
at
once,
develop
it
once,
but
then
run
it
anywhere
and
their
customers
could
be
on.
A
E
I
think
when
we
think
about
kubernetes
and
the
abstractions,
it's
also
important
to
think
that
there's
no
such
thing
as
a
stateless
application
right.
So
it's
good
to
have
abstractions
for
compute,
but
I
don't
have
compute
less
applications,
so
I've
extracted
the
application.
I've
made
it
portable,
but
you're
persisting
something
you're
persisting
data
and
if
we're
talking
about
real
workloads,
they're
followed
the
Sun
they're
machine
learning.
There
are
different
data
services
that
have
to
be
delivered
around
the
world.
E
There
needs
to
be
an
answer
for
how
compute
and
the
data
meet
and
that
abstraction
is
super
powerful,
but
I
need
a
solution
that
brings
that
application
and
the
data
together.
So
if
I
really
want
to
be
multi
cloud,
if
I
really
want
to
be
able
to
deliver
around
the
world
and
do
these
interesting,
real
applications,
I
have
to
solve
the
data
gravity
problem
or
we
have
to
solve
it
as
a
community.
So
we
have
to
be
able
to
bring
the
applications.
E
The
data
together
deliver
it
on
any
infrastructure
and
actually
for
someone
like
us
at
port,
where
it's
what
our
customers
expect
right:
you're
adopting
kubernetes,
because
you're
reap
lat
forming
or
you're
competing
in
some
digital
advantage.
So
you
need
to
be
faster
into
production,
but
you
also
need
to
be
resilient
and
mobile.
So
that's
where
our
customers
are
asking
us
make
sure
that
we're
multi-cloud
make
sure
that
we're
able
to
run
in
public
clouds
on
prem
and
do
it
in
any
environment
seamlessly.
E
A
And
my
next
question
is
going
to
go
to
anybody,
but
Eric
and
Alan,
because
I
want
to
know
for
reals
Eric
actually
named
the
panel
I'm,
pretty
sure
Eric's
three-year-old
son
named
the
panel
multi-cloud
for
reals
busy
so
but
for
real.
Does
kubernetes
make
multi-cloud
easier
and
do
you
need
kubernetes
to
do
this
as
opposed
to
nothing.
B
Like
you
will
III
think
it
does,
and
mainly
how
I'm
looking
at
it
is
just
like
the
the
evolution
that
we're
seeing
of
okay.
Yes,
you
know
the
initial
we're
COO
Brandi
started
at
one
and
the
use
cases
that
initially
served
up.
You
know
it
kind
of
laid
that
groundwork
for
it,
but
then
with
the
emergence
now,
what
we're
seeing
that
we
can
start
running
all
types
of
workloads
and
running
them
in
a
kubernetes
way.
I
think
that's
where
the
real
power
comes
in
at
now.
B
The
concerns
I
have
and
the
things
I'm
starting
to
see
because
I
think
a
lot
of
us
here
have
you
got
the
conference
you're,
seeing
a
lot
of
different
vendors
starting
to
you
know,
position
different
ways
to
accomplish
this.
We
see
things
like
cross
plane
coming
out
with
ways
to
have
like
a
universal
control
plane
and
when
I
look
at
a
lot
of
what
that
architecture
looks
like
I
mean
I,
it's
some
of
it
speaks
to
me,
but
my
concern
is
just
like
one.
B
B
Have
a
lot
of
us
who
are
operators
will
acknowledge
that
you
know
the
experience
above
kubernetes,
it's
all
cloud
native
and
it's
great,
but
if
you
get
into
the
lifecycle
management
of
like
what's
happening
underneath
well
a
lot
of
times
we're
working
around
a
lot
of
that
part
of
it
to
be
able
to.
You
know,
maintain
upgrades
whether
it's
like
security
patching
every
four
six
weeks
or
you
know
being
able
to
know
that
you
know
like
the
securities
coming
to
you.
B
Okay,
can
you
patch
this
quickly
and
then
imagine
now
you're
trying
to
do
this
across
a
bunch
of
different
cloud
providers?
And
so
that's
where,
like
the
evolution
of
cube
already
kubernetes
and
you're,
seeing
things
like
you
know,
cube
fed,
evolve
cluster
API
a
lot
of
these
things
being
as
possibilities
to
be
able
to
help
you
address
these
challenges,
so
I
think
I
believe
that
this
is
gonna,
be
the
route,
but
it's
gonna
be
from
for
us,
multi-cloud
I.
B
Think
for
a
lot
of
other
users,
but
I
do
think
that
there's
still
a
lot,
especially
with
industries
that
are
have
heavy
regulations
or
enterprises,
because
we
do
get
a
lot
of
legacy
industries
or,
like
I,
say
things
we
need
to
plug
in,
and
so
that's
where
some
of
the
challenges
come
in
up.
But
I
do
like
you
know
the
controller's.
The
CRT
is
all
everything
that's
within
it.
I
think
it's
gonna
be
able
to
provide
what
we
need
and.
A
C
I
kind
of
wanted
to
expand
on
that
a
bit
so
even
even
taking
multi-cloud
out
of
it.
So
for
us
we
were
early
adopters
of
containerization
I
mean
we.
We
started
container
izing
our
workloads
even
before
doctors,
so
we
were
LXC
people.
You
know
my
first
project
when
I
started
at
DreamWorks
five
and
a
half
years
ago
was
to
get
Cassandra
to
run
inside
of
docker,
dr
0.7.
C
This
was
before
kubernetes
right
and
so
as
far
as
enablement
goes
doing,
what
we're
trying
to
do,
or
what
we
have
been
able
to
do
just
in
in
a
local
data
center,
was
almost
impossible
without
kubernetes
right.
So
you
know
running
running
1200,
docker
containers
across
15,
different
database
application
types,
persisting
the
data
taking
the
backups
and
doing
all
that,
all
that
workload
that
we
have
having
to
worry
about,
IP
addresses
and
dns
and
service
discovery
is
is
onerous
right.
C
You
know
that
entire
stack
is
it's
very
valuable
for
us
and
then
have
the
underlying
abstractions,
like
port
works,
where
I
don't
care,
if
it's
a
persistent
disk
and
in
in
Google
or
if
it's
you
know,
PBS
or
if
it's
a
local
disk
in
my
local
data
center.
That
allows
me
to
basically
deploy
the
same.
You
know,
resources
anywhere,
so
yeah
absolutely
kubernetes
is.
Is.
A
Swear:
okay-
that
was
the
unbiased
answer
to
my
question.
So
now
we'll
get
the
chance
for
the
people
who
have
helped
build
this
fabulous
technology.
To
add
to
that,
is
there
anything
you
want
to
add
about
what
kubernetes
gets
right
to
really
enable
multi
cloud
I'll
direct
that
to
Alan
and
Eric
I?
Think
one.
D
Thing
that
kubernetes
does
really
well
is
to
API
I
think
that
the
API
is
just
been
designed
in
such
a
way
that
lends
itself
to
you
know,
remove
some
of
the
friction
associated
with
the
underlying
infrastructure,
especially
when
you're
running
in
environments
that
are
very
heterogeneous.
You
know
the
one
area
I
think
where
there's
still
room
for
improvement,
is
you
know,
as
you
think,
about
multi
cloud
and
think
about
the
different
cloud
providers
is
still
a
specialized
skill
set
right.
Typically,
you
know
in
many
of
these
organizations
larger
organizations
where
they
are
multi
cloud.
E
E
That's
super
powerful
and
I
think
the
abstractions
have
evolved
like
when
we
started
there
were
pods,
and
now
we
have
daemon
sets
and
stateful
sets
deployment
objects,
but
we
also
see
to
the
concern
that
Joseph
you're
bringing
up
I
agree
like
there's
now
some
splitting
that
we're
seeing
what
runs
in
kubernetes
this
distro
is
being
adapted
in
a
different
district
and
I.
Don't
know
that
that's
toppings
and
something
that
you
you've
brought
up.
That
I
also
think
about
is
everything
running
in
kubernetes
is
necessary.
E
It's
not
sufficient,
though
every
application
to
the
examples
Allen
and
numeira,
giving
they
have
to
connect
into
some
identity
system
and
so
identity
access
management.
What
I
do
about
the
whole
entire
tooling
around?
It
is
not
necessarily
portable
or
not
necessarily
abstracted
or
should
I
abstract
it
in
some
ways.
So
I
feel
like
it's
a
problem
that
we'll
have
to
figure
out
as
a
community
right
kubernetes
is
the
necessary
ingredient.
It's
so
much
better
than
anything
we
had
before,
but
to
get
to
the
multi
cloud
world
for
real
applications
and
for
all
applications.
E
There
has
to
be
something
that
we
come
together
at
and
the
concerns
are
the
way
the
ecosystem
is
splitting
a
little
bit.
The
conformance
tests
won't
apply
the
same
to
extended
objects
that
makes
sense
they're
extended,
but
if
all
the
core
components
are
moving
to
extended
objects,
now,
I
have
a
problem
with
saying
this
app
runs
across
so
I
feel
like
we
have
to
figure
that
out
as
how
do
we
make
the
meaningful
parts
core
again
and
how
do
we
bring
it
so
that
it
does
solve
the
the
real
problem
of?
E
A
Okay,
so
that
wasn't
even
a
third
of
the
audience
and
that's
probably
less
than
we
thought,
although
I've
been
crowdsourcing
that
question
throughout
the
whole
conference
and
a
lot
of
the
folks
particular
from
Europe
are
saying
it's
about
two
years
out:
we're
hybrid
now
it's
about
two
years
out,
but
we're
getting
there.
So
that
brings
up
my
question:
should
everyone
be
planning
multi-cloud
and
should
every
application,
stateless,
stateful,
every
application
be
thrown
into
a
multiplied
environment?
D
Start
I
say
no
I
say
there
are
applications
to
doesn't
make
sense
to
move
to
the
cloud
and
should
remain
on
Prem
just
but
but
at
the
same
time
you
want
to
get
as
much
value
from
those
applications
as
possible.
So
just
because
an
application
doesn't
move
to
the
cloud
doesn't
mean
that
it's
like
wasted
effort
right.
There's
we're
seeing
a
tremendous
amount
of
investment
and
modernization
in
modernization
of
applications
that
are
running
in
people's
data.
Centers
they're,
not
they
don't
need
to
be
written
off.
D
They
could
join
the
club,
I
think
we're
in
an
environment
right
now
we're
it's.
It's
really
the
flexibility
that
kubernetes
provides
some
enable
you
to
make
that
choice
in
terms
of
not
having
to
lift
and
shift
to
gain
the
values
of
cloud
native.
You
can
start
where
you
are
and
take
advantage
of
it
and
that's
yeah.
C
I
think
it's
the
same
question
that
developers
have
been
asking
themselves
for
years,
which
is:
do
I
optimize
something?
Or
do
I
abstract
it
right
in?
In
many
cases
there
are
good
arguments
for
you
to
take
those
optimizations
for
a
specific
cloud
provider
and
others
you
you
do
want
to
abstract
it.
If
you
can
gain
economies
of
scale
or
whatever
happens
to
be
so
yeah
I
think
there's
cases.
B
B
It
was
a
driver
to
also
bring
us
back
on
on
top
of
like
maintaining
like
latency
SLA,
some
things
such
as
that
so
I
think
you
really
have
to
do
the
due
diligence
on
it
and
just
challenge
yourself
to
see.
Is
it
worth
it,
and
especially
in
this
current
state,
because
you
will
be
adopting
until
some
of
these
tools
start
to
deliver
and
the
promise
of
what
we
are
looking
forward
to
you're
gonna
be
adopting
some
complexity
and
challenges,
and
you
have
to
really
weigh
that
out.
So.
E
Think
I
was
talking
to
a
telecom
operator
yesterday,
one
of
the
largest
in
Europe
and
from
their
perspective,
they're
still
trying
to
figure
out
how
to
hire
COBOL
programmers
for
their
mainframes
right,
I,
never
learned,
COBOL
and
I
went
to
school
in
the
90s.
So
what
I
mean
to
say
here
is
there
are
applications
that
no
don't
need
to
be
multi-cloud,
don't
need
to
be
containerized
per
se.
I.
Don't
want
to
touch
that
mainframe.
It
runs,
leave
it
alone,
but
all
new
applications
should
they
be
designed
for
multi
cloud.
E
Should
they
be
portable
to
kubernetes
and
I.
Think
this
is
where
I'll
try
to
disagree
and
not
steal
right,
I
think
they
should
be
right
if
I'm
gonna
run
this
application,
it's
the
future.
Why
design
it
with
the
baggage
of
what
I'm
trying
to
get
away
from
right?
We
don't
want
to
get
to
a
place
where
everything
we
did
only
lasted
us
for
a
couple
years
or
for
one
lifecycle
and
something
you
said
just
of
yesterday,
that
is
in
my
head.
E
All
day
was
really
that
we
you're
talking
about
the
challenge
of
standing
up
consistent,
kubernetes
clusters,
three
or
four
years
ago
we're
trying
to
solve
this
challenge
of
standing
up
consistent
applications.
So
the
problem
space
is
moved
along
and
that's
a
positive
thing,
but
if
three
years
from
now
we
can
solve
or
we
solved
consistent
applications.
Now
we
solve
consistent
kubernetes.
What's
next
right,
suddenly
we're
tied
into
only
one
if
structure
I
feel
like
we
locked
ourselves
in
yet
again,
so
no
to
the
COBOL
mainframes.
A
A
A
B
You
know
well
for
me,
like
I
said
we
die
I,
always
just
look
for
well
like
what's
best
in
breed
and
for
us
you
know
I
kind
of
feel.
Like
you
know,
especially
with
the
surface
of
like
what
my
team
is
covering
already.
You
know
so
I've
kind
of
felt
like
that
was
like
our
first
adoption.
It
was
just
like
look
I've
already
spending
all
my
innovation
tokens
in
the
data
center
I.
Don't
want
to
risk
it
too
much
further
by
just
trying
to
be
like.
B
Let
me
get
across
all
the
clouds
and
try
to
deliver,
but
I
mean
it.
So
for
me
it
was
more
like
something
that
we
knew
you
know,
but
previous
to
this
I
was
at
another
company
where
we
were
early,
kubernetes,
adopters
and-
and
this
is
like
you
know
one
row
and
I-
can
our
aspirations
at
that
company
at
that
time
was?
We
were
looking
for
multi
cloud
and
I
was
really
selling
an
internally
chart
developers,
and
even
at
that
time
it
was
very
nascent
to
like
take
that
view.
B
But
it's
where
I
wanted
to
see
things
and
then
I
agree.
Eric,
I
think
you're
right,
like
we're
looking
at
what
we
were
wrangling
and
managing
just
getting
like
you
know,
just
the
initial
micro
services
up
and
running
and
getting
them
stable
keeping
them.
You
know
the
life
cycle
going,
and
here
we
are
talking
about
like
trying
to
push
this
needle
a
little
further,
but
it's
a
little
bit
more
complex
and
I
do
find
what's
most
interesting
about
it
is.
B
It
seems
in
my
purview
that
I'm
noticing,
like
companies
like
port
works,
are
talking
about
it.
There's
another
storage
vendor
who
proposed
a
cube
move
as
a
way
to
like
data
migration
as
a
service
and
you're,
seeing
other
ones
coming
in
there
I'm
the
only
thing
I'm
just
kind
of
wondering
about
its.
Are
we
gonna
see
now
the
emergence
of
some?
E
That's
a
that's
a
hard
question:
I
see
companies
building,
CR,
DS
or
abstractions
for
deploying
to
different
cloud
environments.
You
brought
up
a
few
examples
and
I
think
those
are
interesting
and
powerful.
How
they
become
compatible
and
maintained.
I
mean
CSI
is
a
great
example.
A
lot
of
our
customers
have
evolved
in
terms
of
what
they
require
and
the
ecosystems
are
moving
so
fast.
So
one
reason
people
want
abstraction
layers
is
once
you
deploy
enterprise
applications
into
production.
E
You
don't
want
to
be
tweaking
every
few
months,
so
you
kind
of
need
that
ability
to
say
that
these
stacks
are
making
sure
that
everything
is
upgradeable
in
place
and
stable.
So
in
a
constantly
moving
ecosystem.
How
do
we
get
that
unification?
How
do
we
do
it
across
clouds?
There
should
be
some
community
effort,
I
think
it's
probably
right
now
getting
worse
and
so
I'll
have
to
ask
a
punt
on
that
and
come
back
with
an
answer
later.
Maybe
Alan
I'll
pass
it
to
you.
Maybe
you
have
a
thought.
D
See
I
agreed,
the
CRTs
are
becoming
the
de
facto
standard
for
abstracting
away.
If,
like
the
underlying
resources
and
complexities,
I
CCR
DS,
you
know
potentially
extending
beyond
just
kubernetes
resources,
but
I
feel,
like
you
know,
especially
on
on
the
storage
side,
as
stateful
workloads
become
more
mainstream.
There
is
gonna
need
to
be
some
kind
of
agreement
across
the
industry
right
because
storage
is
always
the
trickiest
area
to
tackle,
and
you
know,
and
it's
high-risk
high-reward
so
I
think
folks,
like
port
works,
are
very
well
positioned
to
be
leaders
and
helping
define
that
moving
forward.
D
A
Networking
and
security
are
also
tricky
things
to
tackle
and
I've
run
a
lot
of
meet
ups
on
those
when
I
went
to
meet
up
on
security
and
kubernetes,
it's
packed.
So
so
there's
still
a
few
problems,
we're
all
solving
remember
my
DMS
are
open.
So
if
any
of
you
have
a
question,
you
can
shoot
it
at
me
and
I
will
try
and
get
to
it,
and
I
will
ask
a
practical
question
now.
So,
basically,
if
you
want
to
get
started
on
multi-cloud,
what
should
be
the
first
step.
A
C
B
Yeah
we
say-
probably
you
know,
like
it's
we've
kind
of
learned
this
lesson
of
a
few
of
us
who
you
know
if
you've
been
in
the
Kuban
at
eternity
I
mean.
Obviously
we
we
saw
with
a
lot
of
the
initial
work
loads
started
from.
It
was
always
like
start
from
a
safe
place.
You
know
pick
those
things
that
you
know
are
not
gonna
risk
your
business,
which
of
this
that's
an
obvious
kind
of
easy
thing,
but
but
then
of
you
started
kind
of
trying
to
get.
B
You
know
practical
I
looked
at
like
in
my
world,
you
know
like
we
started
looking
at
like
scheduled.
Jobs
like
this
was
one
we
thought
like
hey.
We
can
actually
do
this,
and,
and
actually
it
was
an
easy
safe
way.
We
didn't
have
to
do
it
multi-college,
but
what
we
wanted
to
do
is
kind
of
tease
out,
like
all
the
things
that
we
may
have
been
not
seeing
you
know,
so
it
gave
us
a
chance
to
kind
of
like
well
I,
didn't
think
about
like
when
we
started
running
on
this
environment.
B
It
was
gonna
introduce
some
challenges
of
like
some
scheduling,
orchestrations
that
we
didn't
think
about,
or
even
just
kind
of
like
working
on
our
ability
to
manage
life
cycle.
You
know
across
two
different
clouds
we
had
to
like
rethink
what
we
are
currently
doing
so
for
me,
it's
just
I
think
it's
it's
the
best
way
to
kind
of
get
yourself
going.
E
I'll
say
that
we've
worked
with
customers
who
are
moving
major
retail
chains
to
Google
and
to
your
question
Lisa.
How
did
they
get
started?
I
think
if
you
want
to
go
multi-cloud,
something
that
they've
done
while
I'm
thinking,
one
in
particular
that's
moving
to
Google,
is
that
they
picked
a
meaningful
workload,
and
so
they
actually
picked
three
and
their
goal
is
to
correctly
architect.
It
have
that
workload
be
representative.
E
Think
I
don't
need
to
solve
that
problem.
I!
Don't
need
to
solve
state
I,
don't
need
to
solve
identity.
Suddenly
they
have
a
second
application
get
introduced
and
they
have
to
react,
protect
everything
right.
So
they
chose
the
wrong
example,
and
it's
always
easy
to
say
that
in
hindsight,
but
I
think
get
started,
pick
a
cloud
pick
an
application
make
it
meaningful
and
then
that's
a
much
better
chance.
Yeah.
C
A
D
So
I'm
last
copy,
but
I'll
take
a
different
approach
actually
in
the
sense
of
I,
wouldn't
treat
it
as
a
one-size-fits-all
I
think
it
depends
on
who
you
are
within
the
organization
right.
So
if
you're
a
developer
right,
yeah
yeah
roll
up
your
sleeves
and
get
into
the
cloud
native
technologies.
If
you're
a
platform
team
start
to
understand
like
what
are
the
benefits
that
each
of
the
cloud
providers
provides
and
what
are
the
applications
that
make
sense?
D
Just
last
thing
you
want
to
do
is
a
platform
team
is
build
out
an
environment
and
wait
for
people
to
come
right.
That's
not
the
right!
It's
it's
important
to
pick
those
applications
that
makes
and
then
build
that
process
show
success
and
then
others
will
follow.
If
you're
a
network
admin,
you
got
to
start
thinking
about
the
complexity
associated
with
multi
cloud
right.
How
do
you
enable
connectivity,
especially
if
you
have
a
hybrid?
If
you
have
your
data
center
and
now
you
have
to
you-
know
interconnect
with
the
different
clouds?
D
How
much
is
that
going
to
cost?
There's
a
lot
of
things
to
plan
right
if
you're,
a
security
person
right,
you're,
gonna,
you're,
gonna
see
an
environment
where
there's
cluster
sprawl
now
right.
Just
like
years
ago,
we
had
VM
sprawl,
you
know
as
more
and
more
clusters
come
up
in
different
environments.
Are
you
gonna,
secure
those
environments?
How
are
you
gonna
ensure
that
you
know
somebody
doesn't
come
along
and
change
things
that
are
outside
of
the
policies,
especially
if
you're
operating
in
the
regulated
environment?
D
So
just
to
close
on
it,
you
know
I
think
it
depends
on
who
you
are
within
an
organization
but
with
with
kubernetes
I.
Think
the
one
thing
that
I
find
that
really
helps
in
these
environments
is.
It
enables
all
these
roles
that
I
just
discussed
to
work
collaboratively
you're
no
longer
an
obstacle
to
each
other,
which
was
in
the
pre
cloud
native
world.
A
C
In
my
perspective,
a
lot
of
the
driving
factor
was
was
cost
in
the
long
run,
but
it
was
you
know.
As
an
example,
we
decided
we
needed
to
do
some
stuff
in
the
cloud
we
started
solving
for
x
provider
and
then
my
management
came
back
to
me
and
said
you
know
what
we've
been
able
to
negotiate
a
deal
with
Y
provider.
So
let's
look
at
them
instead
and
it
cost
us
again
to
throw
away
some
work
and
have
to
do
some
stuff.
C
A
Here
you
say
it's
a
way
to
get
cloud
fighters
to
negotiate
against
each
other
and
lower
the
price
wow.
That's
an
advantage.
Take
note
of
that
one.
You
know
I
think
we're
probably
running
close
to
time.
So
why
don't
we
just
wrap
up
if
there's
any
last
thoughts
that
anyone
wants
to
leave
with
and
I'm
gonna
start
with
that
end,
because
you
guys
are
always
ending
go
ahead.
Ag
I'll.
E
Say
that
isn't
been
incredible
to
see
the
ecosystem
evolved,
we're
all
solving
and
trying
to
make
the
world
better
right,
so
I've
been
in
customer
environments
or
conversations
where
they're
not
innovating,
they're,
not
doing
something
new
and
that's
very
different.
So
we're
up
here
talking
about
how
to
make
everything
better
for
the
whole
ecosystem.
So
I
think
this
is
the
best
place
to
go
solve
those
hard
problems.
D
C
B
Who
are
you
know
looking
towards
multi
cloud,
and
you
know
even
myself
what
I'm
trying
to
do
now
is
get
involved
in
these
conversations
now
you
know
get
let's
get
involved
in
that
community,
so
we
can
up
drive,
because
the
key
things
I
want
to
see
is
like
things
built,
you
know
an
open
standard
so
that
we
have
choice
in
it
and
I
think
that's
gonna
be
driven
by
us
in
the
community
by
having
our
voice
in
their
sand.
You
know,
even
for
myself
the
things
that
I'm
really
excited
about
is
right.
B
Now
we're
rethinking
even
how
we're
doing
disaster
recovery
with
this
kind
of
approach.
So
you
know
it's
in
a
cloud
native
way,
so
that
those
are
the
things
that
are
really
exciting
me
so,
like
I
said,
I
just
advocate
for
more
of
us
to
to
get
out
there
and
have
these
conversations,
so
we
can
help
shape
it
in
and
make
you
know,
kubernetes
be
this
that
help
us
drive
to
accomplish
this.
A
Thank
you
for
having
this
conversation
with
us.
This
is
how
to
find
us.
You
can
keep
the
conversation
going.
We
will
answer
you
and
thank
you
for
attending
this
session.
We
I
look
forward
to
getting
to
know
all
of
you
in
person,
so
you
can
find
us
on
the
show
floor
and
enjoy
the
rest
of
your
conference
thanks.
Everyone.