►
Description
Keynote: Kubernetes: Finally...A True Cloud Platform - Sam Ghods, Co-founder, Box
"Kubernetes is often perceived as ""just another"" container orchestration framework. But the Kubernetes API has its design rooted in more than 10 years of experience with Borg and was explicitly designed to not be specific to Kubernetes - instead, it's a generic way to describe your entire application infrastructure.
This is the first time that we have had a universal interface that we can build real deployment tooling against. We'll talk about what this means for managing applications in the cloud and what opportunities it presents to those who wish to build the future of cloud infrastructure."
About Sam Ghods
Sam Ghods is a cofounder and Services Architect at Box, where he is responsible for service infrastructure. Prior to his current role, Sam led software architecture at Box for eight years as Vice President of Technology. Previously, he was an engineer and cofounder at two startups.
Join us for KubeCon + CloudNativeCon in Barcelona May 20 - 23, Shanghai June 24 - 26, and San Diego November 18 - 21! Learn more at https://kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy and all of the other CNCF-hosted projects.
A
A
So
what
is
a
platform?
Well
I
looked
online
for
a
definition
like
the
merriam-webster
Seb's
kind
of
thing.
I
couldn't
find
what
I
liked
so
I
did
what
any
good
speaker
does
and
I
made
one
up
a
platform,
abstracts
away
a
messy
problem,
so
you
can
build
on
it.
So,
let's
go
over
a
few
more
popular
ones.
A
Linux
gives
you
an
abstraction
over
the
multitude
of
hardware
devices
that
exist
in
the
world,
giving
you
a
consistent
set
of
syscalls
that
you
can
build
against.
Java,
provides
a
high-level,
provides
high
level
abstractions
over
the
low-level
interfaces
of
individual
operating
systems
and
Twilio
takes
a
mishmash
of
operator
specific
API
interfaces
that
are
a
nightmare
to
deal
with
and
provides
a
clean,
consistent,
API
to
communicate
and
platforms
provide
a
lot
of
value
to
the
world,
and
they
do
that
in
very
two.
A
Very
specific
ways:
by
enabling
portability
across
a
wide
variety
of
environments
and
interfaces
and
providing
stable
ap
is
to
enable
extensibility.
So,
let's
take
Linux
as
an
example.
Well,
Linux
runs
on
nearly
all
Hardware
today,
so
developers
don't
have
to
worry
about
directly
supporting
different
network
cards
or
disk
drives
or
file
systems.
Instead,
they
can
focus
on
writing
and
delivering
their
application
code,
allowing
Linux
to
abstract
the
way.
The
messy
problem
of
all
these
different
devices
and
in
terms
of
extensibility,
the
Linux
kernel,
provides
stable,
API
s
and
interfaces
that
enable
a
world
of
tooling.
A
A
A
It's
not
good
guys,
not
good
its
fact.
It's
really
bad
and
while
that
may
be
surprising
to
you,
there's
a
reason
that
it's
so
bad
is.
If
you
objectively
evaluate
cloud
infrastructure
as
a
platform
in
those
two
dimensions,
it
doesn't
do
very
well.
So
let's
do
that
now
and
to
do
that
we're
going
to
put
these
in
a
matrix,
because
how
can
you
evaluate
anything
if
it's
not
a
matrix?
Let's
be
honest
and
let's
start
with
portability.
A
So
let's
say
in
that
box
and
I'm
a
developing,
a
microservice
in
Python,
so
I
go
ahead
and
package
it
up
in
a
docker
image
and
I'm
ready
to
ship
it,
but
to
run
it
not
only
do
you
know,
I
need
to
run
this
image.
I
need
also
a
Postgres
database,
tabacco
and
ideally
I'd
have
a
persistent,
remote
disk
as
well.
A
Now
I'd
also
want
an
auto
scaling
group,
so
my
service
doesn't
get
overwhelmed
and
then
I
do
the
load
balancer
in
front
of
that
auto
scaling,
group
and
I
would
need
service
discovery
for
all
these
components
to
be
able
to
find
each
other
so
to
run
my
application
properly.
This
is
the
kind
of
operational
setup
I
need.
A
A
We
have
the
feature
set
required
by
my
application
that
I
just
drew
up
by
the
micro
service
that
we
have.
So
if
you
want
auto-scaling
to
be
implemented
in
your
application,
each
of
these
different
providers
has
a
different
way
of
looking
at
it.
In
Amazon
you
have
an
auto
scaling
group
and
Google.
You
have
a
dedicated
resource
called
an
auto
scaler
in
Azure.
You
have
a
scale
set
and
OpenStack.
You
have
a
heat
scaling
policy
and
in
bare
metal
you
really
have
no
option
at
all.
So
it's
not
just
that
these
are
named
differently.
A
There
are
actually
fundamentally
different
concepts
that
have
different
nuances
and
api's
and
semantics
to
the
point
that
it's
very
difficult
to
create
an
abstraction
over
them.
Next,
let's
look
at
load
balancing
similar
story,
elastic
load,
balancer
on
Amazon
Google
has
a
load
balancer
as
your
as
a
different
one.
Openstack
is
load
balancers
a
service
and
again
I'm
bare
metal.
We
really
don't
have
a
good
option
and
finally,
with
remote
storage,
again
some
more
story
and
what
about
service
discovery,
so
my
application
can
find
the
database
that's
backing
it
well.
A
So
you
might
be
saying:
well,
ok,
you're,
just
looking
at
the
wrong
layer,
you
need
to
go
up
a
layer.
You
can't
just
look
at
the
infrastructure
providers
and
we
did
that.
So
we
looked
at
a
lot
of
different
options.
Frequently
these
are
called
platforms
as
a
service,
but
the
problem
with
these
is
that
their
primary
goal
has
always
been
to
provide
functionality
first
and
be
an
abstraction
layer.
A
A
A
We
could
build
against
a
cloud
platform
so
for
the
first
one,
let's
imagine
a
tool
that
could
show
you
in
your
entire
cluster,
all
the
different
applications
that
are
running
the
resources
that
they're
using
up
to
update
it
up
to
the
second
and
whether
it's
CPU
a
memory
or
network
and
have
it
visually
displayed
and
sorted
and
organized.
So
you
can
easily
see
at
any
given
time
what's
going
on
in
your
infrastructure,
regardless
of
where
your
infrastructure
is
running
or
what
cloud
infrastructure
your
provider
you're
using.
A
So
let's
call
that
visualizer
there's
a
reason
I'm,
not
in
marketing.
So
next,
let's
imagine
a
tool
that
automatically
balances
instances
of
your
application
across
many
different
cloud
providers.
So
let's
say
we
want
public
cloud
when
prices
are
low,
but
you
want
to
switch
over
to
your
bare
metal
data
centers
when
you're
not
able
to
get
enough
spa
instances
or
depending
on
different
fluctuations
and
market
or
different
traffic
patterns.
You
want
to
switch
between
different
ways
of
balancing
your
application
and,
where
you're
predominantly
running
it
all
without
any
kind
of
manual
intervention.
A
So
you
don't
get
the
advantage
of
the
largest
organizations
who
typically
have
their
own
bare
metal
data,
centers
or
their
own
infrastructures.
You
don't
get
to
leverage
their
engineering
resources
to
help
enable
and
build
tooling
for
the
greater
community.
So
if
we
go
back
and
we
rate
extensibility,
we
get
full
thumbs
down.
A
A
Okay,
why
kubernetes?
Why
well
number
one
it's
optimized
for
all
your
infrastructures?
Okay,
it
runs
on
public
cloud,
private
cloud
bare
metal,
the
five
raspberry
PI's
you
have
in
your
garage
that
are
serving
your
lossless
discography
of
the
Beatles
I,
don't
know
whatever
you
have
kubernetes
will
run
on
it.
It
will
run
amazingly
well
and
it's
optimized
for
all
those
environments
for
all
applications.
So
it
doesn't
matter
if
it's
Linux
or
Windows
doesn't
matter.
If
it's
stateless
or
stateful
guru
Nettie's
is
on
its
way
to
supporting
every
kind
of
application.
You
can
man
it
imagine.
A
If
it's
a
binary
and
it
runs,
it
can
run
in
kubernetes
a
known
and
stable
API
that
you
can
build
incredible,
tooling,
against
that
abstract
away
that
really
messy
problem
of
all
those
infrastructure
providers
and
their
implementation,
details
and,
finally,
and
by
far
I
think
this
is
the
most
important
point.
It's
all
the
folks
in
this
room
right
here.
A
It's
the
unparalleled
team
and
community
that
is
working
on
this
project
and
if
anyone
whatever
had
somebody
asked
me
why
not
mais
au,
s--
or
marathon
or
swarm
or
tried
nor
titus,
it's
it's
because
I
think
this
team
and
the
community
that's
around.
This
project
is
incredible
and
one
of
the
most
incredible
ever
assembled
in
the
open
source
world
yeah
round
of
applause
for
everyone
here
and
and
to
be
specific
for
a
moment.
A
For
example,
you
have
the
folks
from
Google
coming
in
bringing
over
a
decade
of
containerization
experience,
you're
the
folks
from
Red
Hat,
making
sure
kubernetes
runs
super
well
on
every
single
enterprise
environment,
even
the
most
constrained,
and
you
have
people
from
core
OS
working
for
on
making
sure
that
or
basically
making
kubernetes
the
only
sane
way
to
run
infrastructure
and
bare
metal
and
everyone
else
in
this
room
who
are
using
criminales
and
contributing
and
discussing
on
the
mailing
lists
and
issues.
It's
really
incredible
to
watch
and
I've
never
seen
a
community.
A
So,
let's
be
honest
about
this,
though,
and
let's
evaluate
kubernetes
now
on
these
two
dimensions.
So,
let's
start
with
portability.
Well,
if
you
remember
our
micro
service
earlier
every
one
of
those
features,
the
sets
that
I
needed
and
the
cloud
infrastructure
is
represented
by
a
kubernetes
native
object
that
is
portable
across
all
the
different
cloud
providers.
A
And
so,
if
you
remember
this,
really
messy
matrix,
what's
going
to
happen
with
kubernetes
as
your
provider
is
all
that
is
going
to
shift
up
there
and
we're
going
to
be
able
to
abstract
away
all
the
providers
below.
We
finally
have
a
portable
abstraction
to
work
against
in
cloud
infrastructure.
So
what
do
we
do
thumbs
up
so
now?
Extensibility
I
just
want
to
pause
for
a
second
and
say
this
is
what
I'm
most
excited
about
for
the
future
of
kubernetes.
A
So
if
you're
only
going
to
get
one
thing
from
this
talk.
Is
this
next
piece
here?
So
you
remember
all
those
tools
we
imagined.
That
will
be
amazing
if
everybody
that
runs
on
the
cloud
could
leverage
and
invest
in
well
they're
all
projects
today
that
are
either
in
development
or
just
released
for
kubernetes
as
a
cloud
platform.
So
the
first
one
is
dashboard,
the
kubernetes
UI
that
is
able
to
show
you
all
the
resource
utilization
inside
your
cluster.
A
The
next
one
is
cluster
Federation,
which
has
an
amazing
team
working
on
it,
which
will
be
able
to
automatically
balance
your
application
across
different
anything
from
public
cloud
providers
to
private
cloud
and
even
to
bare
metal.
And
if
some
of
these
other
announcement
recently
about
Etsy
the
operator
that
core
OS
released
I
think
it's
honestly,
one
of
the
most
exciting
things
I've
seen
in
cloud
infrastructure,
because
it's
a
never-before-seen
level
of
automation
and
intelligence
around
your
infrastructure,
that's
again
portable
and
agnostic
to
every
provider,
in
every
way
that
you
can
run
your
infrastructure.
A
A
A
I
think
the
amount
of
innovation
and
leverage
that's
going
to
come
from
being
able
to
standardize
on
criminalities
as
a
platform
is
incredibly
exciting,
more
exciting
than
anything
I've
seen
in
the
last
10
years
of
working
on
the
cloud,
and
we
have
an
opportunity
here
in
this
room
to
do
what
AWS
did
for
infrastructure,
but
this
time
in
an
open,
Universal
community-driven
way.
We
can
build
tooling
that
people
today
only
dreamt
of
having
and
truly
uplevel
the
next
generation
of
developers.
Thank
you.