►
From YouTube: Kubernetes in 2015 - Craig Mcluckie
Description
Recorded on Feburary 25th, 2015 at the Kubernetes Gathering in San Francisco, CA, USA
A
I'm,
a
product
guy
at
Google,
I
think
I'm,
one
of
the
sort
of
longest
standing
folks
who
are
working
in
Google's
cloud.
Let
me
just
present
all
right.
Let
me
back
up
here,
and
so
what's
the
product
guy
from
Google
doing
here,
speaking
to
to
everyone
so
turns
out
at
Google
we
got
some
really
interesting
technical
infrastructure.
We
talked
about
a
little
bit,
we
write
papers
on
it
and
it's
pretty
cool.
It's
very
modern.
A
We
launch
a
lot
of
containers
like
we've
used
the
next
application
containers
extensively,
we're
about
2
billion
in
the
week
and
my
job
as
a
product
guy
is
to
take
that
awesome.
Google,
technical
infrastructure
and
bring
it
out
into
the
world
and
more
specifically,
my
not-so-secret
objective-
is
to
try
to
create
a
really
healthy
business
selling.
Then
awesome
technical
infrastructure,
the
world,
and
so
my
mission
has
been
to
bring
a
new
class
of
cloud
service
to
market
and
it
needs
to
be
modern.
It
needs
to
match.
A
This
is
a
pretty
new
thing
like
let's
create
a
cloud
service
and
instead
of
just
like
trying
to
build
a
service
around
some
existing
open
source
project,
let's
go
build
the
open
source
project
and
then
I'm
offer
up
a
cloud
service.
That's
that's
built
upon
it,
and
so
that's
that's
really.
What
kubernetes
is
a
bar
and
that's
what
we're
here
to
talk
about.
So,
let's
just
cover
it
in
a
couple
of
words
like
didn't:
just
get
a
show
of
hands.
How
many
folks
here
have
heard
about
kubernetes
or
no
broadly,
what
it
is?
A
Ok,
great,
so
I
don't
have
to
go
into
too
much
detail,
but,
as
I
mentioned,
Google's
big
into
Linux
application
containers.
So
if
you
actually
track
back
in
history,
I'm
not
going
to
go
into
all
the
details,
but
we
looked
at
Linux.
Application
containers
are
solving
a
really
gnarly
resource,
isolation
problem.
So
we
could.
A
We
could
like
package
things
very
carefully
and
run
a
lot
of
containers
on
the
same
physical
piece
of
infrastructure,
achieve
very
high
levels
of
utilization,
basically
use
every
piece
of
the
animal
and
that's
really
important
when
you're
running
a
business
of
Google
skill.
But
we
didn't
solve
a
pretty
important
problem
with
these
technologies,
even
though
we
did
open
sourced
them,
which
was
making
them
really
accessible
to
to
developers
to
engineers
along
came
darker.
Doctor
did
an
amazing
job
of
popularizing
these
technologies.
They
solved
a
really
gnarly
packaging
and
deployment
problem.
A
They
may
contain
as
accessible
to
engineers
and
they've
created
a
great
way.
To
put
your
code
into
a
container
and
then
run
that
container
somewhere,
but
neither
docker
and
all
Linux
applications
and
containers
themselves
had
actually
solved
some
of
the
problems
that
we
thought
were
really
important.
So
neither
created
a
fabric
or
an
environment
for
an
application
to
run
it.
Neither
we're
providing
a
set
of
common
services
that
that
container
package
technology
could
just
bet
on
independently
of
where
it
was
running
and
neither
was
doing
much
to
handle
a
lot
of
the
realities
of
application.
A
Lifecycle
management,
like
health
management,
doing
seamless,
updates
dealing
with
the
orchestration,
the
mechanics
of
keeping
an
application
running,
and
so
we
decided
to
do
kubernetes
to
bring
those
capabilities
to
to
market
and
it's
inspired
by
a
lot
of
the
technologies
we
used
inside
Google
projects,
codenamed
Borg
and
the
project
that
followed
called
Amiga.
We're
really
some
of
the
underpinnings.
And
if
you
actually
looked
around
the
room,
you
can
see
people
like
Brian
and
Tim,
and
some
of
the
other
guys
that
builds
out
those.
A
It
doesn't
only
systems
with
the
architects
of
those
systems,
and
so
we've
been
very
lucky
and
being
able
to
Conscription
pretty
expert
folks
to
actually
help
us
build
this
stuff,
and
so
kubernetes
creates
this
cluster
environment
for
containers.
It
solves
some
really
hard
problems
and
we
like
a
lot
and
then,
of
course,
there's
Google
container
engine,
which
is
our
hosted
version
of
communities
coming
out
of
the
gate.
It
just
solves
a
lot
of
the
operational
problems
running
the
cluster,
but
over
time
it
will
provide
a
better,
stronger,
faster
way
to
actually
run
these.
A
A
So
don't
how
many
of
you
have
seen
the
matrix
but
I
always
kind
of
wondered
about
the
choice
of
taking
that
red
pill
like
particularly
fiercest
atman
life's
hard
enough
as
it
is,
do
you
really
want
to
live
in
a
world
of
kind
of
crazy
infrastructure
and
all
of
those
sort
of
the
details
of
that?
What
a
cluster
environmental,
creating
this
cluster
first
way
of
thinking
about
application
management?
Is
it
gives
you
this
world,
where
you're
blissfully
abstracted
away
from
the
horrible
and
dark
realities
of
your
infrastructure
management?
Right?
It's
a
much
nicer
world.
A
You
don't
have
to
deal
with.
You
know
a
lot
of
a
lot
of
difficult
things,
and
so
kubernetes
naturally
brings
things
like
micro
service
architecture
to
to
developers
that
are
trying
to
run
applications.
It
allows
the
separation
and
specialization
of
operations.
That's
a
really
important
factor
like
there's
one
of
the
things
that's
made
Coug
already
powerful.
Is
we
have
a
small
set
of
really
intense,
really
dedicated
people
that
build
and
manage
the
underlying
services?
And
then
we
have
another
set
of
people
that
are
perhaps
less
intense.
A
You
know
running
applications
on
top
of
them
and
that
specialization
of
off
of
apps
is
good,
and
then
the
other
piece
is
just
being
able
to
bet
on
a
set
of
services.
I
want
to
monitor
this
I
want
to
log
it
I
want
to
connected
to
the
Internet
I
want
to
be
able
to.
You
know
wire
it
up
to
some
other
things
that
I
might
depend
on
creating
an
environment
that
offers
up
those
services
and
abstracts
away
where
they're
running
just
makes
them
seamlessly
available
to
your
containers
is
very
powerful.
A
A
If
you
actually
apply
modern
systems,
there's
things
that
machines
do
better
than
people,
one
of
those
things
that
machines
do
better
than
people
is
fly
the
airplanes
or
fly
little
drones
right
and
the
model
with
this
is
you
have
a
system,
that's
taking
inputs
and
making
it
a
very
informed
set
of
decisions
based
on
your
intent
right.
So
when
you,
when
you're
piloting
on
these
little
drones,
you
have
a
lie
pad
and
you
say:
hey
go
up,
I'm,
not
saying,
like
you
know,
increase
you
know,
thrust
on
these
things.
A
The
system
is
making
those
decisions
for
you,
and
so
one
of
the
things
we
want
to
do
with
kubernetes
is
bring
this
ideal
of
dynamic
control
systems
to
market,
because
systems
just
like
the
machines
can
do
some
things
a
lot
better
than
people
and
one
of
those
things
they
can
do
better
in
people
is
dynamically,
run
your
applications
and
so
creating
an
intent.
Driven
system
is
good
and
what
this
materializes
is
like
massive
reliability
gates,
because
you
have
something
that's
actively
watching
your
application.
A
It
knows
when
things
go
bad
and
it
knows
how
to
correct
it
and
actually
remediate
the
situation
and
radical
efficiency
gains,
because
you
can
pack
things
together
and
turn
them
down
when
they're,
not
in
use.
You
don't
have
to
set
aside
a
lot
of
idle
capacity.
So
that's
that's
the
second
thing
that
that
you
know
kubernetes
already
brings
the
final
thing
that
I
mentioned
this
earlier,
but
I
just
want
to
push
on
this
point
a
little
bit
like
if
you're
in
the
business
of
cloud
services
and
you're,
not
an
open
project.
A
You
are
a
massive
competitive
advantage
disadvantage
today,
like
I've,
been
humbled
by
the
amount
of
acceleration
we've
got
out
of
our
open
source
community,
particularly
from
the
community
as
a
whole.
Creating
this
very
strong
and
rich
virtuous
cycle,
where
you
know
our
engineers
will
put
something
out
there
and
get
very
immediate
feedback,
and
you
know
like
work
with
the
community
to
make
sure
that
it
works
and
it
fits.
And
frankly,
our
partners
have
been
hugely
important
in
in
progressing
our
our
technology
and
in
making
making
communities
real,
for
instance,
right
in
this
room.
A
A
You
know
out
that
you
know
our
plus
Turing
engineers
don't
have
access
to,
and
then
we
have
people
like
core
OS
that
have
actually
provided
a
lot
of
the
basic
building
blocks
that
we
can
build
on
and
they've
been
working
very
closely
with
us
to
make
this
come
together,
and
so
that
ecosystem
as
a
whole
is
creating
a
lot
of
power
for
the
project,
and
it
makes
us
very
happy,
and
another
piece
of
this
is
just
making
sure
that
everything
is
modular.
You
know
we
recognized
that
this
is
a
pretty
complex
system.
A
People
want
to
plug
out
pieces,
and
so
we've
we've
taken
modularity
very
seriously
and
made
sure
that
things
that
can
be
plugged
up
are
able
to
be
plugged
out,
and
we
encourage
people
to
take
the
take
the
system
and
work
on
it
and
build
it
out
to
meet
their
needs.
So
I'm
gonna
wrap
it
up
with
a
quick
statement
of
the
road
ahead
and
I.
A
Don't
know
some
of
the
folks
in
this
room
were
in
earlier
sessions
today
where
people
were
talking
about
like
what
does
the
world
look
like
18
months
out,
I
thought
I'd
just
take
a
sort
of
a
wild
swing
it
you
know
we
don't
want
to
go
in
five
years,
like
where's.
This
gonna
end
like
what
are
we
going
to
do
when
these
kind
of
technologies
are
are
ubiquitous,
and
you
know
obviously
we're
gonna
have
a
production
version
of
kubernetes.
A
You
know
very
soon,
I
think
by
summer
I
hope,
not
speaking
out
of
turn
for
my
engineers.
But
you
know
beyond
that.
I
think
in
the
limit
there's
a
couple
of
things
that
are
going
to
really
change
the
game
when
these
types
of
technologies
and
this
approach
to
management
become
ubiquitous.
The
first
thing
is
like
a
radically
higher
level
of
visibility
right.
A
So
if
everything's
running
in
this,
you
know
blue
pill
style
cluster
environment,
you
have
a
lot
of
insight
into
what's
running
where
it's
running,
which
version
of
its
running
what
potential
vulnerabilities
it
has
and
you're
able
to.
You
know
generate
that
view
and
that
that
ultimately
leads
to
a
lot
more
insight,
and
it
also
leads
to
a
lot
more
control.
So
we
can
create
this
empowered.
Decentralization
where
you
can
like
someone
who
has
to
deal
with
central
functions,
can
look
across
their
entire
environment
and
and
also
apply
policy.
A
You
know,
define
a
specific
set
of
policies
and
make
sure
that
things
with
vulnerabilities
cannot
physically
run
in
that
environment.
The
other
thing
that
I
think
is
gonna
happen
is
we're.
Gonna
see
a
much
stronger
separation
of
the
operations
functions,
and
this
is
a
really
important
thing
and,
like
I,
think
this
has
made
Google
so
powerful.
The
ability
to
start
thinking
about
infrastructure
operations
as
a
small
and
specialized
discipline
that
just
deals
with
the
racking
and
stacking
of
infrastructure.
A
The
ability
to
think
about
the
cluster
and
common
services
operations
as
being
I,
said
a
small
team
that
provides
a
very
useful
and
common
set
of
services
to
a
very
broad
array
of
people,
and
then
application
operations
is
something
that
exists.
On
top
of
that,
I
think
the
world
would
be
more
secure
frankly,
because
when
you're
running
in
these
types
of
environments,
the
applications
being
run
for
you,
so
you
can
run
with
a
very
low
set
of
privileges.
A
We're
really
excited
to
be
working
this
community.
Our
aspiration
is
to
bring
this
technology
to
the
world
and
we
hope
that
it
will
run.
You
know
well
everywhere,
and
obviously
we
hope
that
no
one
loves
you
or
your
money
better
than
we
do
so
we
hope
to
create
a
delightful
service
for
you
to
run
it
on
our
infrastructure.