►
Description
Want to view more sessions and keep the conversations going? Join us for KubeCon + CloudNativeCon North America in Seattle, December 11 - 13, 2018 (http://bit.ly/KCCNCNA18) or in Shanghai, November 14-15 (http://bit.ly/kccncchina18)
Why Do I Need Kubernetes When I Already Have Cloud Foundry? - Sanjay Patil, SAP (Beginner Skill Level)
Like many of you, several SAP teams have also been asking this question to themselves. While the answer is singular in some cases (either Cloud Foundry or Kubernetes), in many situations, there is room or rather a need for both these technologies. In this session, you will learn why several SAP teams have adopted Kubernetes / container model for developing comprehensive solutions (more than what a typical ‘cf push’ can handle) and deploying them on premise and/or in the cloud. Some of these Kubernetes based solutions are also exposed as services in Cloud Foundry environment, whereby Cloud Foundry applications can consume them via the simple and familiar service binding paradigm.
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,
first
of
all,
I'm
Sanjay
Patel
product
manager,
at
ACP,
working
in
ACP
cloud
platform,
I
work
from
Palo
Alto
office
in
California
I
have
been
in
the
tech
industry
for
close
to
I.
Guess
a
couple
of
decades
now
I
have
worked
on
different
products,
platform
technologies,
application,
server,
integration
broker
or
middleware
like
CORBA
b2b
web
services,
lots
of
gray
here
my
friends
were
telling
me
that
I
had
thick
black
hair
until
I
started
looking
into
cloud
technologies,
it's
just
too
much
and
the
hair
started
turning
gray.
A
So
today,
I'm
going
to
talk
about
how
these
two
things
come
together.
This
is
not
a
product
talk,
obviously
from
ACP,
but
we
wanted
to
share
the
thinking
within
ACP
at
this
point
of
time
about
how
these
two
technologies
relate
to
ACP
and
probably
do
most
of
you.
This
is
how
I'm
structuring
my
talk.
First
of
all,
I
would
like
to
talk
about
cloud
foundry
application
runtime.
What
are
the
target
scenarios?
Acp
has
been
the
foundational
member
of
cloud
foundry
foundation.
We
have
been
part
of
the
foundation
right
from
its
inception.
A
There
are
lots
of
many
ACP
engineers
making
contributions
to
cloud
foundry
open
source
code.
Today,
rod
foundry
technology
has
been
adopted
within
ACP
solutions,
in
particular
ACP
cloud
platform.
So
definitely
we
are
one
of
the
main
supporters
leading
contributors
in
the
cloud
foundry
open
source
project,
so
I
want
to
talk
about
what
are
the
targets
in
areas
for
which
we
see
value
in
cloud
foundry
and
why
we
are
heavily
invested
in
cloud
foundry
technology
and
then
I
want
to
talk
about
some
of
the
cases
where
the
Cloud
Foundry
paradigm
really
does
not
fit
very
well.
A
Very
kubernetes
is
the
right
answer,
taking
some
concrete
examples
of
ACP
solutions
that
are
leveraging
kubernetes
today,
so
it
becomes
clear
where
we
are
using
these
two
different
stacks
and
then
I
want
to
talk
about
the
open
source
project
project
Gardner,
a
new
open
source
project
initiated
by
CP
to
make
some
of
these
kubera
Nettie's
adoption
easier
for
ACP,
as
well
as
our
customers,
because
that's
also
essentially
part
of
vision.
We
have
at
ACP
to
have
Lord
foundry
and
kubernetes
back
integrated
together.
So
that's
the
general
structure
of
my
talk.
A
So
let
me
begin
with
a
very
short,
quick
overview
of
cloud
foundry
environment.
So
what
you
see
right
here
at
the
center
is
the
application
application
in
cloud
foundry
could
be
developed
using
any
programming
language
of
your
choice,
popular
programming
languages
like
Java,
JavaScript,
Ruby,
node,
HTML
Python
go
as
well
as
you
can
use
any
other
programming
language
of
your
choice.
By
using
the
build
pack
mechanism
in
cloud
foundry,
it's
fairly
straightforward
and
to
expedite
the
innovation,
most
developers
would
use
the
backing
services
like
Postgres
sequel,
RabbitMQ.
A
Some
of
this
popular
backing
services-
MongoDB
Redis,
as
well
as
in
the
case
of
a
Seabee
cloud
platform.
They
say:
peas,
flagship,
in-memory,
database,
engine,
SCP,
Hana
and
the
platform
obviously
has
some
of
the
platform
services
like
auto
scale
or
logs
management,
job
scheduler
and
in
most
cases
you'll
also
need
a
service
for
security,
user
account
and
authentication,
using
which
you'll
connect
the
platform
to
your
corporate
ID
provider
and
in
many
cases.
Obviously,
the
cloud
applications
would
need
to
access
data
and
functionality
in
lots
of
scenarios.
A
The
data
and
functionality
recites
on-premise
systems,
so
you'll
need
a
connector
to
access
that
and
lots
of
new
cloud
native
applications
do
want
to
leverage
innovations
in
the
area
of
IOT,
machine
learning,
blockchain
so
and
so
forth.
So
there's
a
packaging
of
innovative
innovation
services
as
part
of
the
environment
and
the
entire
cloud
funding
environment
itself
runs
on
different
public
clouds
like
AWS,
agile,
GCP
and
many
more,
and
this
is
not
just
a
technical
enablement,
but
the
SE
p
cloud
platform
cloud
foundry
environment
actually
runs
as
a
platform
on
these
different
clouds.
A
So
you
can
build
an
application
pick
and
choose
which
cloud
provider
you
want
to
deploy
it
on,
and
you
can
run
that
application
over
there
and
then,
of
course,
the
additional
things
on
this
slide
are
some
of
the
tooling
to
access
the
platform.
So
that's,
basically
the
Bill
of
Materials
are
what
in
what
is
included
in
a
cloud
formed
environment.
So
let
me
also
spend
a
couple
minutes
to
talk
about
the
high-level
architecture
or
the
pattern
of
cloud
foundry
applications.
A
So,
as
I
mentioned
the
application
using
any
Bill
pack,
that's
really
the
heart
and
center
of
cloud
foundry,
environment
or
fonti
platform,
and
then
this
application
uses
any
of
the
backing
services.
So
fundamentally,
the
application
developer
focuses
on
building
the
business
logic
in
you
in
in
a
programming
language
of
his
or
her
choice.
Right
and
then
the
application
developer
would
then
specify
the
dependencies,
like
my
application
requires
certain
messaging
system.
Certain
frameworks,
certain
library
and
the
application
code
is
then
pushed
to
the
platform.
A
It's
a
platform,
then
that
platform
that
looks
into
the
application
code
identifies
what
language
it's
written
in.
If
it's
not
already
specified,
looks
into
the
list
of
dependencies
gets
hold
of
those
backing,
services
does
the
plumbing.
First
of
all,
it
creates
a
droplet,
a
running
instance
of
the
application.
Does
all
the
plumbing
to
have
your
application
micro
service,
talk
to
all
the
backing
services
and
takes
care
of
managing
health
scaling
and
all
the
runtime?
A
So
the
main
idea,
or
the
main
difference
when
compared
to
kubernetes
based
environment,
is
that
over
here
the
application
developer
experience
is
highly
simplified.
Application
developer
does
not
really
worry
about
where
my
micro
service
is
going
to
run.
He
or
she
would
also
specify
some
configurations
about
resource
requirements,
most
of
the
heavy
lifting
of
managing
the
runtime
processes,
their
scaling,
that's
done
by
the
platform.
A
Application
developers
simply
focuses
on
writing
the
code,
so
in
cloud
foundry
you
can
actually
push
the
code
along
with
just
a
manifest
of
dependencies,
and
the
cloud
foundry
environment
takes
care
of
you
know,
realizing
all
the
runtime
processes
and
its
health
management.
So
having
said
that,
this
is
certainly
a
very
powerful
model.
A
We
see
this
quite
useful
in
many
scenarios
where,
for
example,
a
customer
has
several
different
projects
going
on
in
the
IT
and
the
developers
want
to
make
different
choices
or
backing
services
programming
languages,
but
at
the
same
time
the
customer
also
wants
to
have
some
kind
of
central
governess
like
what
technologies
are
used.
If
there
is
a
security
patch
that
has
to
be
applied
to
certain
framework
instead
of
running
around
and
finding
out
which
projects
are
using
that
framework,
it's
better
to
apply
it
in
one
place
and
everybody
benefits.
A
A
There
are
cases
where
the
developer
wants
to
have
the
same
application
package
that
is
built
on
his
desktop,
also
running
in
testing
and
development
in
production,
wherever
that
is
run
right.
So
in
Cloud
Foundry
as
well,
you
could
guarantee
the
same
application
package
across
the
landscapes,
but
then
you
have
to
have
some
strict
processes
that
would
ensure
that
overall
configuration
is
same
right,
whereas
with
containers,
it's
pretty
inherent
that
once
the
developer
has
programmed
the
container,
then
it's
going
to
be
the
same
image
running
anywhere
so
for
this
scenarios.
A
The
next
one
is
obviously
one
of
the
more
important
ones,
which
is
cases
where
the
developer
needs
greater
freedom
or
greater
control
over
the
infrastructure.
Right
cloud
is
pretty
powerful.
Obviously,
cloud
platforms
do
a
lot
of
heavy
lifting,
but
behind
the
scenes
in
order
to
give
a
simpler
experience
to
the
developer,
the
cloud
platform
has
to
move
a
lot
of
things
around
if
there
was
a
banking
service
that
was
running
right
next
to
the
application
at
the
beginning.
Over
a
period
of
time,
the
micro
service
might
have
shift
a
completely
different
note
different
host.
A
So
you
cannot
make
many
assumptions
about
how
the
cloud
platform
or
platform
as
a
service
is
going
to
arrange
the
topology
of
different
parts
of
your
overall
application.
This
may
work
in
many
cases,
but
there
are
cases
where
this
just
would
not
work
like.
For
example,
if
you
have
a
machine
machine
learning
application
in
that
case,
the
application
obviously
has
to
be
deployed
on
specific
hardware
right
typically
GPU
servers.
A
You
cannot
arbitrarily
shut
down
an
instance
in
one
place
and
start
on
any
other
node,
so
binding
the
application
to
specific
hardware
that
shows
up
in
like
machine
learning,
blockchain
and
many
such
scenarios.
If
the
architecture
of
the
application,
particularly
for
multi-tenancy,
that
is
assuming
that
there
is
a
network
of
containers
for
talent,
then
you
need
a
certain
detail
level
of
control
over
how
your
application
process
institutions
are
grouped
together
and
how
they
are
associated
with
specific
tenants.
A
Likewise,
there
are
cases
where
different
processes
must
be
clubbed
together
because
they
have
certain
affinity.
There
are
cases
where
process
instances
should
not
be
running
together.
So,
for
all
these
reasons,
you
would
need
greater
control
over
the
underlying
infrastructure,
also
to
support
resiliency
and
high
availability.
In
many
cases,
you
cannot
just
depend
upon
the
platform
based
on
the
specific
application
requirements
you
need
to
get
down
into
the
details
of
how
do
you
want
to
actually
materialize
high-availability?
So
for
many
of
the
scenarios
I
just
talked
about
containerization.
A
Obviously,
the
main
main
solution
and
for
our
key
station
kubernetes
is
a
is
a
platform
that
that
we
are
going
with
I'm,
not
gonna,
speak
much
on
this
slide.
As
I
said,
this
is
the
community's
conference
or
here,
but
only
thing
that
I
want
to
point
out.
Is
this
whole
idea
of
having
a
container
instance
consuming
compute
storage
networking?
That's
pretty
pretty
common
that
has
been
around
for
a
long
time,
but
with
kubernetes
now
having
this
entire
metadata
that
is
associated
with
the
containers.
A
All
these
powerful
capabilities
that
are
introduced
by
this
metadata
layer
on
top
of
this
containers,
that
is
pretty
powerful
and
that
has
been
utilized
heavily
so
I
want
to
give
you
some
examples
of
ASAP
zone
projects
where
containerization
and
kuberan
it
is
either
Orchestrator
has
been
heavily
used.
One
of
the
main
ones
is
acp
data
hub.
So
the
problem
over
here
is
in
most
customer
scenarios.
There
are
different
data
sources,
different
data
systems
and
most
business
users
want
to
have
applications
that
consume
data
from
all
these
different
systems,
pretty
quickly
right.
A
So
this
is
an
age-old
problem.
How
do
you
get
data
from
different
systems
together?
And
we
have
been
talking
about
this
or
many
many
years,
and
still
the
problem
remains
to
be
not
completely
solved,
if
not
unsolved
right.
So
what's
the
fundamental
issue
or
here
the
issue
is
that
it
will
always
continue
to
be
the
case
that
there
are
different
data
systems
and
data
sources
right.
A
You
cannot
realistically
bring
all
this
data
in
one
place,
so
we
Oshin
that
would
allow
the
data
to
sit
very
days
and
still
allow
communication
between
these
different
data
silos
and
realizes
applications
requirement.
So
the
main
idea
that
acp
data
hub
brings
to
the
table
is
to
have
this
powerful
pipelines
of
data
processing
and
these
pipelines
are
pushed
down
to
the
nodes
that
are
running
close
to
the
data
systems,
so
the
idea
here
is
to
have
so
what's
a
pipeline
pipeline
is
essentially
a
computational
graph,
which
is
a
network
of
operators.
A
So,
as
I
mentioned,
there
could
be
data
systems,
big
data
stores,
traditional
databases,
so
the
operators
are
essentially
processing
data
that
is
specific
to
that
data
system.
Now
you
can
imagine
some
operators
sitting
close
to
the
big
data
system
and
other
operators
sitting
close
to
the
your
cloud
storage
system
and
a
higher
level
application
that
has
this
overall
data
flow
is
then
executed
by
having
these
containers
that
are
running
close
to
this
data
sources.
A
Now,
in
order
to
have
this
container
instances
talking
to
these
different
systems,
you
might
use
different
libraries
may
be
different,
binaries
and
customize
these
operators
in
in
specific,
in
way
specific
to
those
data
systems
there
may
there
may
also
be
cases
that
certain
nodes
in
this
data
flow
that
need
to
be
run
need
to
be
run
on
the
same
host
or
you
might
want
to
separate
them.
So
for
this
kind
of
scenarios,
having
a
containerized
architecture
and
orchestration
via
coronaries
comes
extremely
handy.
A
Blockchain
is
another
example
where
having
specific
Hardware
behind
the
scenes
is
or
all
fairly
commonly
required,
and
in
this
case
we
are
also
looking
into
using
different
kubernetes
clusters
for
different
talents.
Pretty
handy
model,
that's
that's
been
used
by
this
stream,
si
P
conquer
now.
The
previous
two
examples
I
talked
about
one
was
about
AC
and
AC
P
data
hub,
which,
in
my
opinion,
is
more
cosmic
platform,
that's
providing
a
certain
level
of
platform
for
data
based
applications,
the
other
example
of
blockchain
or
machine
learning.
A
A
So
if
you
are
in
the
keynote
today,
there
was
a
slide
on
end
users
using
kubernetes
in
production.
Today.
This
was
one
of
the
solutions
this
team
found
the
set
up
time
for
given
it
is
to
be
extremely
manageable
or
better
compared
to
other
platforms,
as
well
as
the
fact
that
the
Kuban
radius
orchestration
is
not
tied
to
a
specific
container
technology.
That
was
also
an
important
deciding
factor
for
this
particular
project
over
30
services.
In
this
long
list
of
services
as
part
of
the
CP
conquer
application
currently
do
use
queue.
A
Ladies
and
the
rest
of
them
are
in
the
process
of
being
modified
to
use,
containers
so
again,
I
think
this
few
examples
gives
you
a
flavor
of
different
projects
like
with
different
SCP
solutions
using
cuban.
It
is
today
SCP
data
hub
as
a
car
platform,
machine
learning
blockchain
as
backing
services,
SCP
conquer.
As
another
example,
we
are
colleagues
from
other
teams
as
well
over
here,
ACP
hybrids.
They
have
been
using
kubernetes
for
quite
some
time
as
well.
So
here
is
a
list
of
few
of
the
projects.
A
Some
are
services,
some
are
applications
and
there
are
obviously
projects
from
our
customers
and
partners.
We
have
very
similar
requirements
so
for
all
of
these,
what
we
really
need
is
communities
cluster
as
a
service
solution.
That
is
a
secure
and
scalable
right
like,
for
example,
if
a
team
is
using
kubernetes
for
their
project,
then
they
would
want
to
create
a
cluster.
Do
some
development
create
another
cluster
for
testing
multiple
times
testing,
instead
of
maintaining
a
cluster
all
along?
A
If
the
set-up
time
is
much
higher
than
it's
getting
in
your
way
of
continuous
integration,
continuous
delivery
concepts
for
our
customers
having
a
solution
that
works
homogeneously
across
the
public
clouds
and
on-premise,
so
that
they
can
easily
create
kubernetes
clusters
cluster
as
a
service
on
these
different
public
clouds
is
very
critical
so
that
they
can,
they
can
extra,
expand
their
business
into
any
markets
that
they
would
like
to
because
they
can
easily
start
a
cluster
in
a
public
cloud
in
that
different
region.
Have
a
cluster
running
on
top
of
that
and
have
their
solutions
deployed.
A
On
top
of
that,
and
all
of
this
has
to
be
done
with
minimal
TCO-
that's
really
an
important
point.
Right.
Lots
of
open-source
tooling
is
out
there.
They
would
let
you
get
the
system
up
and
running,
but
for
day-to-day
to
operations,
it's
not
always
straightforward.
So
we
actually
looked
at
several
options
to
provide.
You
know
operate
kubernetes
as
I
was
mentioning
in
the
community.
There
are
lots
of
tools.
Many
of
them
are,
frankly
speaking
still
developer
focused.
A
They
have
either
limited
or
no
support
for
day
two
and
they
we
didn't
find
them
to
handle
thousands
of
clusters,
which
is
the
which
is
the
case
in
the
case
of
SA
P,
the
commercial
tools
that
we
looked
at.
They
aim
to
fill
the
gaps
in
the
community
tools,
but
then
there
is
a
risk
of
having
a
vendor
lock-in
and,
in
general,
the
time
to
market
was
fairly
high
by
having
this
commercial
tools
to
work
with
the
solutions
that
we
needed
hyper
cloud
providers.
A
Obviously
they
have
solutions
for
cluster
as
a
service,
and
obviously
we
work
with
all
of
these
partners.
But
what
we
wanted
to
have
is
a
simple
cluster
as
a
service
that
is
independent
of
the
different
cloud
providers,
so
that
our
development
teams
don't
have
to
have
separate
plans
for
testing
their
solution
with
god
provider
a
and
have
another
testing
for
cloud
provider
B.
We
wanted
to
have
a
model
where
they
test
their
solution
on
Q
burn.
A
It
is
cluster
and
it
should
run
uniformly
on
any
public
cloud
as
well
as
in
the
on-premise
systems
set
up
using
OpenStack
VMware.
What
have
you
so
for
these
reasons?
What
sa
P
started
is
a
project
Gardiner
which
is
currently
open,
source
github.com,
slash
Gardiner,
please
do
check
out,
and
if
you
like
it,
please
start
it
as
well.
So
this
is
launched
by
CP
gaining
community
support
fairly
faster.
A
The
main
promise
is
to
be
able
to
create
and
manage
given
a
DS
clusters
on
any
cloud
or
on-premise
with
just
a
simple
single
click
of
your
mouse.
It
is
extensible
as
most
things
in
kubernetes.
You
can
add
new
features.
You
cannot
support
for
new
cloud
providers
as
well
and
one
of
the
unique
aspects
of
this
project
is.
It
uses
kubernetes
itself
for
automated
health
management
and
scaling
of
the
control
planes
and
I'll
talk
about
that
shortly
or
here.
A
But
how
are
you
actually
going
to
keep
all
of
these
clusters
in
sync,
like
maintaining
all
the
control
planes
on
these
different
clusters
is
not
a
trivial
task
to
say
the
to
is
to
at
least
so.
The
implementation
approach
in
gardener
is
that
the
clusters
that
are
actually
running
the
workload
we
take
out
the
control
planes
of
those
clusters
and
make
those
control
planes
as
the
workload
of
dedicated
kubernetes
clusters.
So
the
idea
here
is
to
have
chute
clusters.
A
The
concept
of
Syrian
chute
is
coming
from
like
the
botanical
world,
so
you
have
the
seeds
that
will
sprout
different
shoots.
The
chute
clusters
are
the
one
that
are
running
only
worker
nodes
and
their
control.
Planes
are
running
in
the
seed
cluster
and
then
I
can
quickly
give
a
demo
of
the
gardener
project,
or
here
may
the
demo
gods
be
with
with
us
with
me
rather
so
this
is
a
simple,
fairly
straightforward
demo,
and
this
is
again
something
that
has
been
used
in
sa
P
by
several
different
teams.
A
A
All
right,
sorry
about
that,
okay,
so
this
is
the
gardener
dashboard,
and
so
there
are
many
options
for
administration
members,
and
if
you
look
at
the
secrets
here,
you
can
obviously
configure
your
secrets
for
different
cloud
providers.
Currently
Amazon
Web
Services
is
what
I
can
figure
it
for
our
demo.
We
are
currently
prepared
to
deploy
clusters
as
a
service
on
Azure
cloud.
Dcp,
OpenStack
Ali
and
there
are
more
cloud
provider
as
well
as
on-premise.
Solutions
that
are
becoming
available
shortly
so
to
create
clusters
is
fairly
straightforward.
A
So,
for
example,
I
have
few
clusters
running
over
here
already.
The
interface
for
creating
new
clusters
would
look
something
like
this.
You
can
obviously
provide
the
cluster
name
pick
the
infrastructure
public
cloud
product
that
you
are
interested
in
for
this
demo.
Aws
is
what
I'm
going
with
you
can
pick.
A
specific
region
also
pick
a
specific
version
of
Kuban.
It
is
that
you're
interested
in
and
for
the
cloud
provider.
A
You
can
go
deep
down
and
go
to
the
next
level
and
also
pick
the
zone
that
you're
interested
in
and
the
purpose
for
which
you
want
to
have
this
cluster
created.
Interestingly,
you
can
also
get
into
the
details
of
setting
up
the
node
pools
like,
for
example,
if
you
wanted
to
add
different
node
pools.
If
you
are
not
happy
with
the
machine
types
in
that
node
pool,
if
you
needed
more
powerful
machines
hosts
in
that
particular
pole,
you
can
configure
them
or
here.
In
addition,
we
can.
A
You
can
also
configure
different
operating
systems
right
like
if
you
go
with
any
specific
cloud
providers
cluster
as
a
service
solution.
You
have
a
limited
set
of
operating
systems
and
a
limited
set
of
other
choices,
but
over
here
we
have
much
much
wider
option
of
all
of
these
configurations
for
the
node
pools.
So
with
that
you
could
create
these,
you
can
have
different
add-ons.
You
can
also
configure
the
maintenance
schedule
for
your
clusters,
whether
you
want
the
patches
to
be
applied
automatically
or
not.
So
our
fall
in
the
interest
of
time.
A
I'll
go
into
one
of
these
clusters
that
I
just
set
up
before
getting
onstage
over
here.
So
here
you
can.
This
is
a
cluster
that
has
been
created
and
and
the
cluster
creation
should
take
you
less
than
a
few
minutes.
In
fact,
I
created
a
cluster
when
I
was
on
the
way
from
airport
to
my
hotel
room
from
my
mobile
phone.
It
should
work
from
your
phone
as
well.
I
mean
this
there's
no
big
deal
over
there,
but
if
someone
needs
to
do
that,
you
could
do
that.
So
this
is
a
cube
configure
here.
A
The
main
point
I
would
like
to
draw
your
attention
to
is
this
is
perfectly
standard
cube,
config
right,
so
if
it
is
from
a
command
prompt,
if
I
were
to
look
at
the
cube,
cuddle
config
view,
you
can
use
the
standard
tooling
to
manage
your
is
cluster
from
any
of
your
tooling
or
from
the
dashboard
as
well,
because
we
are
using
the
totally
standard
configuration
and
other
other
mechanisms,
or
here
so
I
believe
this
is
slightly
slow
or,
as
usual,
the
demos
always
have
slight
glitches.
But
we'll
come
back
to
that.
A
And
this
should
be
the
familiar
kubernetes
dashboard
that
you
would
see
for
managing
any
clusters,
or
here
I
just
deployed
a
workload
for
a
simple
workload
with
five
instances:
five
five
parts-
or
here
you
can
get
into
parts
as
well
as,
for
example,
replica
sets
and
for
any
given
replica
set
over
here
right
from
over
here.
Then
you
can
scale
that
to
have
multiple
number
of
parts
as
you
need
it,
you
can
also
edit
the
replica
set,
which
is
typically
not
what
you
might
do.
A
You
might
actually
want
to
edit
the
deployment,
but
just
to
make
the
point
over
here,
you
can
make
the
changes
to
to
the
specification,
or
here
and
and
then
just
by
having
a
submit
or
apply
button.
The
Kuban
arias
would
of
bringing
the
state
of
your
cluster
to
the
specification
that
you
just
made
over
here
and
yeah,
so
yep.
A
So
I
think
this
is
the
commander
Hydra
shirt,
so
this
specific
cube
Kuro
configured
here
and
what
you
would
see
from
the
command
prompt
they
are
exactly
same
and
I
could
use
any
other
to
control
commands
over
here
like,
for
example,
get
the
parts
for
my
hello
web
application
that
I
just
deployed.
For
some
reason.
This
is
slightly
slower
than
normal
or
here
but
I
guess.
This
makes
the
point
of
having
a
single
view
for
creating
clusters
on
any
cloud
provider
that
you
would
like
and
whether
it's
on
premise
lot
in
cloud.
A
All
right,
so
that
was
the
demo
and
as
far
as
architecture
is
concerned,
as
I
was
mentioning,
there
is
the
concept
of
using
kubernetes
itself
to
manage
the
control
planes
of
your
actual
clusters
that
are
running
the
workload.
So
this
first
cluster
that
we
spin
off
is
the
garden
cluster,
which
is
kind
of
the
central
cluster
from
which
then
you
would
access
the
different
target
infrastructure
as
a
service
options.
A
So
for
each
target
is
option,
you
would
then
first
be
enough:
a
single
seed
cluster,
so
this
seed
cluster
will
then
have
the
different
controllers
that
would
be
used
for
managing
your
actual
shoot
clusters
which
run
the
worker
which,
on
the
workload
so
with
that,
as
the
main
innovation
happening
at
acp,
which
is
now
open
source.
Again,
please
do
check
out
on
github.com
slash
gardener.
The
question
then
remains
is
how
do
all
these
things
fit
together
right?
A
So
we
have
different
ACP
projects
as
well
as
customers
and
partners
using
or
wanting
to
use,
Cabernet
DS,
and
we
have
the
cluster
as
a
service
solution
coming
along
si
P.
As
a
strong
supporter
of
Cloud
Foundry,
we
have
a
Cloud
Foundry
based
pass
offering
for
simplified
development
experience
for
agile
innovation.
So
how
do
these
things
fit
together?
So,
as
as
I
just
mentioned
for
agile
innovation,
we
have
cloud
foundry
environment,
we
always
start
with
multi
cloud
as
the
main
main
starting
point,
and
then
there's
got
foundry
environment
running
on
these
different
cloud
environments.
A
You
would
then
create
services
either
directly
running
on
any
of
these
infrastructures
or
from
elsewhere,
and
you
would
bring
those
services
for
consumption
in
the
cloud
from
the
environment
and
you
would
build
the
apps
using
cloud
foundry.
So
this
is
the
current
status
right,
so
you
have
Platte
foundry
as
a
past
solution.
That's
running
on
these
different
multi-cloud
options
and
the
cloud
foundry
platform
is
creating
apps.
Those
are
consuming.
Services
works
beautifully
for
agile
innovation
scenarios,
twelve
factor
apps,
stateless
apps.
This
cannot
typically
be
the
complex
applications
like
acp
concur.
A
But
for
those
scenarios
where
12
factor
apps
is
the
choice
to
be
made.
Cloud
Foundry
works
beautifully
for
communities
environment
as
I
listed
some
of
the
different
types
of
workloads
like
the
data
hub
blockchain.
As
a
service
or
a
CP,
Conqueror
and
other
applications
or
here
again,
multi
cloud
is
our
main
main
main
concept
in
all
of
these
offerings
and
then
use
kubernetes,
the
with
open
source
gardener
as
a
project
to
create
and
manage
clusters
in
a
simplified
and
easier
manner,
and
then
you
would
build
services
on
top
as
well
as
applications.
A
So
this
is
how
the
kubernetes
world
is
handling
problems
and
the
different
types
of
workloads
at
the
CP
and
in
our
ecosystem
going
forward.
However,
we
have
a
vision,
we're
again,
starting
with
multi
cloud
and
having
gardener
provide
kubernetes
cluster
as
a
service
offering
and
providing
solutions,
environment
for
creating
services
and
apps.
A
What
we
would
also
like
to
say
is
the
cloud
foundry
itself
is
running
on
top
of
Cuban
Ares,
and
so
that
cloud
foundry
becomes
available
wherever
kubernetes
is
available
from
any
CP
standpoint,
and
then
the
apps
that
are
created
on
top
of
Cloud
Foundry,
they
would
also
run
as
communities
cluster
workloads.
So
the
main
idea
or
here
is
that
today,
those
teams
which
started
with
using
Cloud
Foundry
environment
as
the
platform
at
the
beginning
and
when
the
applications
outgrew
and
they
needed
to
get
into
the
lower-level
infrastructure
as
a
service
and
started
building
some
cuban.
A
It
is
based
solutions.
These
are
the
same
teams
which
are
handling
both
12
factor
apps,
as
well
as
the
stateful
apps.
So
when
these
teams
have
to
make
the
offering
in
a
totally
different
region,
they
have
to
start
worrying
about
how
do
I
get
my
Cloud
Foundry
platform
set
up?
How
do
I
get
my
cluster
for
the
backing
services
set
up
so
by
having
Cuban
IDs
as
the
underlying
fabric
and
running
both
Cuban
IDs
workloads,
as
well
as
the
entire
Cloud
Foundry
platform
as
a
workload
on
Cuban
it
is.
A
It
becomes
much
more
simpler,
both
from
the
developer
experience,
as
well
as
from
the
operational
efficiency
point
of
view
for
acp
as
well.
So
in
this
regard,
ACP
has
been
driving
a
bunch
of
integration
projects.
Now
some
of
there
are
taking
place
under
the
cloud
foundry
foundation
to
start
with
cross
consuming
services
from
cloud
foundry
from
tinnitus
apps
or
services
in
Cuban.
A
It
is
from
Cloud,
Foundry
apps
and
to
start
with,
using,
for
example,
kubernetes
scheduler
for
the
cloud
foundry
apps
to
begin
with,
and
then
containerize
the
entire
cloud
foundry
platform
and
run
it
as
a
workload
on
kubernetes,
so
lots
of
open
source
activities
both
over
here,
as
well
as
in
the
cloud
foundry
foundation.
So
that
was
pretty
much.
My
last
slide
hope
that
gives
you
an
idea
of
how
ACP
is
looking
at
both
of
these
platforms
and
our
vision
to
bring
these
two
things
together.
Thank
you.
A
Yeah,
so
we
actually
have
some
of
the
main
product
team
members
as
well
here
yep,
so
the
question
is:
will
the
clusters
created
using
gardiner
also
supports
security,
with
role
based
access
control,
and
the
answer
is
yes,
and
this
is
also
one
of
the
important
points
that
I
would
like
to
make?
Is
the
seed
cluster
that
I
mentioned
that
is
running
inside
and
inside
a
target
infrastructure
service?
A
That's
essentially
opening
up
the
normal
end
points
of
kubernetes
for
accessing
the
API
server,
so
you
might
have
your
other
management
cluster
running
elsewhere,
and
now
the
seed
cluster
provides
role-based
access
to
your
management
cluster
running
elsewhere.
This
becomes
an
extremely
important
quality
in
cases
where
you
cannot
really
directly
access.
The
infrastructure
in
the
target
is
because
of
the
geopolitical
or
other
reasons,
or
if
the
infrastructure
is
running
on
premise,
you
just
do
not
open
these
endpoints,
so
this
is
a
role
based
access
control
mechanism
used
for
implementation
of
Gardner
itself.
A
A
So
will
it
will
it
lead
to
something
like
a
marketplace
where
other
services
and
solutions
will
become
available?
And
the
answer
is
absolutely
yes
and
that's
one
of
the
fundamental
ideas
that
when
this
ecosystem
grows,
if
you
look
at
this
model
of
garden,
cluster
seed,
cluster
and
chute
cluster
as
I
was
pointing
out
just
to
take
an
example.
A
Likewise,
then,
you
can
build
additional
extension
communities,
extensions
for
higher
levels
of
extensions
as
well,
so
it
can
essentially
lead
to
an
iTunes
like
model
where
this
just
becomes
a
platform
for
creating
cluster
as
a
service,
and
then
the
partners
and
others
can
bring
their
solutions
and
create
a
marketplace
for
cross
uses
all
right.
Any
last
question
I
think
otherwise
we
are
towards
end
of
this
time
of
this
session.
Thank
you.
So
much
for
coming.