►
From YouTube: OpenShift at Volkswagen Case Study with Marcus Greul at OpenShift Commons Gathering 2019
Description
OpenShift at Volkswagen Case Study
with
Marcus Greul (Volkswagen AG) | Martin Reinke (Red Hat) | Noel O'Connor (Red Hat)
at OpenShift Commons Gathering 2019
Red Hat Summit
A
A
A
Why
is
it
so
complex
I'm
going
to
I'm
trying
to
explain
it
to
you?
An
electronic
system
in
a
car
consists
of
server
components,
components
such
as
sensors,
actuators
and
electronic
control
units
and
those
control
units
contain
one
or
more
software
components
and
those
software
components
air
for
sure
they
need
to
interact
with
each
other,
so
they're,
they're,
communicating
by
different
technologies,
bus
systems,
field,
buses
like
Ken
or
Ethernet,
and
so
on.
A
Integration
testing
means
that
such
electronic
systems
in
the
car
need
to
pass
the
integration
testing
process
for
each
combination
of
components
for
each
equipment
line
of
a
car
and,
of
course,
for
each
model
and
this
for
every
software
version
in
the
car.
So
this
alone
is
a
whole
bunch
of
work
to
do
and
I'm
not
finished
the
the
systems
itself,
the
more
capabilities
those
systems
gain
in
terms
of
taking
over
control
of
the
car.
The
more
test
cases
are
needed
to
approve
these
to
pass
these
integration
testing
process.
A
A
We
would
not
test
in
the
real
world,
at
least
for
security
reasons
or
for
safety
reasons,
additionally
to
the
virtual
environment
or
beneficial
to
the
Intuit
virtual
environment,
and
we
put
a
virtual
car
in
there
a
virtual
car.
No,
a
real
car
is
not
necessarily
need
to
be
built,
two
for
just
testing
a
subset
like
an
electronic
system
or
a
bunch
or
a
couple
of
electronic
systems
to
simulate
the
customers.
Interaction
with
a
system
a
virtual
driver
is
put
into
the
virtual
car
I.
A
Think
I,
don't
need
to
explain
why
the
driver
should
be
virtual
as
well,
so
the
last
component
will
be
the
system
under
test
or
the
test
object.
The
test
object
could
be
a
software
component
of
an
electronic
control
unit,
the
control
unit
itself
or
a
complete
system,
or
a
couple
of
systems
which
represent
a
customer's
function.
A
A
A
Components
such
as
virtual
environments,
data
analytics
other
kinds
of
tools,
test
cases,
test
case
execution,
applications
or
you
have
simulations
traffic
simulations,
for
example,
and
for
sure
the
software
components
which
finally
go
to
to
an
electronic
control
units.
When
all
those
components
are
in
stands
here
with
software
components,
the
test
bench
might
be
set
up
and
being
test
ready
in
just
a
few
minutes
compared
to
let's
say
weeks,
we
would
take
today.
We
would
need
today
months
months,
okay,
but
in
terms
of
exponential
growth.
This
will
not
be
good
enough.
A
A
It
is
possible
to
fire
just
hundreds
of
those
test
benches
automatically
by
the
system.
The
system
should
be
capable
to
do
a
proper
deployment
of
those
test
benches.
They
should
care
about
distribution
of
test
cases
to
those
test
benches
executing
them
automatically
and
for
sure
the
system
should
care
about
proper
resource.
Claiming
and
clean
up.
A
C
Yeah
I
think
this
is
really
awesome
and
I'd
like
to
introduce
myself
a
little
bit.
I'm
Martine
Wonka
I'm
working
the
solution
architect
in
redhead
in
Germany
a
little
bit
more
than
four
years
now
for
Volkswagen
and
I
was
with
that
journey
all
over
the
time
and
I
want
a
little
bit
to
explain
our
journey
until
today
and
maybe
what's
coming
up
next.
C
So
as
Marco
said
two
years
ago,
we
had
this
challenge
for
future
testing
and
to
really
think
on
how
we
could
transform
that
current
base
of
technology
and
usage
within
the
testing,
and
it
is
really
important
to
be
able
to
have
this
transformation
in
terms
of
having
multiple
test
stands
spin
up
in
a
matter
of
minutes.
Instead
of
weeks
or
months,
as
we
said
so
back
in
the
day
is
2017.
C
We
had
that
first
initial
talk
I,
remember
that,
because
my
Technical
Account
Manager
Stefan
was
coming
to
me,
who
knows
those
guys
longer
than
me
and
said
they
have
a
problem.
You
should
talk
to
them,
so
we
are
starting
the
discussion.
They
explained
what
they
just
explained
to
you
and
I
was
thinking.
This
is
really
complex.
I
mean
this
needs
to
have
agility.
This
needs
to
have
highly
complex,
dynamic
connections,
and
why
not
just
think
about
having
this
done
in
containers
Oh
might
be
a
good
idea,
but
nobody
had
a
clue
back
then.
C
C
One
of
the
first
pictures
that
are
in
that
and
you
see,
they're
a
journey.
How
customers
can
how
redhead
can
help
customers
to
adopt
container
technology
in
a
way
that
really
makes
business
outcomes
and
tangible
results,
and
this
is
all
true
except
of
the
timing.
You
see
there
are
some
weeks
and
numbers
how
long
it
takes.
So
we
did
most
of
the
steps,
despite
on
the
timing,
because
the
stuff
that
we
had
to
be
done
was
much
more
complex
and
needed
more
time.
C
So
we
proved
that
GPO
works.
We
even
get
to
a
point
where
we
really
could
work
on
other
things
and
there's
this
little
red
bar
this
tiny
red
bar
in
the
middle.
So
that
was
an
important
step
towards
the
open
innovation
lab
which
no,
we
would
talk
right
after
me,
a
little
bit
more,
and
this
was
really
to
get
this.
What
you
have
seen
as
an
outcome
together
in
a
way
that
we
really
can
do
work
together
for
the
business
of
house
one.
C
So,
as
I
mentioned,
we
had
some
challenges
in
terms
of
technical
realization,
so
we
we
have
now
nvidia
gpus,
available
and
open
shift.
We
also
work
closely
together
with
Microsoft
engineers
in
that
first
PUC
and
currently
so
also
witness
containers
will
come
to
that
solution,
but
we
still
have
a
way
to
go
in
terms
of
new
technologies,
new
architectures
being
introduced
in
terms
of
having
communication
based
on
ipv6
as
well
as
those
few
passes
that
Marcus
has
mentioned
before
introducing
other
stuff,
and
there
are
also
automotive
projects
like
autos,
are
adaptive.
B
So
they
give
me
four
and
a
half
minutes
to
talk
for
stuff
that
took
12
weeks
to
build
I.
Also,
don't
speak
German,
so
you'll
have
to
keep
up
so
I
was
brought
on
board,
let's
say
August
last
year,
untold
hey,
we
want
you
to
put
a
container
Karen
a
container
platform,
I
kind
of
asked
really.
Is
it
the
other
way
around,
but
no?
No,
no!
We
and
we
need
to
get
a
caracal
running
and
testing
and
executing
within
it.
And
yes,
one
of
those
times
you
kind
of
go.
B
We
start
so
one
thing
we
did
a
labs
is
for
we
went
in
there.
We
started
with
this
thing
called
a
vein
storming,
so
we
sat
down
with
the
cost
world
and
used
this
event
storm
we'll.
Take
it
true
solution
from
start
to
finish
or
what
you're
gonna
do
in
that
labs,
engagement
and
work
out
the
major
components
involved.
It
takes
about
two
to
three
days
for
this.
It
took
two
to
three
weeks
because
the
Germans.
B
So
yeah
myself
and
Marcus
here
were
basically
he
was
the
product
owner
and
I
was
the
the
lead
architect
involved.
So
basically
we
we
sparred
with
sabers
every
day,
but
I
was
musing.
So
when
you
come
down
to
it,
basically
it's
it's
all
about
constrained
resources.
Everything
open
ship
does
is
about
constrained
resources,
it's
about
kind
of
you
know
CPU
or
memory
or
apps,
and
all
this
type
of
stuff
what
we
did
with
Leslie.
B
B
We
wanted
something
that
could
stack
up
quickly,
get
this
job
done
and
clear
out
of
the
of
the
environment,
so
we
took
folks
were
going
to
have
this
concept
called
test
environments,
so
we
put
together
a
protocol,
a
virtual
test
environment
which
was
10
or
20
pods,
running
radars
and
light-hours
and
sensors,
and
your
virtual
drive
very
virtual
environment,
all
that
type
stuff,
and
we
kind
of
go
right.
You
know
we
put
this
on
with
plunk
it
on
and
we'll
shove
testers.
The
thing
with
kubernetes
is
basically
it's
all
eventually
consistent.
B
You
tell
it
something
you
wanted
to
do
and
it
get
back
to.
It
gets
back
to
you
when
it's
done
so,
basically,
the
way
we
actually
figure
this
out
is
basically
rather
than
having
setting
up
an
environment
and
basically,
then
throwing
tests
that
we
do
the
other
way
around.
We'd
make
opeth
part
of
the
application
itself,
so
we
did
was
basically,
we
specified
VT
definition,
which
is
a
whole
heap
of
old
ship
resources,
as
I
say,
radars
and
data
source,
and
all
that
type
of
stuff
we'd
separate.
B
We
special
how
many
GPUs
it
could
use,
not
just
for
the
rendering,
but
for
the
short
calculations
of
collision
detection
knowledge.
You
should
see
this
video.
This
is
the
same
video
we
had
a
car
doing
200
KS
an
hour
plowing
into
another
Karen.
It
was
hilarious
but
I,
tactfully,
old
virtualised.
So
what
we
did
was
basically
we
took
OpenShift.
We
basically
created
an
environment,
a
project
namespace.
We
patched
things
like
multicast,
enabling
so
the
base
is
the
components
could
detect
each
other.
They
use
multicast
to
communicate
and
do
service
discovery.
B
We
basically
when
they
were
spinning
up.
We
prepared
the
test
at
on
the
other
side
in
an
asynchronous
approach
and
basically,
when
the
VG's
were
up
ready,
stable
and
ready
to
execute,
they
just
literally
pulled
across
tests
that
had
been
loaded
onto
messaging
queues,
so
the
whole
day
was
asynchronous.
We
weren't
polling,
we
weren't
racing
for
stuff
to
get
ready.
It
executed
in
its
own
time,
yeah
so
base.
B
The
whole
thing
was
asynchronous
event
trip
and
we
had
too
many
events
in
the
end,
because
basically
there
was
a
few
issues
with
BT
restarts
in
the
middle
of
of
testing
and
all
that.
But
the
thing
that
was
important
was
basically
using
the
projects
as
deployable
units.
So
basically
we
could
run
this.
We
could
run
entities
in
parallel
and
they
were
all
isolated
and
the
SDN
part
of
that
is
very
important
for
basically
making
sure
that
Carrick's
doesn't
talk
to
Carraway
over
across
the
network.
So
the
whole
thing
was
temporal.
B
It
was
always
short-lived,
spun
it
off
excu
what
we
need
to
do
when
it
was
all
done.
We
we
shut
it
down
and
looted
from
orbit
and
just
made
them
made
the
environment
ready
for
the
next
set
of
tests
that
were
corner
come
in
these
VT
is
at
the
moment
we
did
ACC
yeah
ACC,
but
the
V
T's
will
be
will
be.
You
have
multiple
types
of
ETSU.
You
know
different
for
collision
detection,
all
this
type
of
stuff.
B
So
the
intention
is
this
big
is
to
use
a
cluster
as
much
as
you
need
and
then
just
get
off
and
let
somebody
else
use
it
so
in
the
future.
So
everything
we've
done
at
the
moment
is
about
virtualizing
the
the
software
components,
components
of
it
in
the
future.
We
want
to
obtain
this
hardware
in
the
loop,
so
hardware
and
loop
is
basically
where
you
have
actually
got
harbor,
who
we've
run
every
time.
We
actually
got
hardware
involved
in
the
testing,
as
was
so.
What
you
want
to
see
is
you're
gonna.
B
You
want
to
test
a
sitting
there
and
he
or
she
will
will
turn
a
steering
wheel
and
the
carroll
actually
turn
in
the
environment,
so
we're
gonna
have
a
whole
bit
fun
doing
that
there's
a
hole
as
Martin
as
Martin
said,
there's
ipv6,
which
is
interesting.
We're
going
to
look
a
whole
lot
of
custom
hardware,
real-time
computes
radar,
sensors
that
are
actually
detecting
things
in
real
time.
We're
looking
at
operators
to
drive
an
off-roader
this
to
basically
simplify
how
people
use
the
platform,
we're
looking
at
thing
called
virtual
cubelet
to
drive
harbour
nodes.
B
We
don't
have
a
wok
here.
We
have
it's
a
theory
right.
We
need
to
figure
it
out,
but
basically
the
end
goal
is
basically
everything
will
be
managed
and
execute
a
winning
openshift,
so
you're
not
having
to
two
pieces
of
information.
You're
floating
around
the
place
you
basically
it's
the
source
of
truth
is
obvious:
we
load
the
ROS
resource
in
there.
It
does
the
job
and
shuts
down
we're.
Looking
at
cloud
bursting.
Microsoft
is
your
AWS?
B
Doesn't
matter
there's
a
whole
heap
of
things
around,
but
this
because
this
stuff
is
expensive,
the
GPUs
are
lots
and
basically
they're
hard
to
install
the
developer
workflow
as
well
at
the
moment,
we're
using
op
ship
resources
which
yeah
they're
good,
but
basically
it's
too
low-level
for
people
who
are
developing
VTS.
So
we
look
at
custom
resource
definitions
to
specify
the
VTE
definitions
and
we're
looking
operators
to
actually
create
and
populate
them
on
board.
We're
also
looking
at
the
the
the
glorious
concept
of
chargeback,
which
is
basically
going
to
be.
B
A
D
There
are
many
challenges
involved
when
it
comes
to
developing
self-driving
cars.
Practical
simulations
and
testing
can
be
extremely
time-consuming
and
costly.
At
VW
we
found
ourselves
at
a
bit
of
a
brick
wall
that
was
until
we
attended
Red
Hat's
innovation
labs.
There
we're
able
to
create
a
new
vision
for
virtual
testing
of
customer
functions
in
vehicles.
D
Our
new
goal
was
to
have
fully
automated
and
virtual
testing,
allowing
us
to
test
functions
earlier
in
trial
development,
giving
us
greater
confidence
using
Red
Hat's,
OpenShift
container
platform.
We
are
able
to
test
and
validate
new
features
for
self-driving
cars
through
a
combination
of
software
and
hardware
systems
by
simulating
thousands
of
testing
environments
and
scenarios
for
specific
car
functions,
such
as
autonomous
driving
in
highways
or
even
autonomous
parking
in
multi-story
car
park
along
with
vehicles.
Software
open
ships,
container
technology
helps
us
put
all
the
necessary
tools,
applications
and
simulation
components
in
instantiate
all
logical
compounds.
D
This
helps
us
to
run
large
quantities
of
test
cases
in
parallel
over
a
short
amount
of
time.
Open
shifts
container
technology
has
been
provide
tools
to
our
success
and,
as
a
result,
VW
has
experienced
transformational
outcomes
allowing
us
to
work
and
think
in
new
ways
when
developing
testing
building
codes.