►
Description
Case Study: Vattenfall - Sustainable Data Centers, OpenShift on OpenStack, Max Schulze at OpenShift Commons Gathering Seattle 2018
https://commons.openshift.org/gatherings/Seattle_2018.html
A
A
Justest
there
we
go
so
I
had
the
pleasure
in
early
September
of
going
to
Helsinki
and
hosting
an
open
shift
Commons
gathering
there,
and
there
I
saw
one
of
the
most
exciting
things.
I
know
it's
kind
of
geeky,
but
one
of
the
most
exciting
things
I've
seen
anybody
do
with
open
shift
and
I'm
gonna,
let
max
Schultz
from
that
infall
go
into
it.
But
if
this
doesn't
make
you
excited
about
the
potential
for
doing
good
with
open
shift,
nothing
will
so
without
further
ado
max.
Thank
you
for
coming
all
the
way
from
Berlin.
Actually.
B
I
also
want
to
say
Thank
You
Diane
for
bringing
me
here.
I
first
thought
that
I
would
be
the
last
person
that
fits
into
this
community,
but
now
that
actually
Chris
and
and
Razer
I
think,
what's
the
name
yeah
they
used
to
what
commodity
then
I
feel
right
at
home
yeah.
This
is
this
is
exactly
what
we
are
all
about
and
I'm
sure
most
of
you
have
no
idea
what
one
for
so
I
will
give
you
a
very,
very
brief
introduction
button.
B
First,
one
of
the
largest
utilities
in
Europe
missed
one
represent
owned
by
the
Swedish
state,
and
it's
actually
one
of
the
biggest
providers
of
renewable
energy
worldwide,
so
top
20
provider,
and
we
have,
of
course,
our
largest
fleet
in
Sweden,
which
is
mostly
hydropower
and
nuclear.
In
Germany.
We
also
have
called
in
the
Netherlands
gas,
and
we
are
the
biggest
builder
in
Europe
right
now
of
new
wind
assets.
B
B
And
when
you
look
at
Sweden
actually
one
of
the
biggest
industries
and
then
now
we
can
talk
a
little
bit
about
more
of
tech
and
it's
the
data
center
industry,
because
Sweden
is
very
cold.
You
have
availability
of
renewable
energy,
then
there's
lots
of
space.
We
actually
see
almost
every
single
hyperscale
AWS
Microsoft
Google
Facebook.
They
are
all
coming
to
Sweden
and
you're
building
data
centers
of
sizes.
That
I
don't
think
even
you
guys
have
seen
it
starts.
It
doesn't
end
at
300
megawatt
size.
B
It
starts
at
300
megawatts
sites,
that's
about
14,
soccer
fields
of
dimensions
and
they're
all
coming
to
the
Nordics
and
when
they
came,
we
started
looking
a
bit
more
at
data
centers,
but
before
I
talk
about
this
I
have
to
boy
you
a
bit,
because
in
order
for
my
whole
presentation
to
make
sense,
I
have
to
explain
you
two
things
about
the
energy
industry.
That
is
important
and
I'll.
Try
to
keep
it
really
short.
B
The
first
trend
that
you
see
in
live
in
in
the
energy
industry
is
renewable
energy
assets,
so
wind,
solar
hydro
has
been
around
for
quite
some
time,
but
wind
and
solar
are
quite
new
and
they
introduce
something
to
the
energy
system
that
the
energy
system
is
not
designed
to
do,
which
is
volatility.
All
of
a
sudden
I
cannot
predict
and
I
cannot
control
the
production
of
energy,
which
is
really
really
difficult.
B
The
second
trend-
and
we
can
do
a
little
experiment.
Its
decentralization
decentralization
means
that
all
of
a
sudden,
your
house
has
a
solar
panel
on
its
roof
and
were
a
battery
in
the
basement
or
a
heat
pump,
or
something
like
this.
How
many
of
you
have
an
energy
system,
maybe
on
their
house
or
in
their
house?
B
Not
that
many?
Ok?
So
if
you
do
this
in
Norway
or
in
Sweden,
I
think
like
about
80%
of
the
hands
go
up,
and
this
introduces
another
big
challenge
because
we
used
to
be
able
to
really
forecast
how
you
consume
power.
Actually,
if
you
have
a
two
family
home
to
two
kids
at
home
and
a
wife
I
can
really
forecast
your
consumption
curve
really
easily,
but
all
of
a
sudden,
if
you
produce
your
own
energy
and
you
produce
energy
that
is
again
volatile.
So
solar,
you
use
your
heat
pump.
B
So
supplier
needs
to
be
always
exactly
matching
demand,
and
this
is
expressed
in
the
energy
system
with
a
frequency
that
is
50
Hertz,
and
this
frequency
is
quite
important,
because
if
we
go
1
Hertz
off
or
your
microwave
clocks,
every
clock
you
have
in
your
house
will
go
about
10
to
15
minutes
per
day
off
track.
So
a
lot
of
these
electrical
devices
depend
on
the
grid
frequency
in
order
to
stabilize
the
frequency
and
to
make
supply
meet
demand
in
a
world
of
volatility
and
renewables.
B
We
can
do
two
things.
We
can
learn
how
to
store
energy,
which
is
what
you
hear
in
an
alien
mask,
say
a
lot
that
he
builds
gigantic
lithium
iron
blocks.
If
I
show
you
the
cost
of
these
lithium
iron
blocks
and
actually
how
unsustainable
these
things
are,
you
will
also
say:
well,
maybe
we
should
do
something
else.
To
give
you
an
example:
when
you
buy
a
Tesla,
you
need
to
drive
it
for
eight
years
and
power
it
entirely
with
renewable
energy
to
make
the
car
co2
neutral.
B
Now,
because
when
you
produce
the
end
of
the
battery,
you
need
so
much
co2
and
you're
using
so
much
minerals
that
you
actually
make
it
at
worse.
So
it's
a
nice
story,
but
it
doesn't
really
work
and
the
other
side
is.
We
can
change
how
you
consume
power,
and
now
you
will
think
I
already
have
energy-saving
light
bulbs
at
home.
That's
not
really
what
I
mean,
but
what
I
mean
is,
for
example,
your
washing
machine
consumes
a
lot
of
power.
What,
if
I
can
send
you
a
text
message
and
say:
hey?
B
Could
you
wait
an
hour
until
you
wash
your
load
or
two
hours,
because
then
we'll
have
wind
and
solar
available
and
that's
what
we're
talking
about?
This
is
demand
side,
flexibility
or
demand
response,
and
we
actually
believe
that
there
will
be
one
of
the
most
important
parts
of
the
energy
system.
Now:
data
centers
data
centers
you
can,
we
can
have
a
very
long
argument
here
of
how
much
they
will
grow,
but
I
think
we
can
all
agree
that
their
growth
is
needed.
B
So
we
need
more
digital
in
the
future
to
run
your
applications
to
run
whatever
deep
learning
machine
learning
you
want
to
do
and
this
by
the
way
is
without
crypto
mining.
Before
people
ask
in
data
centers,
they
don't
just
consume
any
type
of
power.
They
come
to
us
as
a
utility
and
they
say
ideally
I
would
like
100%,
reliable,
uninterrupted
power
supply.
That's
great!
Then
we
usually
agree
on
five
nines
or
six
nines
or
four
nines
yeah.
B
But
if
you
listen
to
what
I
just
said,
if
we
want
the
world
to
run
on
renewables,
it
will
be
very
hard
to
give
you
this
number
right.
It
will
be
almost
impossible
for
me
to
say:
yeah
I
can
give
you
always
up
power.
I
can
do
this.
Of
course
you
argue
with
nuclear
sure
that
works,
but
do
we
want
more
nuclear
assets
around
the
world?
Probably
not
I
can
also
do
it
with
hydro
in
Sweden,
but
in
the
u.s.
B
there's
some
hydro
assets,
but
it
is
very
difficult
to
do
this
with
wind
and
solar
power,
so
we
started
really
looking
into
data
centers
and
and
what
what
is
exciting
for
us
and
the
first
thing
we
saw
when
we
looked
at
the
power
consumption
was,
can
we
make
it
more
flexible?
Can
we
make
a
data
center
actually
consume
power
in
a
more
flexible
manner,
and
the
second
thing
we
we
saw
and
this
actually,
as
an
energy
utility
car,
that's
very
excited.
B
You
should
have
seen
our
colleagues
in
the
heat
Department
and
when
they
found
out
that
silicon
chips
produce
100
percent
co2,
free
heat,
so
literally
a
silicon
chip
takes
power
as
an
input,
one
megawatt
and
makes
about
0.99
megawatt
in
heat.
That's
if
you,
if
you
talk
to
the
heat
guys,
this
is
awesome
because
it's
a
very
efficient
power
to
heat
system
and
they
don't.
They
didn't
think
about
cooling.
They
thought
about.
How
can
we
use
this
heat?
How
can
we
get
it
out
and
the
descent
operators
were
very
confused?
B
I
can
tell
you
that,
and
so,
in
order
to
test
this,
we
decided
that
we're
going
to
build
a
an
actual
test.
Bed
and
now
Retta
will
be
very
happy
about
what
I'm
going
to
say,
because
we
first
went
to
them
because
we
said
maybe
we
can
solve
this
purely
on
a
software
side,
at
least
the
flexibility
topic,
and
they
were
immediately
jumping
onboard
on
this.
Not
because
necessarily
they
saw
an
opportunity
to
make
money,
but
because
they
could
really
align
with
our
vision
to
to
build
a
sustainable
digital
infrastructure,
especially
in
the
Nordics.
B
But
we
are
all
a
bit
more
about
sustainability,
I
would
say,
and
I
said.
Yes,
of
course,
you
can
do
this
a
bit
with
openshift.
You
can
start
moving
kubernetes
continuous
around
and
we
can
figure
out
a
way
how
to
make
it
flexible,
and
we
also
found
some
other
partners
Clos
and
heat
that
does
the
heat
extraction
first
explain
how
this
works.
B
A
Helio
helps
us
look
at
compute
as
more
of
a
commodity
and,
of
course,
Nvidia
could
also
very
excited
because
for
them
cooling,
these
GPUs
is
actually
a
big
problem,
so
it
became
they
were
excited
to
get
rid
of
the
heat
and
we
also
defined
okay.
What's
the
KPI,
how
do
we
define
a
sustainable
digital
infrastructure
and
for
us
it
means
we
reuse
80%
of
the
heat.
B
Now
that's
a
pretty
high
number
and
explain
to
you
how
difficult
it
is
and
we
want
power
consumption
to
be
flexible
so,
but
in
order
to
do
this,
like
I
said
we
built
to
test
sides
now,
and
we
also
have
a
lot
of
confusion
here,
because
I
always
talk
about
containers,
but
I
mean
physical
containers
because
we're
building
container
eyes
data
centers,
but
there's
some
challenges,
because
when
you
talk
to
data
center
people,
you
were
here
a
lot,
they
will
say
yeah.
Of
course
we
can
give
you
the
heat.
B
It's
no
problem,
we
were,
we
will
deliver
the
heat
out
of
the
air.
We
extract
it
and
we'll
give
it
to
you,
but
they
can
give
it
to
us
at
a
temperature
about
100
degrees,
Fahrenheit
or
40
degrees
Celsius,
which
is
useless
heat.
You
can
just
throw
it
away
right
away.
It
doesn't
make
any
sense
because
you
can't
transport
heat
at
a
temperature,
so
what
we
need
is
60
degrees,
Celsius
or
140
degrees,
Fahrenheit,
and
in
order
to
get
this,
we
have
to
do
two
things.
The
first
thing
is
we
use
water.
B
Of
course
we
use
liquid
cooling
technology
to
get
the
heat
out,
but
we
cool
the
chips,
not
with
cold
water
but
with
hot
water,
and
we
flow
the
water
at
such
an
incredible
speed
through
the
data
center.
So
that's
basically
the
inflow
temperature
of
datacenters
55
degrees,
Celsius,
so
about
120
degrees,
Fahrenheit
and
the
outflow
is
140
degrees
Fahrenheit.
B
So
you
can
imagine
how
fast
the
water
has
to
run
to
only
heat
up
by
5
Kelvin,
that's
the
Delta
that
we
need,
and
the
other
thing,
though,
if
you
look
at
the
load
curve
of
a
data
center-
and
you
know
that
the
chip
only
gets
hot
when
it
does
something
right.
So
you
need
workloads,
so
a
datacenter
doesn't
have
a
flat
workload
line
and
it's
not
running
at
100%
capacity,
all
the
time
if
you're
lucky
it
runs
at
40%.
B
So
we
had
two
right
now
we
do
this
with
all
the
sec.
We
we
had
to
manipulate
actually
the
workload.
So
so
sometimes
when
we
have
heat
demand
in
winter,
where
we
really
need
a
continuous
flow
of
six
degrees
Celsius,
we
actually
ramp
up
artificial
workloads.
So
so
we
recycle,
we
basically
run
the
CPU
on
100%
so
to
make
heat.
And
ideally
we
want
to
solve
this
problem
by
concentrating
workloads
on
certain
machines
where
we,
where
we
can
generate
the
heat
and
the
second
problem
when
it
comes
to
energy
flexibility.
B
In
order
to
actually
make
a
data
center
energy
flexible,
you
need
to
physically
shut
down
servers,
so
we
because
you
cannot
preserve
energy.
Just
by
shifting
the
workload
you
actually
need
to
physically
shut
down
machines
and
systems.
So
what
we
started
first
was
to
use
open
shift
to
actually
we
we
are
telling
open
shift,
a
part
of
the
cluster
will
go
down
and
then
it
automatically
starts
moving
containers.
B
But
then
we
also
integrate
really
deeply
with
the
UPS
systems
to
actually
remotely
turn
off
the
servers
now,
which
is
a
bit
trickier,
but
all
of
this,
it's
it's
working
now,
but
it's
there's
one
last
challenge:
it
works
for
for
artificial
workloads,
so
for
form
for
HPC
or
simulations
that
just
need
the
CPU
at
the
moment.
So
far
when
we
throw
the
data
in
and
we
have
some
projects
internally,
where
we
use
really
large
data
sets
to
forecast
weather
phenomenon,
for
example
we're
talking
about
about
a
petabyte
of
data
that
we
need
to
shift
around.
B
So
if
I
want
to
do
this
in
a
data
center,
I
either
need
a
very
big
fiber
connection
or
I
need
to
figure
out
a
way
how
to
just
move
subsets
of
data
around,
which
is
something
to
be
honest
that
we
haven't
solved
yet,
and
I
actually
hope
that.
Maybe
some
of
you
have
some
input
on
this
and
how
to
solve
it,
because
we
would
love
to
do
it
so
when
we
solve
all
these
challenges.
What's
next
for
us,
it's
not
a
commercial
project
at
all.
B
Setups
and
shifting
workloads
from
different
cloud
providers
is
what
we
are
looking
at
from
an
energy
perspective
in
order
to
basically
make
the
data
center
more
interactive
with
the
over
energy
system
and
make
workloads
more
energy,
flexible
and
I.
Think
if
we
can
find
a
way
to
bring
energy
utilities
I'm
a
software
engineer
for
me,
it's
more
logical
to
bring
utilities
together
with
the
data
center
industry
and
the
software
people,
and
and
really
have
them
talk
to
each
other.
B
I
think,
there's
a
lot
of
great
things
we
can
do
in
terms
of
sustainability,
I
think
for
utility.
I
can
tell
you
from
my
own
experience.
Now.
We've
been
working
on
this
for
one
and
a
half
years
is
that
they
stopped
thinking
about
a
data
center.
When
the
cable
arrives
at
the
transformer
at
the
power
station
yeah,
they
have
no
idea
what's
going
on
in
the
data
center
and
they
don't
care,
but
all
of
you
know
that
non
non
of
in
the
data
center.
None
of
the
servers
are
running
at
100%
capacity.