►
From YouTube: Introducing Racklet a commoditized educational and open source data center scale-model for tinkerers
Description
“The future is already here - it's just not evenly distributed” - William Gibson
We’d like to announce our latest open-source project: Racklet. It’s a fully-integrated, Raspberry Pi form-factor server rack and software stack that aims to be a scale model of hyperscaler datacenters. All layers of the stack are 100% OSS/OSH, and will be developed together with the community. It’s reproducible through open PCB designs, 3D printed casing, and commodity, off-the-shelf hardware.
We want to lower the barrier of entry for becoming cloud native. Racklet aims to inspire users to explore how modern server architectures work, in a tangible and educational way. Emphasis is put on security, knowledge sharing, extensibility, and portability.
The goal is to conceptually map to real environments and provide an accessible and well-documented path to welcome future talents to the world of cloud native.
B
A
The
next
session
on
the
enterprise
track
would
be
github's
challenges
for
the
enterprise.
Now
here
on
the
community
track.
We
have
two
of
some
gentlemen,
lucas
and
dennis
I
don't
want
to
pronounce
their
names.
Wrongly
probably,
I
will
learn
once
they
pronounce
their
names.
They
are
joining
us
from.
Oh,
I
can't
remember
the
name
of
the
country,
but.
A
B
B
Perfect
great,
then,
let's
start
hi
welcome
to
this
presentation,
and
yes,
as
as
was
mentioned
here,
we're
today,
gonna
announce
our
new
open
source
project
called
raclet
and
we're
joining
you
from
finland
today
and
we're
very
excited
to
be
at
this
conference
cool.
So
the
cloud.
What
do
you
think
about
when
you
hear
the
cloud?
B
B
And
this
kind
of
there's
a
mismatch
here
in
like
we
can
do
some
of
the
things
through
through
dashboards
without
knowing
what
is
you
know
under
underneath
what
is
happening
in
the
real
world,
but
sometimes
you
really
need
that
kind
of
deeper
understanding.
So
let's
take
an
example
here
of
how
you
know
a
web
app
is,
is
introduced
in
say
when
you're
learning
about
these
things,
then
you
have
this
kind
of
web
app,
simplified
model.
B
You
know
it's
talking
to
the
database
and
that's
you
know
all
great,
but
in
the
real
world
it
more
looks
like
this
kind
of
murder
mystery
that
you
see
here
on
the
right
picture
and
then
that's
just
how
it
is,
and
how
can
we
kind
of
bridge
the
gap
of
this
kind
of
mental
model
issue
that
you
have
this
simplification
and
then
in
the
real
world?
It's
much
more
much
much
more
complex.
B
B
This
will
we'll
get
into
later
and,
as
was
found
in
this
master's
thesis
from
kaspa,
nissan
and
martin
jensen
at
urhus
university
in
denmark,
they
they
studied
this
kind
of
thing
and
and
wondered
like
if
we
have
something
tangible,
something
you
know
like
learning
cloud
system
and
something
you
can
touch
you
can.
You
know
roll
around
with
it
and
you
know
physically
change.
B
The
object
between
the
students
or
the
the
one
that
is
learning
and
to
get
the
full
comprehension
of
the
the
kind
of
large,
complex
systems.
That
is
the
case
in
you
know
these
large
cloud
native
systems
these
days
so
we'd
like
to
introduce
raclette,
which
is
this
kind
exactly
this
kind
of
mediator,
objects
and,
and
we'll
we'll
tell
more
as
we
go
with
this
presentation.
D
Story
so
understanding
the
cloud
in
a
holistic
way
is
really
hard.
This
here
is
a
visualization
of
what
you,
as
the
user
or
as
the
developers,
see
when
interacting
with
kubernetes
or
deploying
your
applications
compared
to
all
of
the
behind
the
scene,
complexity
that
is
required
to
run
a
cloud
in
today's
ever
more
complex
technology
ecosystem.
It
is
crucial
to
also
see
the
bigger
picture,
for
example
in
order
to
understand
error
situations
and
to
make
better
decisions
when
developing
applications.
D
D
B
Yep
exactly
so
like
for
us,
we
we
are
both
students
and
you
know,
that's
not
realistic
to
just
you
know,
go
at
some.
Some
large
corporations
say
we
want
to
buy
all
of
these
servers.
That's
not
within
our
budget
in
any
way.
So
what
can
we
do?
Well,
we
can
build
our
own
cloud
from
commodity
components,
for
example
like
raspberry
pi
and
some
other
things
of
the
logos
listed
here,
and
that
is
much
cheaper
and
much
more
accessible.
D
B
So
here
is
the
infamous
cloud
native
landscape.
Many
were
many
of
these
components
that
if
you
build
your
own
cloud,
if
you
go
that
route
you're
expected
to
integrate
these
things-
and
you
know
it's
just
overwhelming
like
where
do
you
start?
You
really
need
some
good
kind
of
glue
to
bind
all
of
these
things
together.
B
Well,
one
very
common
way
is:
is
bash,
so
shell
scripting
you,
you
write
something
imperatively
like
run
this
command
first
and
run
that
command,
but
it's
it's
kind
of
falls
apart
at
some
level
of
complexity
where
it
it
just
you
know,
doesn't
work
well
in
error
conditions
and
isn't
as
robust
as
it.
You
know
the
glue
needs
to
be
for
this
kind
of
really
complex
and
dynamic
systems.
D
B
B
I
want
to
contribute
so
I
started
to
contribute,
and
the
first
thing
I
I
did
was
to
contribute
porting
kubernetes
to
arm
so
that
it
worked
on
my
raspberry
pi's,
because
I,
the
only
linux
computers
I
had
around
were
raspberry
pi's,
because
you
know
they
were
cheap
and
accessible,
and
and
that's
why
I
did
it,
and
eventually
I
built
this
kind
of
this
kind
of
small
cloud
that
you
see
on
the
picture
and
it
was
fairly
straightforward.
I
had
a
usb
hub
that
was
feeding
power
to
all
of
the
raspberry
pi's.
B
B
D
And
to
do
that
more
robust
solution,
we
can
utilize
the
raspberry
pi
hat
standard,
so
hardware
until
attached
on
top.
So
it's
essentially
a
pcb
that
goes
on
top
of
each
of
the
single
board,
computers
and
using
this
we
can
power,
the
raspberry
pi's
and
the
other
single
board
computers
in
a
standardized
way
using
the
gpio
header,
and
this
kind
of
standardized
hat
form
factor
also
allows
for
monitoring,
voltage
and
current
of
each
single
board
computer
separately,
and
we
eliminate
the
single
point
of
failure.
B
Cool,
so
so
this
that's
kind
of
a
lot
of
problems.
Let's,
let's
take
them
in
order,
so
first
the
sd
card,
we
swap
out
an
ssd
which
we
should
give
us
better
capacity,
resiliency
and
and
also
speeds.
Then
we
we
switched
to
using
immutable
infrastructure,
which
means
that
we
network
boot
when
starting
the
pi.
We
network
boot
a
pre-made
image
from
some
kind
of
registry,
and
then
you
know,
do
all
of
the
security
stuff,
like
cryptographic,
verification
and
and
then
to
do
this
kind
of
network
secure
network
booting.
B
We
need
some
kind
of
hardware
route
of
trust,
so
we
need
some
kind
of
key
to
know.
Is
this
you
know
secure
or
not
and
that
we
place
on
the
this
kind
of
hat
that
dennis
was
talking
about.
We,
we
have
the
the
key
there,
as
you
can
see
in
this
picture,
and
that
is,
you
know,
managed
by
microcontroller.
B
B
But
now
we
get
to
kubernetes
level,
and-
and
this
is
you
know-
kubernetes
is
kind
of
hard
to
upgrade
and
manage
on
its
own,
and
you
know
I
worked
on
cube
adm,
as
I
said
for
quite
some
time
previously
and
while
that
went
pretty
well
and
now
it's
a
staple
in
in
all
of
the
you
know,
bootstrapping
scenarios,
it's
just
a
limited
blog.
You
know
it's
a
building
block
on
which
you
build
larger
things.
D
Well,
to
avoid
this
mutable
and
imperative
pattern
that
you
just
described,
we
can
bake
kubernetes
right
into
the
assigned
operating
system
images,
so
it
comes
from
the
network.
Preloaded
and
everything
is
set
up.
You
can
just
need
to
fire
it
up
and
we
can
also
leverage
git
ops
to
achieve
a
fully
declarative
infrastructure,
so
no
more
running
imperative
commands
anywhere
and
all
in
all.
What
this
means
is
that,
in
order
to
upgrade
a
node,
for
example,
you
only
need
to
do
one
file
change
in
a
call
request
and
that's
it.
B
That's
really
cool,
but
then
the
next
question
is
like
is
these:
are
these
things
that
we
have
described?
Are
they
realistic
is
someone
you
know
got
to
have
thought
about
this
before?
So?
Is
that
the
case?
Well?
Yes,
someone
has
thought
about
this
before
the
hype
scale,
as
the
large
companies
like
google
and
microsoft,
and
amazon
and
facebook,
they
you
know,
have
they
are
running
into
these
problems
as
well,
and
they
have
made
really
excellent
solutions.
B
They
also
have
published
parts
of
these.
You
know
the
blueprints
of
their
data
centers
in
open
compute
projects,
and
you
can
see
one
of
the
kind
of
designs
here
on
the
picture,
and-
and
here
you
know,
we
have
you
with
raclette-
we
have
taken
inspiration
and
and
want
to
make
a
scale
model
of
these
kinds
of
really.
You
know
large,
open,
compute
project
servers.
We
want
to
make
something
smaller,
but
but
still
kind
of
have
the
same
characteristics.
D
And
indeed,
the
architecture
that
we
just
described
to
you
here
with
this
visual
visualization
borrows
many
concepts
from
the
open
compute
project
and
also
from
their
cloud
native
ecosystem.
It
has,
for
example,
optimized
power
delivery.
We
have
hardware
and
software
security
features,
and
we
also
have
a
fully
declarative
configuration
with
immutable
infrastructure.
B
Cool,
so
that
said,
you
know
like
all
these
things
we
these
are
aspirations
that
that
we
can
also
say
you
have
probably
seen
this
kind
of
learning
curve
picture
before
we
want
to
emphasize
that
we're
really
in
the
beginning
of
this
journey,
where
we
have
been
thinking
about
these
things,
you
know
for
a
while
now,
but
it's
it's
really
like
early
stage,
so
we'd
like
to
invite
you
to
join
us
in
journey.
You
know
to
the
top.
B
Let's
see
what
those
are
so
the
first
one,
the
first
value
is
security.
We
emphasize
security
really
highly
in
this
and
that's
a
main
distinguishing
factor
between
raclette
and
some
other
projects,
and
we
want
to
use.
You
know
the
good
stuff
that
has
been
developed
in
the
cloud
native
ecosystem
and
also
in
the
open
source
firmware
ecosystem.
D
And
as
the
second
value,
we
have
interoperability,
and
this
is
really
important
to
avoid
this
very
common
xkcd
standard
scenario
that's
depicted
here
and
our
approach
to
avoiding
this
is
to
combine
upcoming
and
cutting
edge
standards
with
the
least
amount
of
glue
possible.
So
this
means
that
we
want
to
explicitly
avoid
creating
new
and
redundant
standards
that
only
apply
to
this
project.
D
B
D
And
in
addition
to
the
public
availability,
the
raclet
hardware,
design
itself
will
be
also
as
accessible
and
as
reproducible
as
possible.
So,
for
example,
we're
going
to
leverage
3d,
printing,
we're
going
to
leverage
3d
printing
and
an
open
source
cad
workflow
for
all
of
the
structural
components
of
this
project.
B
Cool
and
the
schematics
will
be
also
made
open,
source
or
open
hardware.
In
kicad,
for
example,
we
have
a
markdown
book
online
docsracklet.io,
which
is
work
in
progress
where
you
can
find
the
documentation
for
the
software
components
and
everything
all
the
rest
of
the
stuff
will
be
just
off
the
shelf
components,
which
means
you
can
buy
it
in
your
local
web
store.
D
B
Cool
and
one
of
the
transparencies
one
of
the
values
as
well.
We
want
to
demystify
the
cloud
stack
and
make
it
more
accessible
and
that's
why
we
we're
going
to
use
these
kind
of
good
cloud
native
projects
like
jaeger
prometers.
You
know
openmetrics
to
to
make
all
of
the
things
in
the
system
observable.
D
And
after
that
we
have
maintainability
and
upgradeability
as
the
sixth
value,
so
the
maintenance
burden
of
keeping
a
right-click
cluster
running
and
up-to-date
should
be
really
minimal,
such
that,
for
example,
software
upgrades
should
be
very
much
seamless.
We
don't
want
to
have
an
I.t
team.
I
t
team
specialized
at
just
maintaining
this
rack.
It
should
be
done,
be
able
to
be
upgraded
by
anyone
and
on
the
hardware
side
we
also
support
or
aim
to
support
hot
swapping
and
then
upgrades
of
individual
components
so
that
you
can.
B
Cool
and
the
last
one
is
affordability.
We
want
to
use
cheap
commodity
components
or
inexpensive
ones
such
that
it's
compared
to
you
know
like
these
large-scale
servers,
if
you
would,
if
you
buy
a
stack
of
them,
racklet
is
much
more
cost-effective
and
also
thanks
to
the
modularity
that
denis
was
talking
about.
You
can
also
make
raclette
fit
your
or
your
budget
or
local
availability,
and,
let's
take
a
look
at
the
use
cases.
B
Cool
and
then
we
have
education,
which
is
super
important
in
this
field.
We
want
that.
You
know,
for
example,
universities
and
other
educational
institutes
should
be
able
to
to
get
a
rack
or
a
raclet
easily
and
use
that
for
education
of
these
cloud
computing
systems
and
who
knows
like
using
this
kind
of
system.
Raclet
could
be
the
start
of
someone's
cloud-native
career.
B
D
D
D
B
E
A
C
B
E
E
E
B
A
B
B
By
doing
these
tangible
things
and-
and
you
know
making
it
easier
to
understand,
you
know-
and
this
is
so
complex-
all
of
these
cloud
native
things-
I
you
know-
I
I
don't
have
no
idea-
you
know
how
many
cncf
projects
there
there
are
these
days,
there's
just
so
many
every
day
coming
in
and
and
then
that
would
be
this
would
we
hope
would
be
the
you
know,
perfect
testbed
for
that
kind
of
stuff.
So
you
know
like
you,
you
want
to
try
something
out.
B
You
know
you
just
like
download
it
and
then
see
it.
Then
you
can
plug
wires
and
stuff
and
then
have
it
really
tangible.
That
is,
that
is
why
we
start
a
project,
but,
but
I
said
like
there's
no
code
yet
just
like
design
and
and
these
kinds
of
ideas
and
then,
where
you
know
we're
open
to
all
kinds
of
feedback
from
the
community.
E
B
C
D
Yeah,
I
can
just
add
that
kind
of
we
also
felt
that
there
was
a
huge
discrepancy
between
what
the
hyperscalers
were
doing.
They
were
going
full
steam
ahead
with
their
custom
cloud
solutions
and
we
kind
of
lay
man
normal
consumer
people
here
are
just
kind
of
left
behind.
So
we
wanted
to
research
into
this
state
of
the
art
and
kind
of.
How
can
we
bring
this
into
home,
labs
and
education?
How
can
we
make
people
aware
of
these
systems,
so
that
was
also
one
incentive
behind
this.
E
B
B
C
B
Like
community
management
is
a
super
important
super
important
work,
you
know
helping
people
find
each
other
and
these
kinds
of
things-
that's
that's.
You
know.
People,
for
example,
in
special
interest
group,
contribute
or
experience
that
satoshi
was
talking
about.
It
is
like
just
amazing
so
like
that,
as
well
could
be,
could
be
a
place.
D
And,
of
course,
if
you're
great
at
writing
documentation-
or
you
want
to
learn
more,
then
I
also
or
we
as
a
team,
also
highly
encourage
reading
through
the
the
request
for
comments,
documents
of
raclette
and
kind
of
even
small
contributor
contributions
like
fixing
typos
and
improving
wording
and
saying
that
this
chapter
is
unclear.
All
of
that
is
is
highly
appreciated,
so
feel
free,
there's
so
many
different
things
that
you
can
do
in
wrecklet
and
feel
free
to
just
come
in
and
contribute
yeah.
B
B
D
B
A
A
E
Contribute
to
future,
possibly
dogs.
You
know
anything
basically
so
yeah
awesome
presentation
of
some
conversations
we've
had
so
far
and
yeah.
I
think
we
are
on
time
or
out
of
time,
let's
see
yeah
exactly
so.
I
would
want
to
ask
them
this
one
question,
but
I
don't
have
enough
time
for
that,
but
I
think
I
will
have
to
proceed
anyways.
So
that
is
you.
I
see
a
student
and
we
are
also
an
open
source
developer
so
like.
How
do
you
manage
your
time?
E
C
B
Yeah,
that's
that's
also
why
it
fits
perfectly
with
this
conference
to
to
tell
people
about
it
that
you
know
now
we're
just
like
starting
to
ramp
up.
We
haven't,
you
know
said,
as
we
said
like
done
much
more
than
the
idea
part
and
and
now
we're
gonna
start
doing
stuff
when
we
get
a
bit
more
time
in
the
summer,
because
because
yeah,
it's
tight
with
university.