►
From YouTube: Node.js, in a Galaxy Far, Far Away (and 300 million pockets) by Bill Fine, Joyent, a Samsung Company
Description
Node.js, in a Galaxy Far, Far Away (and 300 million pockets) - Bill Fine, Joyent (a Samsung Company)
Samsung holds the #1 global market share position in mobile phones, TVs and refrigerators, and has deployed over one billion connected devices. Increasingly, however, it is the software applications on those devices (instead of the physical characteristics of the devices themselves) that differentiate Samsung products.
As Samsung evolves and grows its software application portfolio it is migrating those applications to a Node.js and Docker-based backend platform, designed to support over 100 million daily users accessing 100s of petabytes of data,.
This presentation will provide a detailed overview of Samsung's Node.js and Docker-based architecture, lessons learned and challenges that remain, plus insights into Node.js-based, IoT apps coming to a Galaxy near you (in the not too distant future).
A
So
bill
was
originally
going
to
give
this
talk,
but
a
month
ago
he
asked
me
to
do
it,
and
I
was
more
than
happy
to
I
love,
node
I
love,
what
we're
doing
with
it.
So
here
we
go
so
the
slides
and
all
the
links
to
everything
I'm
going
to
cover
are
available
at
that
link
at
the
bottom,
so
jst
com,
/
ni.
A
A
So
adding
new
features
or
fixing
bugs
was
very
difficult,
and
this
is
due
to
the
fact
that
what
we
had
actually
built
back
then
was
a
monolith
and
it
had
lots
of
issues,
namely
the
release
cycles
are
very
slow.
We
had
again,
as
I
said,
cascading
failures.
If
you
tried
to
fix
a
bug,
sometimes
a
new
bug
would
appear
and
it
was
generally
just
difficult
to
work
with
so
back
in
two
thousand
eight
and
two
thousand
nine.
A
They
started
switching
over
to
running
JavaScript
on
the
server
and
at
that
time
no
did
not
exist,
so
they
were
running
everything
with
spider
monkey
or
starting
to
go
that
route,
and
they
found
that
it
was
perfect
for
doing
a
sink.
Io
programming,
as
we
all
know,
but
then
note
came
out
and
node
got
a
little
bit
more
stable
and
the
people
at
joint
saw
this
and
decided
to
just
go
full
in
on
node
itself.
A
So
we
hired
ryan
dell,
the
creator
of
node
and
then
Isaac's
schluter,
who
is
the
NPM
creator,
and
we
began
migrating
that
big
monolith,
ruby-on-rails
implementation
over
to
micro
services,
using
node
and
throughout
this
process
we
supported
lots
of
core
improvements
and
note
itself
which
I'll
talk
about,
but
I
went
back
and
I
looked
at
the
public
announcement
on
whenever
we
actually
decided
to
adopt
node
and
I
found
some
pretty
well
here.
Let
me
read
it
to
you
and
you
can
figure
out
what
you
think
is
about
this
okay.
A
So
back
then,
in
to
2010
mark
Mayo,
our
CTO
at
the
time
said
we
believe.
Node
is
a
foundational
technology
that
will
enable
developers
to
build
the
next
generation
of
apps
that
will
finally
rusher
in
the
Internet
of
Things
era.
This
is
back
in
twenty
ten
people,
weren't
doing
node
iono
seemed
pretty
advanced
at
the
time,
are
pretty
pretty
insightful,
so
the
other
quote
in
the
announcement
was,
we
knew
the
asynch
model.
A
Node
and
forest
was
right
and
we
knew
JavaScript
would
work
too,
as
we'd
already
fallen
in
love
with
JavaScript
outside
of
the
browser
building
the
giant
smart
platform.
So
we
adopted
a
node,
but
at
the
time,
will
act
a
lot
of
tools
that
we
needed,
especially
as
we
were
trying
to
move
into
more
enterprise
customers.
So
one
thing
that
we
did
was
added
DTrace
probes
into
note
itself.
A
We
also
added
support
for
doing
post
mortem
debugging
with
md
b,
which
is
the
modular
debugger
that
comes
with
smart
OS,
and
we
created
and
supported
the
creation
of
several
modules
that
helped
with
observability
that
had
dtrace
probes
built
in
so
bunion
via
singh
can
rest
if
I
and
as
we
started
to
just
destroy
or
not
destroy
but
separate.
All
of
our
concerns.
In
that
monolith
we
were
implementing
microservices,
and
this
aligned
very
well
with
our
core
ethos,
which
is
a
as
a
unit.
Shop,
should
not
surprise.
A
You
is
the
UNIX
philosophy
and,
as
a
result
of
going
the
micro
services
route,
all
of
the
components
that
we
built
were
bounded
and
we're
very
easy
to
take
down
or
to
add
to
and
iterate
quickly
we're
able
to
add
new
features
very
simply,
and
we
started
to
see
places
that
we
could
actually
use
note
that
probably
other
people
wouldn't
expect
to
find
node.
So
at
the
core
of
the
stack
that
actually
does
the
orchestration
for
our
or
plat
or
for
our
cloud,
it's
called
Triton.
We
have
knowed
everywhere.
A
So
here
are
some
examples
of
places
that
we've
used
node
that
you
probably
wouldn't
expect
it
to
be
used.
So
a
dns
server.
We
have
it
running
as
our
ldap
server.
We
have
it
managing
all
of
the
knicks
on
our
computing
ads.
We
even
have
a
key
value
store,
that's
backed
by
postgres.
It's
all
written
a
node
and
we've
learned
through
this
journey.
That
note
is
perfect
for
CLI
tools,
so
the
tools
that
actually
talk
to
Triton
and
manta
are
all
done
in
node
and
pretty
much
all
of
our
CLI
tools
are
all
done.
A
A
node-
and
here
are
some
other
ways
that
we've
are
other
places
that
we're
using
node.
So
we
have
a
telemetry
agent.
We
have
dhcp
server
and
NFS
server
and
the
point
of
all
this
is
not
to
say,
like
hey,
look
at
all
these
great
places
that
we
were
using
note
it's
more
to
get
you
to
think
about
other
places
beyond
just
the
API
layer
or
just
this
orchestration
layer
between
your
mobile
services
and
some
legacy
services
that
node
actually
can
work
in.
A
So
if
you
look
at
this
architectural
diagram
of
Triton,
which,
as
I
said,
is
the
orchestration
stack
for
managing
all
the
infrastructure
in
our
data
centers,
you
can
see
all
the
green
are
places
that
are
written
a
known.
So
all
those
components
are
node,
which
is,
as
you
can
see,
pretty
much
everywhere,
and
that
is
all
open
source.
So
going
you
know
full
in
with
node.
We
also,
we
already
had
a
value
of
open
sourcing
everything
we
do,
but
note
especially
encourages
this.
A
I.
Have
this
example:
microservices
app
it's
a
pretty
typical
node
microservices
app
using
docker,
and
if
you
want
to
actually
see
this
running,
if
you
come
to
our
booth,
I
can
I
can
walk
it
walk
through
it
with
you,
but
it's
it's
simple.
It
has
three
microservice
sensors
that
are
pulling
data
from
a
smartthings
hub.
A
So
it's
reading
temperature
humidity
motion
and
then
it's
just
throwing
those
into
a
serializer
microservice,
that's
save
to
in
flux,
and
then
you
have
a
simple
graph
display
on
the
front
end
through
web
sockets
and
to
make
all
that
possible
without
the
need
for
a
load
balancer
anywhere.
We
created
something
called
container
pilot.
So
this
implements
pattern.
We
call
the
autopilot
pattern
and
it
means
that
all
of
your
containers
and
your
apps
are
self-aware
in
self-operating
and
it
has
support
for
console
or
at
CD.
A
A
All
you
need
to
do
is
copy
over
the
container
pilot
binary
and
then
pass
that
in
to
your
command
or
entry
point,
because
container
pilot
will
also
act
or
run
as
the
ped
one,
which
is
typically,
where
you'd
have
your
init
process
running
so
it'll
reap
any
zombie
processes
for
you
and
take
care
of
starting
up.
All
of
your
services.
A
So
if
you
look
at
the
configuration,
though,
for
container
pilot,
it's
pretty
simple,
it's
just
JSON
configuration
you
specify
if
you
want
to
use
console
or
at
CD,
and
then
you
specify
the
services
that
you
want
container
pilot
to
register
for
you
with
console
or
at
CD.
So
in
this
case
it's
registering
the
front
end
and
then
we
have
a
health
check
which
is
all
run
within
the
actual
container
itself,
which
means
you
don't
have
to
expose
your
heartbeat
or
health
endpoints
publicly
and
then
and
I.
A
In
this
example,
it's
just
scaling
serializer
22
and
since
we're
using
console
for
discovery,
all
the
all
the
humidity
motion,
temperature
and
front-end
are
notified
that
there's
another
instance
of
the
serializer
and
they'll
act
accordingly
and
if
you're
using
Triton,
it's
very
simple
to
use,
docker
and
deploy
your
your
containers.
All
you
have
to
do
is
set
your
docker
host
environment
variable
to
point
to
one
of
our
data.
A
Centers
and
then
you
can
use
that
same
command,
docker
compose
up
and
am
you
have
five
containers
created
for
you
and
any
of
them
that
actually
required
to
be
exposed
publicly
get
a
public
IP
address
and
all
of
the
networking
is
taken
care
of.
For
you
very
simple,
another
really
cool
benefit
of
using
container
pilot
is
we
have
promethea
support
built-in
so
for
doing
ops
reporting
like
in
this
example
where
we're
reporting
on
the
free
memory
in
the
container
it's
all
built
in.
A
So
you
just
add
this
to
your
configuration
and
then
it'll
register
that
with
Prometheus
for
you,
so
I
have
a
new
option.
Now
my
drop-down
tell
me
the
amount
of
free
memory
in
my
front
end
containers-
and
here
you
can
graph
that,
so
you
can
scale
out
the
number
of
front-end
instances
you
have,
you
can
drop
it
back
down
and
you
can
see
the
the
memory
graph
and
anything
else.
You
really
care
about
about
your
application.
A
So
in
this
example,
we're
saying
give
me
the
serializer
addresses
for
any
instance
that's
healthy,
and
if
you
don't
have
any
serializers
that
are
healthy,
don't
even
do
anything,
don't
try
to
connect
and
then,
in
the
next
part,
we're
just
using
Seneca,
which
is
a
micro
services
toolkit
for
node
and
creating
a
connection
so
with
container
pilot.
There's
a
lot
of
great
advantages,
as
I've
mentioned,
some
of
them.
That
I
think
are
notable,
though,
or
the
fact
that
developers
can
actually
move
faster.
A
A
So,
if
you're
interested
in
that,
please
reach
out
and
we're
doing
more
work
to
make
Triton
itself
more
observable.
So
we
have
this
open
tracing
feature
that
we're
trying
to
add
in
and
for
everything
else
that
we're
doing,
including
this
we
have
it
all
publicly
available.
So
if
you
go
to
giant
on
github
/
RFD
release
for
discussion,
you
can
find
everything
that
we're
planning
again.
The
slides
are
at
JSC,
ecomme,
/
and
I,
and
thanks
y'all.