►
Description
Ever wondered what software gets developed at a national laboratory? Thanks to the open source efforts of LLNL, you don't have to wonder any longer!
Join Ian Lee, Computer Engineer at Lawrence Livermore National Laboratory, as he highlights the ongoing expansion of the open source community inside one Department of Energy research lab. These developers bring software and tools which support science and security, issues of national importance, into the open source community. Ian will also describe the unique challenge of bringing decades old software projects and historically closed source developers into the open source community for the first time.
About GitHub Universe:
GitHub Universe is a two-day conference dedicated to the creativity and curiosity of the largest software community in the world. Sessions cover topics from team culture to open source software across industries and technologies.
For more information on GitHub Universe, check the website:
https://githubuniverse.com
A
Great
you
want
to
welcome
you
all
thank
you
for
making
it
out
to
the
talk
open
source
software
and
national
security.
These
are
two
topics.
You
generally
don't
hear
in
the
same
sentence,
but
as
we
heard
in
the
keynote
this
morning,
there's
actually
quite
a
lot
of
work
being
done
in
this
area,
in
the
federal
government,
in
the
open
source
communities
around
government
and
in
between
agencies,
both
inner
sourcing,
open
sourcing,
some
clothes
sourcing
and
there's
some
real
nuances
and
difficulties
in
some
parts
of
the
federal
government.
A
It's
an
evolving
process,
as
al
Vaughn
said
this
morning,
we're
going
to
be
working
on
this
for
a
while
and
I
think
the
best
is
yet
to
come.
My
name
is
Ian
Lee
I'm,
a
computer
engineer
at
Lawrence,
Livermore
National
Lab
for
those
of
you
that
maybe
aren't
as
familiar
with
what
that
is,
we're
actually
only
located
about
50
miles
east
of
here
out
in
Livermore
California
and
we're
one
hello.
B
A
We
go
we're
one
of
the
Department
of
Energy's
research
labs.
We
grew
out
of
the
after
math
of
the
Manhattan
Project
Lawrence,
sorry,
Los,
Alamos,
National
Laboratory,
founded
back
in
the
nineteen
forty
in
fifties,
and
our
mission
is
to
develop
and
solve
the
big
science.
The
big
computing
problems
of
our
time,
so
we're
located
on
one
square
mile
out
in
the
eastern
part
of
Livermore.
For
those
of
you
that
have
been
out
there.
A
It
actually
just
got
incorporated
and
came
officially
part
of
Livermore
in
the
last
few
years,
I
was
a
former
Air
Force
Base
that
they
convert
it
over
in
and
that
we
currently
uses
our
home.
We
have
about
sixty
three
hundred
employees
and
we're
one
of
actually
17
do
II
national
laboratories
across
the
country
in
the
orange
or
in
the
bottom
left.
A
You
can
see
we're
actually
one
of
three
nectaries,
along
with
Sandia
and
Los
Alamos,
which
are
responsible
each
year
as
one
of
our
main
objectives,
reporting
to
the
president
on
the
state
and
readiness
of
the
nuclear
stockpile
stewardship.
So
the
idea
here
is:
we
have
these
nuclear
stockpiles
these
weapons
since
their
early
90s.
You
can't
go
out
to
the
desert
and
test
them
there,
so
what
have
we
done
to
grow
out
of
it?
Well,
we've
had
to
develop
big
computing
simulations
to
solve
these
problems
in,
to
be
ants
able
to
answer
the
question.
A
If
called
upon
would
this
stockpile
be
ready?
You
may
also
have
seen
us
recently
in
star
trek
into
darkness.
We've
done
a
little
bit
of
work
there,
so
we
actually,
this
is
shot
from
inside
of
NIF
NIF,
the
National
Ignition
facility,
where
for
those
of
you
that
saw
the
movie,
the
the
warp
core
is
sort
of
the
focal
point
there.
A
A
The
big
physics
questions
the
big
science
questions
of
the
day
and
actually
having
to
reinvent
everything
that
the
open
source
community
is
built
already
would
be
a
ridiculous
waste
of
taxpayer
dollars
and,
as
was
mentioned,
the
keynote
you
know
trying
to
reduce
the
duplication
across
across
the
federal
agencies
is
one
of
the
main
goals
as
we
strive
forward
into
the
future
here.
So
because
we're
a
national
laboratory,
because
we
grew
out
of
a
time
in
the
nation's
history,
where
we
had
nuclear
development
happening,
we
actually
have
quite
a
lot
of
silos.
A
We
have
many
pillars
inside
of
just
Lawrence
Livermore
between
lawrence
livermore,
Los,
Alamos,
Sandia,
the
other
national
labs
and
other
parts
of
the
federal
agencies
right.
So
we
have
do
we
as
a
whole.
We
have
the
djinns
community
agencies,
we
have
D
many
many
many
parts
of
the
federal
government.
A
We
all
sit
in
these
individual
silos
and
what
open
source
is
allowed
us
to
do
is
actually
to
bridge
those
gaps
to
start
piercing
holes
into
these
silos
and
to
make
the
code
the
resources
and
the
knowledge
that
we're
having
to
develop
to
solve
our
problems
available.
More
broadly,
so
what
does
that
look
like?
Well,
you
know,
obviously
we're
here
at
github
universe,
the
open
source
community
as
a
whole
has
really
embraced
git
and
github
com
as
a
hosting
platform
for
developing
software.
Internally,
though,
the
picture
is
often
very
different.
A
As
I
mentioned,
we
were
founded
back
in
1952
in
the
aftermath
of
World
War
two
and
at
that
time,
the
largest
supercomputers.
What
was
actually
qualified
as
the
first
supercomputer,
the
cdc
3300,
was
1
million
instructions
per
second
right.
That's
I,
don't
know
your
cell
phone
back
ten
years
ago,
sort
of
timeframe
and
what
was
a
supercomputer
at
the
time
nowadays.
A
Obviously,
we
have
much
bigger
systems
than
that
are
currently.
Our
largest
system
is
16,
pedo,
flops
and
actually
developed
on
those
systems
is
not
something
that
the
general
open
source
community
General
developers,
folks
that
are
developing
software,
for
your
laptop,
actually
have
to
deal
with
so
I
mentioned,
we've
been
around
for
60
years
just
over
60
years,
and
it's
been
an
evolution
of
computing
equipment
technology
and
we've
tried
to
be
this
entire
time.
At
the
forefront
of
that
development.
A
A
Timox
simulations
math
simulations
meshing
and
all
of
this
really
grows
out
of
the
need
to
have
these
large
computing
resources
available
right,
they're,
not
necessarily
designed
originally
for
solving
these
sorts
of
problems.
In
fact,
a
lot
of
the
supercomputers
in
the
last
20
years
have
been
collections
large
large,
where
you
essentially
just
take
many
many
CPUs
common
CPUs
like
you,
would
find
in
your
desktop.
You
put
them
all
together
and
you
put
them
all
facing
at
the
same
problem
all
at
once.
The
petascale
system
that
we
have
right
now.
Our
largest
is
Sequoia.
A
It
has
1.6
million
CPUs
in
it,
and
it's
actually
on
the
top
500
list
for
2012.
For
a
limited
time,
we've
had
some
of
the
biggest
systems,
but
what
does
that
actually
mean
so
to
give
you
an
idea
for
those
either
familiar
in
the
HPC
world?
One
of
the
biggest
list
sort
of
the
the
Grammys,
if
you
will
of
the
supercomputing
world
of
hpc,
is
the
top
500
list.
A
We
have
internally,
and
these
are
the
systems
that
we
run
on.
These
are
the
systems
that
are
scientists.
Our
engineers
are
researchers
that
are
doing
research,
including,
but
not
limited
to,
if
heard
news
recently,
the
cancer
moonshot
initiative
is
a
collaboration
involving
Lawrence
Livermore
and
involving
these
large
computing
resources.
So
having
a
bunch
of
hardware,
as
I
said,
you
know,
take
a
bunch
of
CPUs,
you
can
put
them
all
together
in
a
rack,
turns
out
in
the
case
of
Sequoia,
it
sits
on
about
4,500
square
feet.
A
So,
to
give
you
an
idea,
that's
about
the
size
of
a
house
plot
in
California,
so
it's
a
good
chunk
of
real
estate,
but
in
order
to
actually
do
anything
with
it
do
anything
useful
with
it.
You
actually
need
an
operating
system
right,
so
linux
mac
windows
are
operating
systems.
We
have
actually
had
to
develop
our
own
this.
The
earlier
version
of
this
was
called
chaos,
the
clustered
high-availability
operating
system.
It
was
grown
out
of
the
linux
program,
linux
project
at
Lawrence
Livermore.
It's
a
collaboration
between
the
tional
laboratories.
A
Again
when
we're
not
the
number
one
spot,
where
also
often
rotating
amongst
Sandia
Los,
Alamos
oak
ridge,
with
the
titan
supercomputer
most
recently
at
oakridge,
and
the
reality
is
there's
kind
of
a
different
software
stack
than
you're
used
to
developing
with
on
your
mac
and
where
you're,
the
only
user
and
the
reality
is
that
we
have
a
need
for
actually
scheduling
out
these
jobs
for
doing
things
like
high
availability
file
systems
for
low-latency
interconnects.
One
of
the
biggest
challenges
in
computing
today
is
IO
bandwidth.
A
We
have
both
incredibly
large
data,
again
file
systems
on
the
order
of
10
petabytes
single
file
systems
that
we
have
to
deal
with.
We
have
network
connections
between
these
hundred
or
1.6
million
CPU
cores
and
how
to
actually
work
with
that
and
not
have
the
contention
that
would
slow
down
an
application
in
something
like
a
Hadoop
environment
where
you're,
maybe
not
necessarily
carrying
about
you,
know
the
most
optimized
Layton
sees
between
things.
You
can
sort
of
distribute
out
the
work
and
collect
it
back.
A
It's
a
very
different
computing
environment
to
an
a
large
MPI
environment
like
this,
where
you're
passing
messages
between
these
cores
and
broadcasting
across
the
entire
system,
doing
reductions
and
doing
these
large
operations,
so
we've
actually
built
our
own
stack
that
we
develop
in
collaboration
with
again
Los
Alamos
and
Sandia
to
have
a
unified
operating
system
stack
and
the
reality.
Is
we
build
right
on
top
of
Red
Hat
Enterprise
Linux?
In
our
case,
we
take
the
base
set
of
packages
there
about
3300
packages.
Moves
don't
need,
but
we
also
add
a
lot
on
top
of
that.
A
So
I
wanted
to
highlight
a
few
of
the
open
source
projects.
Again,
we
have
had
to
build
these
computer
systems
to
do
our
own
physics.
Simulations
and
a
lot
of
the
pieces
that
have
fallen
out
of
this
are
have
made
their
way
into
the
open
source
community
and
actually
actually
have
done
that
quite
a
while.
So
slurm
is
a
job
scheduler.
This
is
a
resource
manager
to
actually
in
general,
you're,
not
running,
one
user
doesn't
have
the
entire
system
all
at
one
time.
A
The
reality
is
that,
most
of
the
time,
it's
subdivided
up,
you
might
have
let's
say
100
users
running
simultaneously
jobs
that
maybe
I
need
a
few
hundred
nodes.
Someone
else
needs
a
couple
thousand,
but
you've
got
1.6
million
here.
So,
actually,
dividing
that
up
to
be
able
to
schedule
that
work
to
happen
efficiently
and
to
schedule
things
to
schedule.
Different
resources
that
are
needed
is
as
a
common
problem
and
it
becomes
quite
interesting
when
you
have
these
sorts
of
large,
though
slurm
was
originally
started
back
in
2001.
A
A
Actually,
this
was
full
of
one
of
our
very
successful
technology
transfers
where
a
product
an
open
source
product
that
we
had
developed
internally,
a
few
the
developers
went
to
found
a
company
where
they've
actually
been
able
to
support
this
across
the
entire
worlds.
There's
HPC
centers
in
flint
in
france,
in
the
rest
of
europe,
in
aids
that
are
using
salaam
to
do
their
jobs,
getting
the
resource
management
across
their
systems.
A
Flux
is
a
currently
prototype
system.
It's
in
very
active
development.
It's
actually
a
framework
of
components,
and
it's
designed
to
be
the
evolution,
or
this
may
be
the
potential
successor
for
firm
at
some
point,
and
the
idea
here
is
when
slurm
was
developed
back
in
2001,
some
of
the
things
that
are
commonplace
now
we're
not
common
in
computers,
even
in
super
computers.
A
At
the
time
things
like
GPUs
things
like
strange
new
to
memory
file
systems,
all
these
sorts
of
things,
weren't
necessarily
something
that
the
resource
manager
knew
how
to
handle
and,
for
the
most
part,
they've
been
added
their
support
for
them.
Now
for
scheduling,
okay,
I'm
going
to
need
a
job
I
need
this.
Many
CPUs
I
need
this
much
memory.
Now
you
can
handle
okay,
I
need
these
gpus,
but
as
the
types
of
devices
that
we're
seeing
come
out
of
the
open
source
community
and
the
the
software
community
as
a
whole
continues
to
evolve.
A
There's
reason
you
know
set
of
requirements
instead
of
pieces,
so
flux
is
the
evolution
again
of
slurm
of
job
scheduling,
where
the
idea
is
that
you
can
subdivide
out
pieces
that
you
can
make
plugins
as
very
common
in
the
software
development
world
nowadays.
But
again
was
not
in
decades
past
to
have
these
modular
applications
and
actually
together
the
pieces.
A
Efsun
links
so
one
of
the
things
that
we
deal
a
lot
with
is
these
large
file
systems
and
we
have
everything
from
millions
and
millions
or
billions
of
tiny,
tiny
files.
Right
essentially,
let's
say
a
little
identifier
is
little
messages,
little
pieces
of
information
from
these
large
multi
physics
simulations.
We
also
have
environments
where
we
have
our
simulations,
where
we
have
very
few
but
very,
very
large
files.
A
The
reality
is
that
the
file
system
file
systems
that
are
available
in
the
commodity
world
in
the
open
source
world
in
your
map
distribute
don't
always
necessarily
solve
the
same
problems,
but,
for
example,
if
your
Linux
user,
you
might
be
familiar
XD
or
file
system
and
think
well,
why
don't
we
just
use
that
that's
good,
it
turns
out.
It
may
be
a
fine
file
system,
certainly
very
battle-hardened.
A
However,
it
only
sports
up
to
volume,
teen
terabytes
so
and
talking
about
petabyte
files
is
order,
a
magnitude
greater
you
can't
use
necessarily
these
same
products
and
pad
here
again
is
the
ZFS
on
Linux
project
root
of
the
DFS
on
solaris
or
the
open
ZFS
community.
It
was
originally
available
through
Solaris,
back
development
actually
started
in
2001
was
released
to
solaris
in
2005,
and
the
reality
is
that
lawrence
livermore
actually
led
the
work
to
pour
initially
ZFS
to
linux.
Work
began
in
2008
and
in
2011
is
when
the
first
actual
release
happened.
A
You
can
see,
there's
a
nice
spike
right
there
around
the
time
it
got
release
and
we've
seen
a
fairly
steady
community
of
adoption,
and
one
of
the
great
successes
is
that
as
of
this
year,
it
actually
ships
now
in
Ubuntu.
So
it's
now
become
as
easy
to
get
and
use
this
files
as
apt-get
install
ZFS
and
enable
the
the
pooling.
So
again,
this
is
a
project
that
grew
out
of
Lawrence.
Livermore
is
managed
by
Lawrence
Livermore,
at
least
the
Linux
proportion
of,
in
collaboration
with
many
open-source.
Many
vendors
across.
A
A
But
the
reality
is
if,
if
I'm
user,
a
and
there's
a
second
user
B-
and
we
both
want
to
build
up
to
tools
in
the
HP
in
in
the
general
commodity
world,
that
might
be
fine
right,
I
have
two
different
things
and
they're
just
going
to
I'm,
mostly
just
going
to
use
the
default
configuration
I'm,
probably
not
going
to
tweak
things
in
the
HPC
environment.
There's
oftentimes,
multiple
MPI
versions,
multiple
platforms
that
we
care
about,
IBM,
vs,
intel
linux
versus
blue
jean.
We
have
several
different
versions
of
each
package.
A
We
have
up
to
seven
different
compilers
intel,
GCC,
clang,
various
other
open
source
projects
that
are
being
done
for
a
compiler
work
or
MPI
work,
and
when
you
have
things
like
applications
that
have
48
dependencies,
you
end
up
with
combinations
like
7,500
possible
ways
to
combine
and
compile
and
Link
all
this
together
and
the
reality
is
that
if
you
had
to
do
it,
where
every
user
did
this
individually
and
make
that
available,
it
doesn't
scale
very
well.
So
spac
was
developed
to
so
that
problem
by
building
some
of
the
dependency
requirements
into
the
actual
system.
A
Once
we've
run
these
big
simulations,
what
do
you
actually
do
with
that?
And
how
do
you
share
that
data
across
researchers,
different
collaborators,
bursa,
tease
other
government
partners
and
for
that
matter,
the
rest
of
the
world
so
es
GF?
Is
a
Lawrence
Livermore,
originally
instigated
texture,
a
collaboration
between
several
agencies,
Lawrence
or
more
NASA,
several
others
where
we're
trying
to
distribute
out
these
large
petabytes
worth
of
data
in
a
distributed
way
across
the
world
where
federated
sites
can
work
together.
A
It's
sort
of
distributed,
AWS
style
where
you
have
a
bunch
of
data
and
you
can
get
it
out
from
various
different
points,
and
the
idea
here
is
to
enable
international
collaboration
to
have
this
sort
of
decentralized
database
around,
in
this
case
climate
science
data.
But
we
have
several
other
projects
that
are
working
with
other
types
of
data
geo
data.
Again
as
we've
as
I
mentioned
earlier.
We
have
some
work
happening
in
the
cancer
in
the
bio
realm
and
how
to
share
that
data
safely
and
securely
across
the
different
environments.
A
Is
some
active
research
that,
where
we're
pursuing
and
as
we
build
up
these
large
simulations,
we
have
this
large
amount
of
data
coming
out
of
our
of
our
systems?
How
do
you
actually
visualize
that?
How
do
you
visualize
a
petabyte
of
data
I
mean
even
just
wrapping
your
head
around
that
number
as
a
whole?
It
can
be
difficult,
even
a
terabyte
or
a
gigabyte,
for
that
matter
actually
brought
many
zeros.
It's
quite
a
challenge.
I
mean
to
actually
wrap
your
head
around.
What
that
really
means,
and
so
what
we've
had
to
do
is
actually
develop.
A
Tools
in
this
case
visit
is
a
visualization
tool
that
we
developed
to
actually
visualize
and
do
computations
and
do
reductions
and
different
sorts
of
overlays
and
graphics
around
these
large
multiphysics
and
other
associated
codes,
so
that
work
is
done
in
a
very
large
community.
I
wasn't
able
to
include
the
the
github
stats
here,
we're
actually
in
the
process
of
moving
it
from
svn
again.
This
is.
These
are
a
lot
of
a
lot
of
the
projects
that
we
have
grew
up
in
a
time
when
git
and
github
did
not
exist.
A
We
have
projects
internally
that
have
been
around
for
10
20
30
years
that
this
software
code
base
has
actually
been
alive,
and
there
is,
you
know,
some
amount
of
effort
to
move
that
forward
to
move
big
code
base
and
some
of
the
product,
some
of
the
things
that
you
could
get
away
with
doing
at
the
time
with
svn.
Maybe
it
was
considered
bad
form,
but
you
could
do
it
things
like
checking
in
binary
files
into
your
source
code
at
the
time
it
worked
nowadays
generally
considered
bad
form,
there's
other
better
ways
to
do
it.
A
A
To
give
you
an
idea,
github
limit
is
two
gigabytes
for
your
get
repos
in
case.
You
didn't
know
that
so
again,
underpinning
these
this
research,
this
hpc
development
relations
that
we're
doing
the
data
science
that
we're
doing
are
a
number
of
research
activities.
There's
a
lot
of
work
being
done
in
mesh
refinement
in
adaptive,
meshing
in
visualization
of
flows
of
different
sorts
of
materials.
How
things
react
to
different
sorts
of
physics,
cases
how
heart
valves
operate?
A
There's
some
work
that
we're
doing
a
project
called
cardioid
that
is
actually
simulating
the
human
heart
and
how
it
beats
over
time
and
having
a
mesh
that
represents
that
heart
and
how
that
changes
over
I
mean
how
you
evolve
at
that
simulation
over
time
is
some
active
work
that
we're
doing
in
collaboration
with
IBM,
and
a
lot
of
that
work
is
happening
out
in
the
open.
A
lot
of
these
projects
have
been
moving
into
the
open
or
have
existed
in
the
open
for
a
long
time.
A
You
know
talking
about
the
percent
of
our
projects.
Move
forward
being
open
sourced
as
a
federal
requirement
is
very
interesting
opportunity
and
I.
Think
code
gov
will
give
a
lot
of
visibility
into
some
of
these
projects
in
a
way
that
will
allow
us
to
see
what
else
is
happening
in
the
federal
government,
both
for
ourselves
to
see
what
other
sister
agencies
are
doing,
but
also
for
the
public.
The
people
to
see
is
actually
be
infant
source
communities
that
have
grown
up
around
federal
research
and
national
laboratory
research.
A
So,
to
give
you
an
idea
of
what
that
looks
like
our
github
com,
/
llnl
is
our
organization
and
we
actually
have
again
a
long
history
of
repositories.
The
moment
we
have
just
over
90
repositories
there,
where
we
have
source
code
in
the
last
year,
we've
actually
added
40
of
the
repositories
there
before
that,
our
first
one
that
showed
up
on
github
was
created
in
2008.
It's
recently
been
moved
into
this
organization
and
you'll
see
a
wide
variety
of
code.
There's
there's
quite
a
different,
quite
a
few
different
projects.
A
You
know
it's
a
lot
of
C
and
C++.
These
are
the
things
that
you're,
typically
using
HP,
see
that
you're
doing
these
simulations.
For
you
want
that
performance.
You
want
every
iota
of
performance,
but
of
cropping
up
like
Python
and
shell
to
actually
do
the
sorts
of
integration
and
glue
projects
that
tie
all
these
things
together.
Spac
is
a
Python
project.
A
We
have
some
shell
code
for
actually
doing
build
and
test
integration,
we're
very
heavily
involved
in
the
build
back
community
to
actually
build
and
test
these
different
systems
and
configurations
and
combinations
on
many
different
compilers
and
platforms.
We're
working
very
closely
with
the
open
source
community
and
we're
very
excited
about
the
opportunities
that
things
like
the
federal
source
code
policy
will
make
available.
A
A
We
started
some
of
this
work
in
December
and
we've
actually
started
moving
and
migrating
some
of
our
posit
stories
that
have
sat
in
svn
or
sad
as
tar
balls
sitting
on
a
website
which
is
obviously
the
way
developers
want
to
work
with
them
in
2016
and
we've
been
migrating
them
and
moving
them
into
these
sorts
of
environments
that,
in
general,
the
open
source
community
has
adopted
long
ago
right.
We
have
what
was
a
hundred
and
fifty
actions
that
are
available
through
github.
To
actually
take
advantage
of.
A
A
Project
is
something
that
we're
able
to
use
with,
along
with
Jekyll
and
mark
down,
to
actually
build
our
websites
and
build
some
of
our
documentation,
both
internally
as
well
as
externally.
So
again,
I
mentioned
some
of
this
work
started
just
before
the
source
code
policy
and
one
of
the
early
things
that
we
wanted
to
do
again.
We
have
a
long
history
here
and
it's
and
it's
going
to
take
us
some
time
to
get
everything
sort
of
pointed
in
the
same
direct
visual
eyes
and
displayed
in
the
in
the
same
way.
A
But
we've
actually
seen
is
we
have
currently
about
eight
major
organizations
on
github
that
are
associated
with
Lawrence
Livermore.
Think
places
like
the
sgf
that
I
mentioned
the
chaos
organization,
which
did
the
the
high
availability
operating
system
the
precursor
to
the
toss
operating
system
that
we
use
today.
The
lnl
organization
is
our
is
our
primary
one.
A
So
we
have
organizations
like
the
flux
framework,
where
it's
sitting
with
its
own
collection
of
plugins,
but
how
to
actually
visualize
that
and
put
a
put
a
lens
on
that,
so
that
you
can
actually
find
the
connections
between
all
these
things
has
been
something
that
we've
actually
been
able
to
use:
github
com
and
again
github
pages
and
Jeckle-
to
build
a
lens
into
that.
So
we
have
a
front
end
portal,
software,
lnl
gov,
and
the
idea
here
is:
there
is
one
site
where
you
can
go
to
look
into.
What
are
all
the
different
repositories
right?
A
The
github
com
search
will
be.
Will
get
you
part
of
the
way
there
you
can
do
things
like
searching
in
code.
You
can
do
things
like
filtering
down
a
repository
name,
but
it
doesn't
at
the
moment
give
some
of
the
clearer
you
know.
Building
to
it.
Narrowing
down
to
things
like
I
want
to
look
for
code
that
has
this
particular
language.
That's
maybe
in
this
particular
domain.
Maybe
it's
HPC,
maybe
it's
multiphysics!
Maybe
it's
matt
simulation
something
else
and
how
to
actually
drill
down
to
that
this.
A
These
are
problems
now
that
the
federal
source
code,
work
and
code
cover
at
empty
to
solve
by
having
a
central
location
where
all
the
federal
agencies
can
actually
put
their
code
repositories
their
code
inventories
and
make
them
up.
This
is
some
work
that
we
did
again
kind
of
in
parallel
or
just
before,
to
actually
put
a
lens
onto
our
own
repositories
in
our
own
source
code
internally
and
make
that
visible
externally.
A
So
I
won't
belabor
this
too
much
as
we
talked
about
it
quite
a
bit
in
the
keynote,
but
the
federal
source
code
policy
is
one
of
these
great
new
things.
There
is
a
lot
of
change
happening
in
the
government
around
open
source,
consuming
open
source
developing
and
producing
open
source,
as
well
as
khan's
talks
about
inner
sourcing.
A
So
yesterday
there
was
a
training
session
that
talked
and
dealt
with
how
to
inner
source
so
essentially
internal
open
source
development,
where
you
have
projects
internally
that
you
open
up
to
the
rest
of
your
employees
to
work
on
to
collaborate
on
to
suggest
ideas
on
to
fork,
to
pull
request
to
get
all
these
things
together
and
I.
Think
the
the
federal
policy
here
gives
a
lot
of
avenues
to
start,
laying
the
groundwork
for
government
to
in
a
lot
of
wit,
to
catch
up.
A
You
know
we
are
a
bit
behind
in
terms
of
what
the
open
source
community
is
doing,
and
what
this
allows
us
to
do
is
all
get
oriented
in
the
same
direction.
To
start
to
take
advantage
of
some
of
these
resources
like
github
com,
the
integrations
that
it
affords
and
the
work
being
done
at
other
agencies
since
we're
a
national
laboratory
again
growing
out
of
the
Manhattan
Project
age,
where
no
one
talked
to
anyone
about
what
they
were
doing.
No
one
knew
anything.
A
That's
that's
really
a
lot
of
what
the
culture
still
is
today
in
government,
you
have
groups
inside
of
various
different
agencies,
our
own
and
other
agencies
that
don't
necessarily
talk
to
each
other.
They
don't
have
a
way
to
see
into
what
each
other
are
doing
and
as
al
von
mentioned,
this
leads
to
a
lot
of
duplication
right.
A
I
think
we're
going
to
have
some
great
developments
and
I
think
there'll
be
a
lot
of
new
upcoming
work
coming
out
of
that,
as
the
pilot
rolls
out
as
the
code
gov,
it
rolls
out
in
the
next
few
months
and
as
agency
is
start
to
get
aligned
with
each
other
and
start
to
get
some
visibility
into
what
each
other
are
doing
so
currently
there's
this
is
this
is
a
small
slide.
I
had
to
cut
it
down.
The
original
list
was
about
1,800
government
agencies
across
the
world
that
have
organizations
on
github.
A
So
here
it's
a
lot
of
us
organizations,
but
there's
a
lot
of
work
being
done
in
other
countries.
Uk
has
a
lot
of
work.
There's
a
talk.
I
believes
a
keynote
tomorrow
on
what
the
UK
is
doing
with
open
source
and
open
government
open
data.
These
are
the
the
terms
that
again
very
prominent
in
the
open
source
community
they're
not
for
most
of
but
in
the
federal
world
in
the
government
world.
A
lot
of
this
is
sort
of
new
to
many
many
folks
involved,
and
so
we're
working
out
that.
A
A
That's
grown
up
inside
of
github,
so
one
of
the
things
that
Chris
mentioned
was
these
these
forums
and
while
there's
not
a
forum
necessarily
for
government
at
the
moment
in
the
way
that
he
showed
there
is
actually
a
community
growing
up
inside
of
github,
where
government
employees
are
able
to
meet
share
best
practices
to
share
information
like
hey
as
anyone
worked
on
the
security
policy,
getting
the
security
approval
for
running
a
particular
piece
of
software.
You
know
every
agency
repeating
this
process
over
and
over
and
over
again
is
saving
no
one.
Anything
really.
A
A
The
lnl
open-source
Twitter
account
is
actually
sending
out
notices
a
newer
one
where
we're
sending
out
information
about
the
work
that
we're
doing
about
new
projects
that
we
have,
and
we
very
much
welcome
those
that
interested
in
getting
involved.
If
you
have
a
project
that
you
think
solve
some
problem,
if
you
have
some
sort
of
expertise
in
file
systems
or
compilers,
or
just
how
to
make
someone's
life
easier
a
lot
of
times
in
these
hpc
systems,
we
have
a
lot
of
over
complexity.
Sometimes
that's
good!
A
There's
a
lot
of
complexity
in
the
system,
but
a
lot
of
times
there's
a
much
simpler
approach
that
the
rest
of
the
community
has
found
that
we
could
take
advantage
of.
So
this
point,
I
say
thank
you
and
open
it
up
for
questions.
There
is
a
microphone
if
you
can
come
up
and
ask
you
the
Mike,
and
we
have
about
six
minutes.
A
It's
a
good
question,
so
so
what
are
the
challenges
around
moving?
Do
II
and
agencies
national
character
isn't
open.
First,
we
one
of
the
biggest
challenges
actually
revolves
around
the
type
of
work
that
we
do,
because
our
portfolio
is
so
diverse.
It
covers
everything
from
this
is
pure
computer
science.
A
These
things
can
become
fuzzy
for
developers
right,
especially
if
you're
working
in
both
areas
detangling
those
two
things
and
the
reality
is
that
we
have
a
lot
of
a
lot
of
guidance
for
what
it
means,
how
you
have
to
actually
get
something
relief,
whether
it's
at
the
end
of
the
day.
It's
generally
a
lot
of
bureaucracy
and
I.
A
Think
a
lot
of
the
work
that
we're
doing
right
now,
as
we
head
into
this
next
year,
is
going
to
be
breaking
down
those
barriers
and
trying
to
ease
the
processes
and
try
and
separate
out
things
that
are
clearly
can
be
open,
source
and
developed
in
the
open
and
things
that
do
need
that
extra
level
of
scrutiny.
Because
the
end
of
the
day
is
you
know
when
things
go
pup
on
on
the
classified
next,
we're
controlled
side
of
things.
People
go
to
jail.
So
it's
actually
something
that
you
know.
Employees
have
to
care
about.
D
A
Stick
public
API
is
so
I
think
I
think.
A
lot
of
this
is
agencies
actually
moving
to
embrace
api's
that
are
in
the
public
already
right,
rather
than
reinventing
our
own
wheels.
Some
of
the
work
that
Alvin
sort
of
alluded
to
through
the
through
the
issues
on
the
code
gov
website,
issue
tracker
on
github
are
things
around
API.
So
one
of
the
big
piece
there
is
that's
not
necessary,
going
to
be
the
source
code
for
all
the
government
or
where
the
source
codes
hosted
a
lot
of
times.
A
It's
still
going
to
be
hosted
internally
at
the
various
agencies,
but
creating
these
sort
of
open
API
is
to
get
the
data
into
and
out
of
code.
Gov
is
going
to
be
one
of
the
questions
that
well.
It
is
one
of
the
questions
that's
being
discussed
very
actively
on
the
issue.
Tracker,
so
I
think
those
those
sorts
of
API
is
continuing
to
work
together
and
also
embracing
open
standards
right
rather
than
developing
our
own
cryptography
or
transfer
protocols,
or
all
these
sorts
of
things.
A
There's
a
lot
of
industry
backing
behind
doing
these
things
already
and
the
reality
is,
and
history
has
been
out
that
the
proprietary
solutions
generally
don't
work
as
well.
They
don't
see
the
adoption,
they
don't
see
the
same
level
performance
and
they
don't
see
the
same
level
of
security.
Broadly
speaking,
you
know,
cryptography
is
a
big
people
and
I,
so
I
think
adopting
things
that
the
are
happening
in
the
open
source
and
not
reinventing
our
own
kills
will
be
one
of
the
main
uses
of
api's
I.
B
Beyond
from
source
graph
here,
thanks
for
the
great
talk
and
I'm
actually
live
blogging
it.
So,
thanks
for
sending
me
your
slides,
yeah
as
well,
I
was
wondering
if
you
had
any
like
stories
of
other
government
agencies,
some
of
the
logos
on
that
slide
kind
of
reaching
out
and
using
some
the
libraries
that
you
presented
in
this
talk,
and
you
know
how
is
that
interaction
been?
How
do
your
developers
work
with
their
developers
they
file
issues?
What's
that
conversation
like
yeah.
A
So
if
I
just
hit
on,
you
know
some
of
the
ones
that
I
showed
as
some
of
the
examples
we
have
things
like
slurm
these
job
schedulers
that
are
used
across
multiple
different
data
centers
and
one
of
the
benefits
of
being
out
on
get
that
not
every
agency
employee
has
to
get
accounts
on
every
other
agency
in
order
to
actually
work
on
the
code,
and
it
also
cuts
down
on
the
alternative,
which
is
essentially
tar,
something
up
and
send
it
an
email
right,
not
a
very
efficient
process.
It
turns
out
place.
A
Things
like
github
have
solved
that
problem,
largely
so
that
sort
of
sharing
and
actually
doing
things
in
the
open
on
github
eliminates
those
barriers.
We
also
have
some
things
that
are
more
sensitive,
that
we
collaborate
on
that,
where
you
know
where
users
we're
collaborators
do
have
to
get
accounts
on
each
other
systems
and
that's
often
one
of
the
biggest
slowdowns.
To
collaborations
since
we're
a
federal
agency,
we
live
on
the
bulb
fifth
we're
at
the
end
of
FY
16
octobre
one
starts
fiscal
year,
17
and
projects
often
times
only
run
for
one
year.
A
So,
if
any
any
delays,
and
actually
starting
that
collaboration,
can
can
really
hurt
the
developer,
what
the
developer
can
actually
get
done
in
terms
of
actually
working
and
collaborating
on
different
things.
I
think
you
know
there's
there's
a
lot
of
other
examples,
some
of
the
ones
I
should
here
on
the
slide
like
actually
the
NSA
has
a
fairly
popular
project
right
now
for
a
sort
of
operating
system
configuration
hardening
things
very
interesting
right,
it's
a
problem
that
we
all
face
and
we
all
learn
little
pieces
of
the
puzzle
and
we
should
put
those
all
together.
E
Ian
I'm
jamie
jones
from
github
government
team.
I
just
wanted
to
give
a
quick
note
for
people
in
the
audience
and
watching
at
home.
You
mentioned
government
people
getting
involved
in
github
government
at
github
com
is
actually
the
best
practices
community
you
were
mentioning
before,
which
is
open
to
anyone
with
a
state.
Local
or
international
government
account
to
come
and
collaborate
and
talk
about
what's
going
on,
so
that
can
answer
some
more
of
everyone's
questions,
as
well
as
helping
them
to
get
involved
so
easy
plug.
A
F
Hi,
thank
you
for
the
talk,
I'm,
a
UCLA
student,
so
in
universities,
one
of
the
growing
fields
is
using
analyzing
big
data
using
supercomputers,
and
you
mentioned
Sequoia.
As
you
know,
that's
a
very
strong
computer
and
I
know.
A
lot
of
companies
like
Nvidia
are
creating
efficient,
small
super
computer
or
server
racks
forties.
Where
do
you
see
the
future
of
that
be
in
like
a
university,
because
it's
kind
of
hard
to
you
know
it's
a
lot
of
money.
So
I
just
wanted
to
see
your
point
on
that
yeah.
A
I'll
answer
real,
quick
because
we're
running
out
of
time
but
and
we
can
continue
offline
but
I-
think
that'll
actually
open
the
doors
to
a
lot
of
these
universities,
but
they're
still
going
to
have
a
lot
of
the
same
problems
right
because
you
have
GPUs,
you
have
these
very
heterogeneous
resources.
You
have
larger
file
systems
if
you
actually
want
to
have
the
data
so
we're
going
to
run
into
the
same
albums
and
I
think
what
it's
going
to
do
is
move
the
you
know,
a
lot
of
the
software
that
we've
been
developing.
A
That
I
talked
about
from
this
is
a
problem
that
only
the
federal
government
does
because
they're
the
only
ones
that
have
a
billion
dollars
to
spend
on
a
supercomputer.
To
now,
like
you
said
university,
it's
an
industry
and
you
know
smaller
time,
operations
able
to
get
the
same
sort
of
computing
or
but
they're
still
going
to
run
into
a
lot
of
the
same
io
scheduling,
data
storage
issues
that
we
face.