►
From YouTube: New User Training: 01 Introduction to NERSC
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
All
right
so
welcome
everybody
I'm
happy
to
welcome.
You
here
looks
like
it's
a
little
bit
of
a
sparse
turnout
this
morning,
so
rainy
I
appreciate
everyone
coming
in
the
rain,
so
my
name
is
Rebecca
Hartman
Baker
I
am
the
leader
of
the
user
engagement
group,
and
so
today,
I'm
gonna
talk
with
you,
I'll
just
give
you
an
overview
of
nurse
and
what
we
do
and
sort
of
how
we
interact
with
users
and
how
we
expect
that
users
would
interact
with
that.
A
So
here's
my
agenda,
so
I'm
gonna,
give
you
a
little
introduction
to
nurse
and
who
we
are.
Why
we're
here
talk
about
the
hardware
that
we
have
the
software,
how
to
interact
with
nurse
and
then
overview
of
user
responsibilities
and
expectations?
Okay,
so
nurse
stands
for
the
national
energy
research,
scientific
computing
Center.
So
that's
why
we
call
ourselves
nurse
get
instead
of
the
whole
name
right
too
long,
so
it
was
established
in
1974.
It's
the
first
unclassified
supercomputer
Center.
A
Now
the
original
mission
of
nurse-
and
we
weren't
called
nurse
back
then,
was
to
enable
computational
science
to
study
magnetically,
controlled
plasma
experimentation.
That's
why
we
had
a
nurse
today.
Our
mission
is
to
accelerate
scientific
discovery
at
the
d.o.a
office
of
science
through
high
performance
computing
and
extreme
data
analysis.
Nurse
is
a
national
user
facility,
so
we
have
users
from
all
around
the
country
and
actually
all
around
the
world.
A
So
we
have
more
than
7000
users
and
800
projects
and
our
users
use
about
600
different
codes.
We
have
hundreds
of
users
who
are
active
daily
on
our
machines.
Our
allocations
are
primarily
controlled
by
the
Department
of
Energy,
so
so
80%
goes
to
what
we
call
our
cat
energy
research
computing
allocations
program.
That
sounds
convincing,
at
least
if
it's
not
right,
and
so
we
give
out
well,
they
give
out
about.
You
know
ten
thousand
to
ten
million
our
words.
Actually
we
have
some
hundred
million
our
awards.
A
Now
users
submit
a
proposal
and
then
do
a
program
managers
choose
from
those
proposals,
so
the
proposals
go
to
a
specific,
do
a
program
manager.
So
if
you're
you
know,
if
you're
in
nuclear
physics,
let's
say
it
would
go
to
the
nuclear
physics
program
managers.
So
eighty
percent
goes
to
the
air
cap
Awards.
A
Ten
percent
goes
to
the
do
e
Oscar
leadership
computing
challenge,
which
we
also
call
a
LCC,
which
is
an
acronym
with
an
acronym,
so
Oscar
I
never
can
remember
what
it
stands
for,
but
it's
like
advanced
scientific
computing,
research,
I,
think
and
then
over.
The
remaining
10%
is
in
our
nurse
reserve
and
we
use
that
for
overhead
if
somebody's,
if
we
need
to
refund
somebody's
job
that
didn't
run,
we
used
our
overhead
for
that.
We
use
it
for
education
and
training.
A
We
use
it
for
Directors
Awards
for
like
right
now
we
have
this
scale,
science
Awards,
so
that
comes
through
our
reserve.
Okay.
So
from
the
do
a
point
of
view,
this
is
how
you
I
was
right:
advanced
scientific
computing,
researchers,
Oscar
domain
okay,
so
this
is
kind
of
a
pie
chart
of
how
our
allocations
were
distributed,
based
on
last
year's
distribution
of
hours.
So
you
can
see
they
go
to
a
wide
variety
of
different
areas.
A
So
probably
some
of
the
biggest
ones
are
BER
and
BES.
So
that's
bait
by
a
biological
and
environment
research
is
BER
and
BES
is
basic
energy
sciences,
and
so
those
are
kind
of
broken
down
into
several
different
sub
areas
in
this
pie
chart.
So
you
can
see
we
have
users
from
a
wide
variety
of
different
scientific
disciplines
that
are
using
our
machines.
A
I
mentioned
before
that
we
have
over
600
codes
that
run.
If
you
look
here,
the
top
code
here
is
vast
and
vast
accounts
for
more
than
10%
of
all
of
the
hours
that
are
run
on
our
machines,
but
the
top
10
codes
make
up
half
of
our
workload
and
then
the
top
25
codes
make
up
two-thirds
of
our
workload
so
well,
we
do
have
a
lot
of
people
from
a
lot
of
different
areas
using
our
machines
and
we
have
more
than
600
coats
that
are
being
used.
A
Ok,
so
we
are
very
focused
on
science,
nurse
users
actually
produce
and
publish
more
than
any
other
any
other
Center
in
the
world.
We
have
about
2,000
publications
per
year
that
a
site
nurse
so
yeah.
We
actually
have
probably
more
publications
than
any
other
Center
in
the
world
as
far
as
we
know,
so
we
love
our
users
and
we
want
to
help
you
all,
but
we
need
you
to
help
us
to
help
you.
A
So
if
you
don't
acknowledge
nurse
in
your
publications,
then
nobody
will
know
how
useful
nurse
gives
and
then
we
won't
get
as
much
money
and
then
we
won't
be
able
to
provide
services
to
you
anymore,
so
be
sure
to
acknowledge
us
in
your
publications.
Also,
we
love
user
success
stories.
So
if
you
have
any
user
success
stories,
like
really
cool
publications
about
your
super
cool
science,
then
please
send
send
us
your
links
to
your
publications.
We
might
even
interview
you,
we
could
make
a
article
about
it.
It's
be
super
exciting.
A
It
would
be
a
win-win
for
everyone.
So
please,
when
you
use
nurse,
acknowledge
us
and
then
give
us
your
success
stories,
because
we
want
to
hear
about
it.
Okay,
so
we
have
a
lot
of
systems
here
at
nurse
our
flagship
system
is
Cory,
so
Cory
is
currently
in
the
top
ten
of
the
most
powerful
supercomputers
in
the
world.
A
I
can't
remember
where
we
are
right
now
exactly,
but
Cory
has
two
different
types
of
nodes,
so
it's
got
about
2000
Haswell
nodes
at
about
990,
300,
Kol
nodes
and
we'll
talk
more
about
these
later
Edison
is
our
other
big
machine.
Edison
has
five
thousand
five
hundred
and
seventy
six
nodes,
and
so
both
of
these
machines
have
scratched
systems
that
they
are
attached
to.
A
Edison
is
also
attached
to
Cory's
scratch
system
and
then
Cory,
of
course,
has
a
burst
buffer,
which
you'll
learn
more
about
this
afternoon
very
neat
resource.
So,
in
addition
to
having
our
machines
there,
we've
got
some
clusters.
We've
got
a
cluster
called
gene
pool
that
houses
our
PDS
F
cluster,
our
hei
resources
in
there
we've
got
other
resources
like
visualization
and
analytics
resources.
We've
got
data
transfer
nodes,
we've
got
science
gateways,
we're
all
connected
into
es
net
and
then
also
I
should
mention.
We
have
our
global
file
systems
very
powerful,
very
large
capacity
in
particular.
A
A
So
when
it
comes
to
our
HPC
systems,
Edison
Edison
is
great.
Edison
is
a
large
and
stable
machine.
It's
not
the
new
hotness
anymore,
so
it
has
shorter
accuse.
Then
Cory
does,
and
it
also
has
a
lower
charge
factor
we'll
talk
more
about
what
that
means
later,
but
basically
it
means
you
can
get
more
CPU
hours
for
cheaper
if
you
go
on
Edison
than
if
you
go
on
Cory
now,
Cory,
like
I
mentioned
before,
has
two
different
types
of
nodes.
It's
got
the
has
little
nodes
and
it's
got
the
KL
nodes.
A
A
We
have.
We
allow
single
core
jobs
on
those
Haskell
notes,
and
we
have
longer
wait
time
limits
for
some
smaller
jobs.
Now
the
KL
nodes,
those
are
really
the
new
hotness.
They
are
the
best
that
we
have
they're,
really
good
for
performance.
The
issue
here
is
that
the
the
KL
architecture
and
I
think
we'll
learn
more
about
this
later.
The
candle
architecture
has
a
lot
of
very
small
low
powered
cores,
but
a
lot
of
them
all
right.
It's
got
68
cores
per
node
versus
these
others
that
have
maybe
32
cores
per
gram.
A
So
if
you
can
exploit
a
mini
core
architecture
like
that,
then
these
core
ekl
nodes
are
perfect
for
you,
and
this
is
where
we
like
people
to
run
all
the
really
large
stops,
because
remember
we
have
more
than
nine
thousand
nodes
of
this.
So
if
you
can
run
across
nine
thousand
notes
as
awesome
as
where
we
want
you
to
be
okay,
so
I
mentioned
before
we've
got
some
pretty
awesome
file
systems,
so
we've
got
different
types
of
file
systems
that
nurse
we've
got
global
file
systems,
local
file
systems,
and
we
have
a
long-term
storage
system.
A
A
It
is
not
to
perform
well
in
parallel
jobs,
so
we
encourage
people
to
not
run
your
parallel
jobs
from
your
home
directory.
You
have
a
quota
on
that
home
directory
and
we
can't
change
it
if
we
just.
We
just
won't
change
it,
because
that
the
purpose
of
home
is
primarily
for
storing
some
data,
such
as
source
code
or
shell,
scripts,
okay,
or
maybe
some
binaries,
but
that's
about
all
that
we
really
want
you
to
use
home
for
because
we've
got
other
file
systems
that
are
way
better
for
other
purposes
that
you
would
be
using.
A
So,
in
addition,
we've
got
a
project
space
for
everyone,
so
it's
mounted
also
like
home.
It's
also
mounted
on
all
of
our
platforms.
It
has
medium
performance
and
parallel
jobs.
We
can
change
the
quota
there.
We
can
extend
your
quota,
it
has
a
snapshot
backup.
So
it's
got
a
seven-day
history,
just
just
like
home
did
I
forgot
to
mention
that.
So
what
that
means
is
that
if
you
accidentally
delete
a
file,
you
can
go
back
into
the
snapshot
and
you
can
retrieve
that
file
within
seven
days
and
prod
the
project
system.
A
Ok,
then,
we've
got
some
local
file
systems,
so
we've
got
our
scratch
file
systems,
so
these
are
large
temporary
storage
systems.
There
there's
a
local
scratch
on
Edison
and
then
quarry
scratch,
which
would
normally
just
be
local
to
Cori.
We've
also
mounted
in
on
Edison.
So
you
can
you
can
access
quarries,
scratch
from
either
machine,
but
you
can
only
access
the
Edison
scratch
file
systems
from
Edison.
These
are
optimized
for
read/write
operations,
but
not
for
storage.
A
Excuse
me:
we
do
not
backup
the
scratch
systems
and
in
fact
we
have
a
purge
policy.
So
if
you
leave
your
data
on
for
12
weeks
on
one
system,
it's
eight
weeks
on
another
212,
if
you
leave
it
there
without
doing
anything
to
it
without
accessing
it
without
writing
to
it.
We're
gonna
delete
it.
But
scratch
is
really
perfect
for
staging
your
data
and
performing
your
computations.
A
That's
where
we
want
you
to
do
those
things
and
then
after
you're
done,
you
need
to
clean
up
after
yourself
and
put
put
the
output
into
a
more
appropriate
storage
system.
So
another
one
is
the
burst
buffer,
so
burst.
Buffer
is
sort
of
a
temporary
per
job
storage
and
it's
really
a
high
performance
file
system
made
out
of
SSDs.
So
it
is
really
really
really
fast
for
readwrite
types
of
operations.
It's
only
available
on
quarry,
that's
one
of
unique
features
of
quarry
and
it
is
perfect
for
getting
really
good
performance
from
I/o
constrained
code.
A
A
Then,
finally,
we've
got
our
HP
SS
system,
so
that
stands
for
high
performance
storage
system.
It
is
archival
storage
for
eat-in,
frequently
accessed
data,
so
it
is
sort
of
it's
sort
of
a
hierarchical
storage
system.
So,
on
the
front,
we
have
this,
these
high-performance
disk
arrays
and
that's
kind
of
where
your
data
goes
when
it
first
gets
ingested
and
but
then,
after
a
while
it
hasn't
been
accessed.
It
goes
into
the
back
end,
which
is
a
bunch
of
tape
drives
now
y'all
may
be
seen
tape.
Are
you
kidding
me
Rebecca
like?
A
Why
would
you
use
tape?
Tape
is
actually
really
great.
It's
really
low
cost.
It
doesn't
require
in
any
electricity
or
power
to
maintain.
It
just
needs
to
be
in
a
safe
environment
for
tapes,
and
so
that's
why
we
use
it,
but
for
more
information
about
HP
SS,
again
we'll
have
later
presentations,
you'll
get
to
learn
a
lot
more
about
it,
okay,
so
using
nurse
filesystems.
This
is
this
is
an
analogy
that
I
like
to
use
with
people
so
computing,
it's
kind
of
like
baking.
Right
like
you,
have
these
baking
ingredients?
That's
your
input!
A
You
have
this
output,
which
is
like
a
cake.
Let's
say
we're,
gonna
bake
a
cake.
Okay
and
the
computer
is
kind
of
like
the
oven
right.
That's
where
you
like,
where
all
the
good
stuff
happens.
Right.
Where
are
you
you're?
Actually,
taking
these
strange
ingredients
putting
them
all
together,
putting
them
in
the
oven
and
out
comes
a
delicious
cake.
Okay,
so
I
would
liken
the
home
and
project
systems
to
your
pantry
or
your
fridge
right.
That's
where
you
store
your
ingredients
for
your
baking
right.
Hp
SS
is
like
your
freezer.
A
That's
where
you
have
like
the
frozen
blueberries
or
something
that
you
don't
use
very
often.
Sometimes
you
need
them,
so
you
would
bring
about
and
then
scratch
is
your
kitchen
countertop.
Okay,
that's
where
you're
gonna
stage
all
of
your
ingredients,
you're
gonna,
put
them
all
together
and
then
you're
gonna
bake
them
in
the
oven
and
then
you're
gonna
stage
them
out
onto
the
countertop
again
right.
A
All
right,
I
already
said
exactly
this.
So
when
you're
baking,
you
take
all
of
your
ingredients
out
of
the
pantry
right
and
you
put
it
on
the
countertop
you're
like
okay
I
need
my
flour.
I
need
my
baking
soda.
You
know
I
need
my
buttermilk,
whatever
I
put
them
all
out
on
the
counter,
all
right
and
then
I
dig
into
them
and
I
mix
them
up
in
my
mixing
bowl
or
whatever.
Then
I
put
it
in
my
pit,
my
cake
pan
right.
So
it's
the
same
thing
when
you're
doing
your
computations
right.
A
So
you've
got
your
data.
You
need
all
of
this
data
in
order
to
learn
what
you're
gonna
learn
from
your
computations,
you
put
it
all
in
there,
you're
already
you're
all
set
from
when
it
finally
runs.
Okay.
So
after
baking,
you
really
should
clean
up
after
yourself.
So
in
this
case,
instead
of
it's
your
own
kitchen
and
you
can
attract
all
the
roaches
you
want-
and
nobody
cares
right
here.
A
It's
like
it's
a
public
kitchen
and
you
need
to
clean
up
after
yourself,
because
we
only
have
a
finite
amount
of
counter
space,
and
so
you
know
somebody
else
is
gonna
need
to
use
that
space.
So
it's
okay
to
let
your
cake
cool
on
the
kitchen
counter,
but
you
need
to
leave
a
space
clean
for
the
next
user.
Okay.
A
So,
after
a
while,
if
you
don't
clean
up
we'll
clean
up,
but
we're
not
gonna
clean
up
in
the
way
that
you
like,
because
what
we're
gonna
do
is
we're
just
gonna
get
the
trash
can
and
we're
just
gonna
dump
everything
into
it.
And
that
includes
your
cake.
Okay,
so
don't
make
that
mistake.
Don't
leave
your
output
on
scratch
for
12
weeks
and
expect
that
it's
gonna
still
be
there
when
you
come
back
because
it
won't
because
we'll
get
rid
of
it.
A
Okay,
so
let's
talk
about
software,
so
both
of
our
machines
are
crazy
for
computers
and
their
OS
is
a
version
of
Linux
that
is
optimized
by
Cray.
Now
the
machine
on
machine.
We
provide
compilers,
three
different,
compiling
environments
and
we'll
learn
more
about
these
things
in
more
detail
as
the
day
progresses.
We
have
many
libraries
that
are
available
that
some
of
them
are
provided
by
Cray.
A
Others
are
provided
by
us
at
Nernst
and
then
we
also
add
nurse.
We
provide
a
lot
of
applications,
so
we
compile
and
support
many
different
software
packages
for
our
users
and,
like
I,
said
there
will
be
more
details
on
this
in
the
later
presentations.
This
is
sort
of
an
overview,
so
one
big
thing
that
we
do
is
we
provide
a
lot
of
chemistry
and
materials
applications.
A
That
chemists
use
a
lot.
Okay,
so
switching
gears
again
we're
gonna
talk
about
how
to
interact
with
us
at
nurse.
So
these
are
three
primary
ways
in
which
you
will
interact
with
us.
So
the
first
thing
is
you'll
interact
with
nurse
consulting
and
possibly
with
the
nurse
operations
folks
and
then,
hopefully
with
the
nurse
user
group.
So
we
here's
our
consulting
team.
We
are
composed
of
people
from
three
different
groups:
user
engagement.
That's
my
group!
That's
me
at
the
top
there,
these
little
tiny
pictures,
the
application
performance
group
and
the
data
science
engagement
group.
A
So
we
all
comprise
this
consulting
team.
We
all
we
all
spend
time
talking
to
users
answering
your
tickets,
things
like
that,
so
in
2017
we
handle
seventy
four
hundred
tickets
from
two
thousand
three
hundred
forty
two
unique
users.
Okay.
So
it's
a
lot
of
work
for
us,
a
lot
of
different
areas
that
people
ask
us
about,
so
primarily
software
and
running
jobs.
Those
are
our
two
big
ones.
A
A
A
So
the
only
thing
we
ask
in
exchange
really
is
again
help
us
to
help
you
I,
remember
I
just
said
we
had
7400
tickets
in
a
year
so
provide
us
with
specifics.
What
is
the
problem
that
you're?
Having
what
machine
is
it
on?
When
did
it
happen,
what
modules
were
loaded?
How
did
you
try
to
fix
it
or
work
around
it?
So
if
you
just
did
this
message-
and
you
say.
A
My
code
died,
okay,
well,
what
code?
What
happened
when
it
died?
How
did
you
compile
it?
You
know
what
error
messages
did
you
get?
You
know
right
like
we'll
probably
reply
and
we'll
ask
you
all
those
questions,
but
sure
would
be
easier
if
the
first
time
when
you
send
us
message,
you
say,
I
was
running
vast
and
I
submitted
with
this
particular
input
file
and
here's.
My
my
batch
script
and
I
got
the
following
error
and
it
was
on
quarry
and
it
was
job
number
such-and-such
right.
A
Okay,
so
we've
also
got
operation
staff
who
are
on
site
all
the
time,
all
the
time,
24/7
365
or
366
days
per
year,
and
they
supervise
the
operation
of
a
machine
room.
They
make
sure
that
bad
things
don't
happen
to
to
these
machines,
which
are
worth
tens
of
millions
of
dollars,
so
I
mean
they're.
Even
there
on
Christmas
they're,
even
I
mean
they're
there
on
Thanksgiving
they're,
their
holidays,
they're
there
at
2:00
a.m.
A
all
the
time,
so
our
operations,
folks
they
answer
the
phone
and
they
will
forward
it
to
us
consultants
during
business
hours,
if
applicable.
So
so
you
might
talk
to
them.
If
you
ever
call,
although
I
see
a
lot
of
young
faces
around
here,
I
think
young
people
don't
like
calling
which
that's
good
I
feel
the
same
way
so
operations.
They
know
exactly
what's
going
on
with
the
machines,
though,
so
they
can
actually
be
very
helpful
with
some
tasks,
so
they
can
help
you
reset
your
password.
They
can
help.
You
kill
some
jobs.
A
They
can
make
limited
changes
to
your
reservation.
If
you
have
a
reservation,
that's
running
so
operations,
don't
discount
them
and
just
say
oh
I,
just
must
be
to
a
consultant
whose
operations
can
really
help
you
out
there
they're
a
bunch
of
smart
folks
down
there,
I'm
always
impressed
whenever
I
talk
with
them.
Okay.
A
Okay,
so
then
we've
got
the
nurse
user
group.
So
this
is
a
community
of
nurse
users.
As
I
said,
we
have
more
than
7,000
users
they're
a
great
source
of
advice
and
feedback
for
us
at
nurse,
so
we
asked
them
their
opinions
and
they
tell
us-
and
that's
great
I-
mean
there's
nothing
better
than
being
able
to
just
opine
and
not
having
to
actually
do
anything
about
it
right
and
that's
what
they
do.
A
So
we've
got
an
executive
committee,
that's
three
representatives
from
each
office,
so
remember
the
DOA
offices
that
I
told
you
about
that
that
that
used
the
Machine.
So
we've
got
three
representatives
from
each
office
and
then
we
also
have
three
members
at
large
and
then
we
also
have
monthly
teleconferences,
hosted
by
nurse
usually
on
the
third
Thursday
of
the
month,
11:00
a.m.
to
noon.
So
we
already
had
it
last
Thursday.
It
was
the
third
Thursday
of
the
month,
okay,
so
you
as
a
user.
A
What
do
we
expect
out
of
you
couple
of
things
so,
first
of
all
be
kind
to
your
neighbor
users
right,
don't
abuse
the
shared
resources
so,
for
example,
everybody
logs
into
the
login
node,
so
don't
just
go
on
the
login
node
and
just
take
all
resources
and
run
something
that's
really
computationally
or
memory
intensive
on
the
login
nodes.
That's
not
nice!
Okay,
use
your
allocation,
smartly
or
wisely.
A
We
have
limited
allocations
right,
I
mean
there's
limited
time
and
number
of
notes
right.
So
if
you
blow
through
your
whole
allocation,
there's
not
a
lot,
we
can
do
for
you,
so
use
it
smartly
and
don't
don't
abuse
it
pick
the
right
resource
for
your
job
and
your
data,
so
small
jobs
are
really
great.
On
Cori
has
well
they're,
not
so
good.
Well,
they're,
okay
on
Edison
at
this
point,
they're
really
not
so
good
on
the
on
the
Cori
KL
nodes,
that's
really
for
the
big
job,
so
that
would
be
my
recommendation.
A
Another
thing
is
back
yourself
back
your
stuff
up.
So
remember.
If
you
leave
your
cake
on
scratch
for
12
weeks,
we'll
throw
it
away
so
don't
do
that
acknowledge
nurse.
Can
your
papers
so
that
we
can
continue
to
get
funding
and
be
sure
to
pay
attention
to
security.
So
don't
share
your
account
with
other
people.
Please
don't
do
that
we'll
have
to
disable
your
account
if
we
find
out
that
you're
doing
that.
So
don't
do
that.
It's
bad
plan
all
right.