►
From YouTube: 1. Introduction to NERSC
Description
Learn the basics about NERSC: our resources, the way we operate, and how you can help us help you!
Slides for all sessions can be downloaded from here: https://www.nersc.gov/users/training/events/new-user-training-june-21-2019/
A
All
right
so
good
morning,
everybody
good
morning,
local
people
good
morning.
Yes,
thank
you
good
morning,
people
out
there
in
the
ether-
and
you
don't
have
to
say
good
morning,
that'd
be
too
much
work
for
you,
okay,
so
my
name
is
Rebecca.
Hartman
Baker
I
lead
the
user
engagement
group
here
at
nurse.
So
if
you're,
a
nurse
user
I
assume
you
all,
are
you
get
an
email
from
me?
Every
Monday?
That's
like
my
claim
to
fame
like
when
I
go
to
conferences.
A
I'm
walking,
I
have
like
my
name
tag
and
people
are
like
I
know
you,
you
send
me
an
email,
every
Monday,
yes,
I!
Do
yes,
okay,
so
in
my
presentation,
I'm
gonna
give
you
an
introduction
to
nurse
we're
going
to
talk
about
the
hardware
that
we
have
here
at
nurse,
the
software
that
we
have
and
then
we're
going
to
talk
about
interacting
with
nurse
how
you
can
interact
with
us
and
then
some
information
about
user
responsibilities
and
expectations.
A
Okay,
so
nurse
stands
for
national
energy
research,
scientific
computing,
Center.
So
that's
why
we
call
ourselves
nurse
okay,
so
we
were
established
in
1974
as
the
first
unclassified
supercomputer
Center.
Our
original
mission
was
fairly
specific,
so
we
were
originally
created
to
enable
computational
science
as
a
complement
to
magnetically
control
plasma
experiments.
A
So
today
our
mission
has
really
branched
out
from
that.
So
today
we
are
here
to
accelerate
scientific
discovery
at
the
do
II
office
of
science
through
high
performance
computing
and
extreme
data
analysis.
So
nurse
is
a
national
user
facilities.
That
means
we
have
users
from
all
over
the
country,
we're
open
to
anyone
in
the
country
and
actually
we
have
a
number
of
users
internationally
as
well.
So
we
have
about
7,000
users
as
of
last
year
and
we
have
800
ish
projects
and
these
seven
thousand
users
in
800
projects
use
more
than
600
codes.
A
We
have
hundreds
of
users
who
are
active
every
day.
Our
allocations
of
time
on
these
machines
and
resources
is
primarily
controlled
by
the
Department
of
Energy
itself,
so
80
percent
of
our
time
goes
towards
the
do
e
annual
production
Awards,
which
is
called
our
cap,
which
stands
for
an
energy
research
computing
allocations
program.
So
that's
way
too
long.
That's
why
we
call
it
or
cap,
and
so
people
can
get
anywhere
from
ten
thousand
to
ten
million
hours,
sometimes
even
a
hundred
million
hours
on
our
machines
that
they
can
use
for
a
full
year.
A
So
this
is
a
proposal
based
system
and
the
successful
awardees
are
chosen
by
do
e
allocation
program
managers.
So
an
additional
ten
percent
is
given
by
the
do
e
Oscar
leadership
computing
challenge.
So
this
is
for
more
sort
of
high
risk,
but
potentially
high
payback
sort
of
projects,
and
then
10%
is
our
reserved
for
nurse
because
we
do
stuff
with
it.
We
we
have
like
strategic
projects.
We
have
overhead
for
when
we're
doing
trainings
or
something
like
that.
So
that's
that's
kind
of
how
the
breakdown
goes
here.
A
So
now,
like
I,
said,
the
Department
of
Energy
is
the
ones
who
decide
mostly
what
gets
on
nursing
supercomputers.
So
it's
specifically
it's
in
the
office
of
science.
So
there's
a
couple
of
different
offices
within
the
OE.
One
of
them
is
the
National
Nuclear
Security
Administration.
So
they
do.
They
do
like
our
nuclear
arsenal.
A
A
So
there
are
six
offices
plus
the
small
businesses
innovation
research
office.
So
these
six
offices,
each
they
each
get
kind
of
like
a
portion
of
nurse
time
to
give
to
their
users
or
to
whoever
they
want
to
give
it
to
so.
This
is
kind
of
the
breakdown
you
can
see
in
this
pie
chart
the
breakdown
of
the
types
of
science
that
gets
done
at
nurse
and
how
much
of
the
allocations
that
they
get.
A
Okay,
so
I
mentioned
before
that
we
have
over
600
coats
running
at
nurse-
it's
probably
a
higher
number
by
now,
but
you
can
see
in
this
breakdown.
We
have
this
breakdown
of
what
codes
are
running
at
nurse,
so
600
sounds
pretty
overwhelming,
but
honestly,
there's
about
ten
codes
that
make
up
close
to
50%
of
our
workload
and
then
there's
15
more
so
total
25
that
make
up
about
two-thirds
of
our
workload.
A
So
then,
there's
just
a
bunch
of
other
ones
that
people
are
running
so
nurse
has
a
huge
focus
on
science,
so
our
users
they
published
more
than
any
other
Center
in
the
world.
As
far
as
we
know,
we
have
about
2,500
publications
per
year,
so
in
2018
we
had
14
publications
just
in
nature
31
in
Nature
communications,
11
in
science-
and
you
know
these
single
word
journals
are
like
the
most
prestigious
ones
right.
So
that's
pretty
amazing.
We
also
have
six
Nobel
Prize
winning
users,
and
so
we're
really
proud
of
that.
A
So
something
something
that
you
should
understand
is
that
you
know
we're
funded
by
the
United
States
government
right
and
so,
if
nobody
ever
tells
the
United
States
government,
how
great
we
are
right,
then
they'll
stop
funding
us
right
if
nobody
tells
them.
Oh,
we
really
used
nurse
and
it
was
invaluable
to
our
research.
Then
maybe
they
won't
find
us
anymore,
and
that
would
be
bad.
A
I
would
be
sad,
I
would
be
unemployed
and
then
you
all
would
be
sad
too,
because
you
wouldn't
have
a
computer
to
use
so
be
sure
to
acknowledge
nurse
in
your
publications.
Because
that
way,
then
we
can
say:
look
look
at
these
great
publications,
this
great
science,
it's
coming
out
of
nurse
right,
so
we
have
a
little
blurb
that
we
like
you
to
put
in
there.
That's
a
you
know:
I'm
not
going
to
read
it
to
you,
because
you're
y'all
can
read,
but
it's
available
on
our
website.
A
So
you
just
want
to
put
that
in
your
research
paper
like
in
the
acknowledgments
section,
and
then,
if
you
have
some
super
cool
science
that
you
did
with
nurse,
send
them
to
us.
Let
us
know,
because
we
love
to
publicize
stuff,
that
our
users
do,
because
most
of
us
are
working
here
because
of
the
science.
You
know
we're
not
believe
it
or
that
we're
now
working
here
for
the
big
bucks.
I
know
it's
incredible
you
can.
A
A
A
We
named
our
machines
after
great
scientists,
so
Edison
Cory
is
named
after
Kirti
Cory,
who
was
the
first
woman
to
get
the
first
American
woman
to
earn
a
Nobel
Prize
in
science,
and
she
discovered
the
Cori
cycle,
which
is
some
kind
of
metabolic
cycle
in
the
cell
and
I've
told
you
everything
I
know
about
that.
Ok,
so
that
we're
gonna
get
Perlmutter
at
the
very
end
of
2020
it'll
be
available
to
users
in
2021
and
then,
hopefully
in
2024,
we'll
get
a
new
machine
nurse
tan.
A
A
Asta
system
and
it's
gonna
be
three
to
four
times
more
powerful
than
Cory,
and
this
is
the
first
system
that
we've
ever
had
that
we've
really
designed
it
to
meet
the
needs
of
simulation
and
data
analysis
from
experimental
facilities,
so
both
of
these
things
combined
into
one
machine.
So
we
don't
know
exactly
what
it's
gonna
look
like.
You
know,
because
when
you
buy
a
Cray,
basically
buy
it
so
that
you
can
get
the
mural
on
the
front
I'm
just
joking
about
that.
A
But
you
do
get
a
cool
mural
on
the
front
when
you
pay
lots
of
money
for
a
supercomputer,
so
this
is
just
kind
of
a
example
of
what
it
could
look
like,
but
that's
probably
not
what
the
machine's
gonna
end
up.
Looking
like
just
so,
you
know.
Okay,
so
you
may
wonder
why
we're
calling
it
Perlmutter.
This
seems
like
really
long
name
so
Saul
Perlmutter
was
the
winner
of
the
2011
Nobel
Prize
in
Physics.
A
He
discovered
the
accelerating
expansion
of
the
universe
and
when
he
was
working
on
this
research,
he
was
really
a
pioneer
in
using
supercomputers
combining
large-scale
simulations
with
experimental
data
analysis.
So
Saul
Perlmutter
is
actually
still
alive
and
he
works
at
Berkeley
Lab,
and
so
we
had
to
ask
his
permission
if
we
could
name
the
machine
after
him-
and
he
said
yes,
but
on
the
condition
that
we,
when
people
SSH
into
it,
they
just
can
SSH
into
Saul
that
nurse
stuck
up
because
Perlmutter
was
too
long.
A
Okay,
so
let's
talk
about
the
systems
that
we
have
here
at
nurse,
okay,
so
until
last
month,
in
fact,
just
over
a
month
ago,
we
had
Edison,
but
Edison
has
now
been
decommissioned.
So
that's
why
he
has
that.
That
said,
no
no,
no
sign,
as
my
son
calls
it,
but
we
still
have
Cory.
So
Cory
is
a
cray
XC
40
machine.
A
It
has
about
2,400
nodes
with
Haswell
cores
and
most
so
those
are
just
kind
of
standard,
CPU
cores
that
everyone
is
used
to,
and
then
it
has
almost
10,000
nodes
with
Intel
KL,
Xeon,
Phi
cores,
and
so
what
these
are.
These
are
many
core
processors,
they're
low
energy,
but
actually
very
powerful
in
the
aggregate.
A
So
Cory
also
has
a
unique
feature
called
the
data
warp
you
might
have
heard
of
that
as
burst,
buffer
right
and
so
burst.
Buffer
is
an
all
flash,
very
fast,
very
powerful
file
system,
and
so
in
fact,
I
believe
you're
gonna
learn
more
about
that
later
on
today,
okay
and
then
we
have
a
lot
of
other
things.
We've
got
a
scratch
system
that
is
attached
to
Cori,
it's
a
scratch
file
system
and
we'll
talk
a
little
bit
more
about
what
that
means.
A
In
a
few
minutes,
we've
got
our
project
file
system,
we've
got
our
home
file
system
and
we've
got
HP
SS,
which
is
like
a
table,
long
term
storage
system
and
then
we're
all
connected
to
all
these
internet
to
the
internet
and
through
IES
net,
which
is
the
energy
scientist
network,
which
is
a
really
powerful
network
that
serves
science
in
the
United
States
okay.
So
let's
talk
about
Cori
a
little
bit
more
so
I
mentioned.
A
We've
got
these
Haswell
notes
right,
so
we've
got
about
2500
of
those
or
so,
and
then
we've
got
these
K
and
L
notes.
So
he's
has
no
notes.
So
when
we
were
designing
this
machine,
we
wanted
to
have
sort
of
something
that
people
could
use
really
well
for
like
processing
their
data.
Okay.
So
that's
why
we
have
these
half
long,
so
they're
really
good
for
throughput
of
of
data,
that's
kind
of
like
what
we're
trying
to
do
the
cues
on
this
allow.
Actually
you
can
just
run
a
single
core
job.
A
A
Excuse
me
very
long,
wait
time
if
you
want
to
use
those
notes
now,
on
the
other
hand,
we've
got
the
KL
node,
so
these
are
really
good,
like
I
said,
they're
really
good
for
performance,
they're
very
low
power,
but
they
have,
and
they
have
like
a
whole
bunch
of
cores
on
them,
but
that
if
you
can
exploit
that
kind
of
an
architecture,
you
can
have
really
good
performance
on
these
nodes.
We
really
like
large
jobs
on
the
KL
partition
of
the
machine.
A
If
you
run
a
job
that
uses
at
least
1,024
nodes,
you
even
get
a
discount
hasn't
that
great
ok,
it's
like
four
times
as
big
as
that
has
well
sideof
of
the
machine.
So
this
is
really
the
best
place
to
run
your
jobs.
If
you
can
at
all
use
it,
it
has
much
shorter
queues,
and
then
we
have
something
unique
called
the
flex
queue
which
I
believe
we'll
talk
about
later
in
a
later
presentation.
But
the
idea
with
the
flex
queue
is
that
you
can.
A
Okay,
so
let's
talk
more:
let's
talk
about
our
file
systems,
so
I
mentioned:
we've
got
home
project
scratch.
We've
got
the
first
buffer,
we've
got
HP
SS,
so
home
is
your
home
directory.
That's
where
your
home
directory
is
at
nurse,
it's
kind
of
small.
We
keep
it
small
on
purpose,
because
the
purpose
of
this
is
for
you
to
kind
of
store
some
data
like
your
source
code
or
your
shell
scripts
or
something,
but
we
don't.
A
A
A
A
A
Okay,
now
we
talk
about
some
local
file
systems,
so
we've
got
scratch
adverse
buffer.
So
scratch
is
a
large
temporary
storage
system
and
it's
really
optimized
for
readwrite
sort
of
operations,
but
not
really
for
storage.
It's
not
backed
up.
This
is
very
important
if
you
have
your
data
on
scratch
and
it
goes
way
too
bad.
So
sad,
it's
not
backed
up.
We
also
have
a
purge
policy.
So
if
you
have
a
file
that
is
not
accessed
in
the
past
12
weeks,
then
we
reserve
the
right
to
delete
that
file.
A
A
Okay,
so
on
Cori
we
have
the
burst
buffer
and
it
is
sort
of
a
temporary
per
job
storage
system.
So
it's
a
high
performance,
SSD
file
system.
So
it's
how
most
file
systems
are
spinning
disk,
but
this
one
is
solid-state
drives.
So
that's
why
it
performs
so
much
better.
It's
available
only
on
Cori,
and
this
is
really
great
file
system.
If
you
are
trying
to
run
your
stuff
and
you
find
that
on
scratch
that
you're
getting
bad
performance
because
your
code
is
constrained
by
the
I/o
operations.
A
So
if
you
have
an
I/o
intensive
application,
then
you
should
consider
using
the
burst
buffer.
Okay,
now
we're
going
to
talk
about
HP
SS,
which
is
a
long
term
storage
system.
So
HP
SS
stands
for
high
performance
storage
system
and
it's
an
archive
okay.
So
what
you
do
is
you
take
any
data?
That
is
not
that
you're
not
going
to
access
every
day,
but
you
still
want
to
save
it
and
you
put
it
into
HP
SS,
it's
a
hierarchical
storage
system,
and
actually
it's
really
fascinating
how
this
works.
A
So
in
at
the
very
front
you
know
kind
of
the
front
line
of
the
machine.
It
has
some
disk
arrays,
just
like
spinning
disks,
just
like
regular
what
you
would
expect
and
then
in
the
back
end
it
has
a
tape
subsystem,
and
that
is
where
your
data
kind
of
migrates
to
so
when
you
first
put
your
data
on
there.
It's
in
this
front
system,
this
these
spinning
disks
or,
if
you've
recently
accessed
that
data.
A
Okay,
so
let's
talk
about
using
nurse
file
system,
so
kind
of
like
file
system
at
a
catch
okay.
So
this
is
an
imperfect
analogy,
but
I
think
it
gets
my
point
apart
across.
So
let's
say
that
computing
is
like
baking.
Okay
and
your
input
is
like
your
baking
ingredients
and
your
output
is
a
cake
right
now
who
doesn't
love
cake,
I,
don't
know
so.
A
Basically,
in
this
scenario,
nurse
is
like
a
gigantic
shared
kitchen
space
with
all
the
latest
gadgets.
Okay,
so
you
can
think
of
the
computers
as
being
like
the
oven
right
like
when
you're
doing
your
computations.
That's
when
you're
baking,
your
cake
and
you're.
Getting
your
output
of
cake
home
and
project
are
kind
of
like
your
pantry
or
your
fridge.
Okay,
and
it's
where
you
kind
of
keep
your
ingredients.
A
Hp
SS
is
like
the
freezer.
Where
you
keep,
you
know
the
frozen
blueberries
or
things
that
you
don't
access
very
often,
but
sometimes
you
might
need
them,
and
then
scratch
is
like
your
kitchen,
countertop,
okay,
so
nobody's
nobody's
looking
at
me
with
great
disdain,
but
probably
people
think
I'm,
not
so
far.
Okay,
so
anybody
anybody
do
baking
or
anything
anybody
anybody
like
to
do
that.
Okay,
it's
in
my
name,
so
I
love
it.
So
when,
when
you
bake
first
thing
you
want
to
do,
is
you
were
to
stage
all
of
your
ingredients
right?
A
You
want
to
put
them
all
on
your
kitchen,
countertop
make
sure
they're
all
there
right.
Does
anybody
ever
not
done
that
and
then
you're
like
oh
crap,
I,
don't
have
any
baking
powder
or
whatever,
and
then
you
have
to
go
to
the
store.
Do
you
have
to
buy
it
right?
So
that's
the
best,
so
you
want
to
put
them
all
in
there
and
you
want
to
put
them
on
your
kitchen
counter.
So
likewise,
when
we're
doing
computation,
it's
the
same
idea
right,
we're
staging
our
input,
data
and
our
executable
onto
the
scratch
filesystem.
Okay,.
A
So
you
remember,
this
is
like
a
shared
kitchen
space.
So,
like
that'd,
be
super
rude.
If
you
just
left
all
your
dirty,
you
know
your
dirty
dishes
and
your
rotting
eggshells
and
stuff
all
over
the
counter
right
super
rude.
So
it's
okay!
If
you
just
kind
of
let
your
cake
cool
there
for
a
while,
but
you
got
to
get
it
all
up
and
you
got
to
clean
up
after
yourself,
because
if
you,
if
you
don't
clean
up
after
yourself,
then
we
will-
and
it
won't
be
in
a
way
that
you
like.
A
So
you
know
at
you
know,
maybe
if
you're
at
home-
and
you
left
all
of
your
dirty
dishes
out-
maybe
maybe
your
mom
would
come
by
and
because
she
needed
to
use
the
kitchen
too,
and
she
would
you
know,
get
rid
of
the
eggshells
and
clean
up
all
your
bowls
and
stuff
but
know
what
we're
gonna
do.
Is
we're
going
to
go
through
and
we're
gonna
throw
everything
out
if
we're
to
put
it
all
in
the
trash,
including
your
cake?
Okay,
so
that's
the
analogy!
A
A
A
So
yeah,
that's
a
good
question,
so
the
HP
SS
space
is
it
unlimited.
It
is
limited,
but
it's
not
very
limited.
So
so,
yes,
you
could
take
you
what,
ideally,
what
you
should
do
is
you
should
take
your
cake
or
your
results,
and
you
should
put
it
into
HP
SS
already,
there's
some
kind
of
long-term
storage
like
that,
but
try
not
to
clean
up
your
eggshells
and
put
them
in
there
too,
because
I
have
a
gross
okay,
good,
yeah.
A
No,
no,
there
are
not
nodes
with
local
SSD
scratch
right.
So
it's
a
global
scratch
filesystem
and
it's
it's
a
completely
like
separate
set
of
cabinets
where
the
filesystem
is
located
and
then
the
burst
buffer
is
located
on.
There
are
some
burst
buffer
nodes
that
are
within
the
machine.
So
that's
where
the
SSDs
are,
but
they're
not
local
to
any.
Given.
A
So
so
this
is
a
complicated
question
we
have
had
times
when
we've
had
network
issues
that
have
impacted
it.
However,
the
Cray
Ares
network,
which
is
what
is
behind
the
scenes
on
these
machines,
has
sort
of
like
this
adaptive,
routing
that
it
can
do
so
generally
like
if,
if
there's
a
link
down
or
something
it'll,
just
kind
of
route
around
that,
so
the
problems
of
course
would
come
if
we
had
so
many
links
down
that
it
couldn't
route
around
it
right
like
if
there
really
literally,
was
no
path
from
A
to
B.
A
That
could
be
a
problem,
but
we
have
we,
you
know
we
we
maintain
the
machine,
we
try
to
take
a
care
of
it,
and
so
we've
we've
never
really
had
anything
like
that
happen,
and
if
it
were
to
happen,
then
that
would
be
an
outage
of
the
machine.
It
just
would
not
be
working.
We
would
consider
it
an
outage.
A
Okay,
so
let
me
talk
about
software
now,
so
the
Cray
supercomputers,
they
run
an
OS.
That
is
a
version
of
Linux.
Okay,
it's
a
little
bit
optimized
for
the
for
those
platform,
but
it's
basically
it's
Linux.
We
provide
compilers
on
the
machines
and
we
have
a
lot
of
libraries
that
you
might
need
to
use
in
your
in
your
compiling
your
code.
A
So
we
have
some
that
are
provided
by
the
vendor
and
then
still
others
that
we
provide
from
nurse
and
then
also
we
compile
and
support
many
software
packages
for
our
users,
particularly
in
the
computational
chemistry
and
materials
science
area
and
we'll
you'll,
see
more
details
in
later
presentations.
So
I
just
wanted
to
show
you,
though
these
are
like
a
whole
bunch
of
chemistry
and
materials
applications
that
we
provide.
A
Ok.
So
let's
talk
about
interacting
with
nurse,
so
first
we're
going
to
talk
about
consulting
our
operations
group
and
then
we'll
talk
about
the
nurse
user
group.
Okay.
So
look
at
all
these
beautiful
faces
here.
This
is
the
consulting
team,
so
whenever
you
send
in
a
ticket
or
you,
you
make
a
phone
call,
you're
gonna
reach
one
of
these
fantastic
folks.
A
Okay,
now,
sometimes
when
you
send
in
a
ticket,
we
don't
know
how
to
handle
it,
so
we'll
pass
it
on
to
somebody
else,
but
those
pictures
are
of
the
people
who
take
the
tickets
and
figure
out
what
to
do
with
them.
Okay,
so
we
have
some
service
level
agreements,
so
we
will
reply
to
you
within
four
business
hours.
You
know
business
hour
is
8:00
to
5:00
Monday
through
Friday,
okay.
So
if
you,
if
you
send
in
a
ticket
at
5:01
p.m.
today,
you
won't
get
a
reply
until
Monday.
A
Okay,
so
just
know
that
we
will
help
you
resolve
your
problem
and
keep
you
apprised
of
the
progress
we
will
attempt
to
accommodate
any
needs
that
don't
fit
within
our
operating
structure.
Okay,
so
we
have.
You
know
we
have
rules,
we
have.
We
have
the
way
that
we
do
things
for
a
reason
that,
and
it
works
for
most
people,
but
if
it
doesn't
work
for
you,
then
you
know
you
can
send
in
a
ticket
and
we
can
talk
about
it
and
we
can
figure
out
how
we
can
best
accommodate
you.
A
Okay,
so
a
few
tips
and
tricks
here
help
us
to
help
you,
okay,
so
if
you,
if
you
send
in
a
ticket
that
says
my
code,
doesn't
work,
okay,
that's
not
very
helpful
to
us
right.
We
don't
know
what
code
we
don't
know
what
the
senior!
What
did
you
do?
Why
is
it
not
working
like?
Is
it
it
worked
yesterday
and
now
it
doesn't
work
today
or
is
it
that
it's
never
worked
or
and
what
is
it
that
isn't
working?
Are
you
able
to
compile
it?
Are
you
able
to
submit
a
job
to
run
it?
A
You
know,
what's
like,
what's
going
on
so
the
more
specifics
that
you
can
provide
the
better
cuz,
the
better?
We
can
help
you
because,
amazingly
none
of
us
are
mind.
Reader's
I've
been
trying
for
a
long
time,
but
it
doesn't
work
okay,
so
we
have
a
whole
staff
of
operations
staff
and
they
are
on-site,
24/7,
365
and
a
quarter
days
per
year
to
supervise
the
operation
of
the
machine
room.
So
if
you
have
a
problem
at
2:05
a.m.
A
on
Christmas
morning,
there's
somebody
here
who
who
can
help
now
what
they
do
is
they
they
primarily
supervise
the
operation
of
the
machines,
make
sure
that
you
know
everything's
running
smoothly.
So
they
can't
help
you
compile
your
code,
but
if
the
machine
is
down
or
if
there's
a
problem,
they
know
about
it,
they
can
tell
you
so
you
can.
You
can
give
them
a
call
because
they're
always
there,
you
can
give
them
a
call
find
out
what's
going
on
and
they
can
do
some
simple
tasks
that
you
know
that
are
within
their
domain
of
expertise.
A
So
they
can
help
you
reset
your
password.
They
can
kill
your
jobs
for
you.
If
you
need
your
job
to
be
killed,
they
can
make
a
few
changes
to
a
reservation
that
is
still
running,
but
there
are
a
really
great
resource
and
some
really
amazing
people
who
work
down
there
and
I
say
down
because
they're
directly
below
us.
A
Ok,
so
then
we've
got
the
nurse
user
group,
so
the
nurse
user
group.
We
call
it
nug,
it's
it's
our
nurse
user
community.
So
it's
a
really
great
source
of
advice
and
feedback
for
nurse.
So
we
like
we
like
feedback.
We
like
to
know
what
users
are
thinking
and
what
they
want,
so
that
we
can
do
better
for
them
and
do
what
they
need
it.
So,
there's
an
executive
committee
which
is
three
representatives
from
each
office
in
the
office
of
science.
A
So
you
remember:
I
had
that
list
of
offices
in
the
office
of
science,
so
there's
three
representatives
from
each
of
those
and
then
there's
three
members
at
large
and
also
we
generally
have
monthly
teleconferences,
hosted
by
nurse,
usually
on
the
third
Thursday
of
the
month
at
11
a.m.
so
you
can
feel
free
to
tune
in
to
those
and
then
also
we
have
our
nurse
user
group
annual
meeting,
which
is
going
to
happen
next
month
on
Friday
July
19th
in
Maryland,
somewhere
in
Maryland
in
the
DC
area.
A
So
you
can
find
all
of
that
on
our
website
as
well.
Okay,
so
you're,
a
nurse
user,
really
happy
that
you
are
I
just
want
to
give
you
some
ideas
of
our
responsibilities
and
expectations.
So
the
first
thing
is
be
kind
to
your
neighbor
users.
So
don't
abuse
the
shared
resources,
so
in
particular
the
login
notes
are
a
shared
resource.
So,
if
you're
doing
something,
that's
really
compute-intensive
on
the
login
node,
then
you're
ruining
the
experience
of
the
other
people
on
that
login
node,
because
then
they
can't
get
anything
done.
A
Okay,
another
one
is
to
use
your
allocation
smartly.
So
when
your
allocation
is
gone,
there's
not
a
lot
that
we
can
do
to
give
you
more
allocation.
So
if
you,
if
you
squandered
your
allocation
on
something,
you
know
it's
kind
of
too
bad,
we're
really
sorry,
but
that's
about
it
for
us.
So
another
thing
is
to
pick
the
right
resource
for
your
job
and
your
data.
A
So
if
you're
running
some
small
jobs,
that's
really
great
on
the
quarry
has
well
notes,
but
probably
not
so
great
on
the
K&L
notes
like
if
you
want
to
just
run
on
a
single
core,
probably
the
KL
nodes.
It's
not
the
place
for
you
to
do
that
back
up
your
stuff,
especially
from
scratch,
because
scratch
has
a
purge
policy
and
so
we're
gonna,
throw
it
all
away.
We're
gonna,
throw
away
your
cake.
Okay,
if
there's
anything,
if
there's
any
like
take-home
message
today,
that
you
understand
is
that
nurse
is
gonna,
throw
away
your
cake.