►
From YouTube: 01 - Welcome and Introduction to NERSC
Description
Part of the NERSC New User Training on September 28, 2022.
Please see https://www.nersc.gov/users/training/events/new-user-training-sept2022/ for the training day agenda and presentation slides.
A
A
A
Okay,
so
let's
get
started
with
our
introduction.
So
nurse
is
an
acronym
for
the
national
Energy
Research
scientific
Computing
Center.
So
we
were
established
in
1974.
We
were
the
first
unclassified
supercomputer
Center.
The
original
Mission
of
nurse
was
to
enable
computational
science
as
a
complement
to
magnetically
controlled
plasma
experiments.
Our
mission
has
changed
a
bit
today.
Our
mission
is
to
accelerate
scientific
discovery
at
the
doe
office
of
science
through
high
performance,
Computing
and
extreme
data
analysis.
A
A
The
allocations
at
nurse
are
primarily
controlled
by
the
department
of
energy,
so
80
of
all
allocations
go
through
the
doe
production
awards
called
ercap
and
ercap
Awards
people
anywhere
from
you
know,
100
hours
to
ten
thousand
hours
of
node
hours
on
Nurse.
This
is
a
proposal
based
award
system
and
it's
and
the
awards
are
chosen
by
our
doe
program
managers.
A
A
nurse
doesn't
have
any
control
over
that
that
80
of
the
awards,
another
10,
goes
to
the
doe
Oscar
leadership,
Computing
Challenge,
and
so
these
are
projects
that
are
more
high
risk,
but
High
potentially
High
reward,
and
then
the
remaining
10
belongs
to
the
nurse
reserve,
and
so
we
use
that
for
our
projects.
So
some
of
it
goes
to
overhead
for
nurse
staff,
but
a
lot
of
it
also
goes
to
a
science
or
things
that
we
are
particularly
interested
in.
So
over
the
past
couple
of
years.
A
Much
of
that
Reserve
has
gone
actually
to
covid
research,
for
example,
so
our
users
produce
really
amazing
groundbreaking
science.
So,
as
far
as
we
know,
users
of
nurse
produce
more
Publications
than
users
of
any
other
Center
in
the
world.
We
get
about
2500
Publications
per
year
that
our
naming
nurse
as
as
a
resource
that
they
used
so
doe.
A
All
right
thanks,
sorry
about
that,
so
the
doe
office
of
science
gives
out
the
awards
and
and
then
there's
kind
of
these
programs
within
the
office
of
Science,
and
each
of
them
gets
a
share
of
the
pie.
As
you
can
see
so
next
I'm
going
to
talk
about
Nurse
Hardware.
So
you
know
we're
always
trying
to
get
better
systems
every
couple
of
years
for
our
users
so
that
you
can
do
more
science
and
use
the
latest
and
greatest
Technologies.
A
We
will
get
our
next
system.
Our
nurse
10
system,
which
is
expected
to
be
an
exaflop
system,
okay,
so
this
is
a
kind
of
a
diagram
I
guess
of
all
of
the
things
that
we
have
going
on
here
at
nurse,
so
we've
got
Pearl,
mutter
and
Corey
and,
as
I
said,
Corey's
going
to
be
retiring,
so
we're
going
to
basically
ignore
Corey
today,
so
we've
got
these
two
systems
and
they
are
connected
to
our
nurse
Network
and
they're,
also
connected
to
some
important
file
systems
that
we
have
and
so
I'm
going
to.
A
Tell
you
briefly
a
bit
more
about
those,
but
later
presentations
will
provide
you
with
even
more
information
Okay.
So
Corey
has
two
partitions
to
it.
So
it
has
some
notes
that
are
of
the
Haswell
CPU
architecture
and
some
notes
that
are
the
Cano
architecture,
and
so
we
can
kind
of
differentiate
between
the
types
of
jobs
that
people
might
want
to
run
on
Corey.
A
So
that
has
well
know
if
we
could
serve
those
to
be
really
good
for
throughput
and
you
can
even
do
single
core
types
of
jobs
and
it
has
kind
of
longer
well-time
limits
for
these
smaller
jobs.
But
it
has
pretty
long
cues
the
k,
l
nodes.
They
are
really
good
for
a
performance
if
you
have
the
type
of
code
that
can
take
advantage
of
many
cores
and
we
encourage
people
to
run
large
jobs
on
it
and
and
it's
a
partition
of
the
machine
is
this.
A
Partition
is
four
times
larger
than
the
other,
so
it
generally
has
much
shorter
cues,
and
then
we
also
have
a
cue
called
the
flex
q
that
can
increase
your
throughput
and
give
you
a
discount
on
the
charges
that
we
charge
you
for
using
the
machine
now
Pearl
better,
which
is
our
new
machine.
This
is
the
focus
of
today's
training.
A
Pro
matter
has
two
partitions:
it
has
some
GPU
nodes
and
it
has
some
CPU
based
nodes,
and
so
the
GPU
knows
they
just
have
an
immense
compute
power
because
of
those
gpus,
and
we
really
love
it.
When
people
run
large
jobs
using
many
gpus,
we
really
encourage
that,
and
it's
really
great
for
any
coach
that
can
exploit
GPU
compute
power.
A
On
the
other
hand,
we
have
the
CPU
partition
and
it
has
some
very
powerful
CPUs
in
it.
But
of
course,
CPUs
are
a
completely
eclipsed
by
by
gpus
right,
so
it
the
total
is
only
about
10
of
the
GPU
compute
power.
It
is
equivalent
in
compute
power
to
all
of
Corey,
which
we
had
just
talked
about,
and
it
functions
more
like
a
traditional
cluster
that
people
might
be
used
to,
and
so
we
would
encourage
this.
A
A
Now
I
am
just
breezing
through
all
of
this,
because
there
will
be
a
lot
more
depth
of
coverage
later
on
about
how
to
use
these
resources,
so
your
home
file
system.
So
this
is,
this
is
mounted
on
all
of
our
systems,
and
so,
when
you
log
in
to
pearlmutter
or
or
Quarry
but
promoter,
because
we're
focusing
on
parameter
today,
the
directory
that
you
get
dropped
into
by
default
is
your
home
directory,
and
so
it's
a
permanent,
relatively
small
amount
of
storage.
A
We
give
you
40,
gigs
of
storage
there
and
it's
like
I,
said
it's
on
all
platforms.
It
is
not
tuned
for
performance.
It's
just
there
to
store
things.
We
like
I,
said
we
give
you
40
gigs.
We
cannot
change
that
quotas,
it's
just
our
policy
because
there
are
other
places
where
you
can
store
more
data
and
it.
It
also
has
these
snapshot
backups.
So
it
has
a
seven
day
history.
A
So
your
home
directory
is
the
perfect
place
for
storing
small
data
such
as
Source
codes
or
shell
scripts.
That's
what
I
do
with
my
home
directory
now.
We
also
have
the
community
file
system.
Okay
and
every
project
has
a
directory
in
the
community
file
system,
and
so
this
is
more
permanent,
larger
storage.
It
is
also
mounted
on
all
platforms,
so
you
can
access
it
on
Corey
or
Pearl
Mudder.
A
It
has
so
so
performance
for
parallel
jobs,
like
you
can
run
your
parallel
jobs
on
there.
It's
not
going
to
be
as
good
as
your
scratch
and
then
you
can
change
your
quota.
We
can
change
your
quota
on
the
community
file
system.
If
you
need
more
space
on
there,
we
can
generally
accommodate
then
another
nice
thing
about
the
community
file
system.
It
also
has
these
snapshot
backups,
so
with
a
seven
day
history.
A
A
Okay,
so
next
we'll
talk
about
scratch
file
system,
so
a
scratch
file
systems
are
local
to
the
machine.
They
are
a
large
temporary
storage
place,
they're
really
optimized
for
read,
write
sort
of
operations,
not
really
for
storage,
because
that's
not
what
they're
for
they're
not
backed
up.
So
if
you
delete
a
file
there
or
it
gets
purged
which
I'll
talk
about
shortly,
it's
too
bad
you've
lost
it.
A
Finally,
we
have
our
hpss
long-term
storage
system.
So
hpss
stands
for
high
performance
storage
system.
It's
an
archive
really,
you
can
think
of
it.
That
way.
It's
an
archival
storage
for
infrequently
accessed
data
and
it's
actually
functions
in
a
hierarchical
fashion.
So
if
you
put
data
on
it,
it
first
ingests
that
onto
an
array
of
disks,
which
is
something
that
we're
all
we
used
to
in
file
systems.
A
Okay,
so
I'm
going
to
talk
about
using
nurse
file
systems,
because
I
think
this
is
a
really
important
message
for
everyone
to
understand
so
I'm
going
to
use
an
analogy,
which
of
course
is
imperfect,
but
I
when
I
was
growing
up
in
the
Southern
United
States
I
learned
that
the
way
to
a
man's
heart
is
is
through
his
stomach.
I
think
that
the
way
to
a
user's
brain
is
also
through
their
stomach.
A
So
we're
going
to
talk
about
baking
and
I
apologize
to
people
for
whom
it's
almost
lunch
time,
but
let's
say
that
Computing
is
like
baking
and
let's
say
that
the
input
to
whatever
you're
Computing
is
kind
of
like
the
ingredients
that
you
would
use
for
your
baking
and
your
output
is
a
cake.
I
will
say
that
for
the
sake
of
argument,
so
nurse
is
kind
of
like
this
gigantic
shared
kitchen
space,
where
it
has
all
the
latest
kitchen
gadgets.
So
you
can
think
of
the
super
computers
as
muffins.
A
A
Similarly,
when
you're
working
on
a
supercomputer
you're
going
to
Stage
all
of
your
data
and
your
executable
onto
the
scratch
file
system,
then
after
baking
you
you
need
to
clean
up
after
yourself
it's
okay,
to
let
your
cake
cool
on
the
kitchen
counter
of
our
shared
kitchen
space.
But
you
need
to
leave
space
the
space
clean
for
the
next
person
who
comes
along
to
use
the
kitchen
resources.
A
So
we
will.
We
will
let
you
cool
your
cake
for
a
while
we're
hoping
that
you're
going
to
throw
away
your
your
used,
eggshells
and
you're,
going
to
take
your
bowls
and
you're
gonna,
wash
them
in
the
sink
or
whatever,
to
continue
this,
this
bad
analogy.
But
if
you
don't
do
that,
then
after
a
while,
we
will
clean
up
after
you,
but
it's
not
going
to
be
in
the
way
that
you
would
want
so
we're
gonna
throw
all
your
stuff
in
the
trash,
including
your
cake.
If
you
leave
it
there
too
long.
A
So
that's
why
we
really
encourage
you
to
clean
up
anything.
That's
on
scratch!
Don't
leave
it
on
there
for
longer
than
definitely
not
longer
than
12
weeks,
but
don't
leave
it
on
there
for
very
long
so
that
you
make
sure
that
you
get
it
off
of
there
and
into
a
more
persistent
form
of
storage
where
it
will
be
safe
from
us
cleaning
up
after
you,
okay,
I'm,
going
to
talk
about
software
now
so
on.
A
On
a
pearl
meter,
we
have
the
crate
they're
made
by
hpe,
slash
cray,
and
the
cray
supercomputer
operating
system
is
a
version
of
Linux,
it's
kind
of
an
optimized
version
of
Linux.
We
provide
compilers
on
machines.
We
also
provide
many
libraries.
Some
of
them
are
provided
by
our
vendor.
Some
of
them
are
provided
by
nurse
ourselves.
We
have
some
applications
that
we
compile
and
support,
and
then
you
know,
as
I
said
before,
for
more
details
check
out
our
later
presentations
and
you'll
learn
a
lot
more
about
this.
A
So
we
do
provide
a
lot
of
chemistry
and
Materials
Science
applications
for
our
users,
because
they
are
popular
and
they
are
so
well
used.
We
also
have
a
very
rich
data
ecosystem.
You
can
see.
We
have
all
of
these
different
types
of
applications-
libraries,
databases-
you
know,
data
management,
all
kinds
of
things-
for
people
to
use
workflows.
So
that's
something
to
keep
in
mind.
A
We
keep.
We
have
some
policies
at
nurse,
so
one
of
them
is
that
we
we
keep.
Our
software
version
defaults
consistent
for
the
allocation
here.
That's
our
goal
anyway.
Some
software
we
may
have
to
make
exceptions
because
there
may
be
security
issues
or
we
may
have
a
major
operating
system
upgrade
that
we
are
required
to
do,
and
then
the
the
software
defaults
that
we
had
are
no
longer
compatible
with
that
operating
system.
So
sometimes
we
do
make
an
exception
for
that.
A
But
that's
generally,
our
goal
is
to
keep
the
software
version
defaults
for
the
whole
allocation.
Here
now
we
classify
our
software
at
nurse
into
four
support
levels,
so
we've
got
priority,
provided
minimal
and
restricted,
and
so
the
only
things
you
really
need
to
know
is
that
restrictive
software
we
don't
allow
it
on
our
resources.
So
if
you
have
export
controlled
software,
please
don't
bring
it
to
nurse
we
have
generally.
A
Otherwise,
we
there's
some
that
we
provide
with
a
high
priority
and
we
make
sure
that
it
works
correctly
and
then
there's
others
that
we
have
sort
of
less
of
a
commitment
to
okay.
Let's
talk
about
interacting
with
nurse,
so
we're
going
to
talk
about
Consulting
and
account
support
and
operations
and
our
nurse
User
Group.
So
we
have
a
big
team
of
folks
who
are
lined
up
to
help
you
out
with
any
kind
of
issues
that
you
might
run
into
at
nurse,
and
so
this
is
pictures
of
most
of
them,
but
not
all
of
them.
A
A
So
we
promise
that
our
first
response
will
be
within
four
business
hours
and
our
business
hours
are
Monday
through
Friday,
eight
to
five
Pacific
time
except
holidays.
We
will
help
you
to
resolve
your
problems
and
keep
you
apprised
of
the
progress
of.
What's
going
on
we'll
attempt
to
accommodate
your
needs
that
might
not
fit
within
our
operating
structure,
but
we'll
try
to
see
what
we
can
do
maybe
help
you
to
fit
within
that
if
possible,
and
of
course
we
always
welcome
feedback
and
constructive
criticism
from
our
users.
A
Now,
when
you
are
submitting
a
ticket
to
us,
please
help
us
to
help
you
and
provide
some
specifics
about.
What's
going
on
so
sometimes
we
get
tickets
where
somebody
just
writes
in,
and
they
say
my
job
failed
and
that's
not
very
helpful
to
us.
We
can't
help
you
if
you're
just
telling
us
that
your
job
failed.
You
need
to
provide
us
with
more
information
and
you'll
hear
more
about
and
another
presentation
later.
A
Another
thing
that
we
offer
is
user
appointments,
so
we
started
off
offering
office
hours,
which
was
just
basically
an
open
Zoom
meeting
where
you
could
join
to
get
help
with
a
particular
topic,
but
the
problem
with
office
hours,
especially
for
all
of
you,
professors
out
there.
You
may
understand
this.
Is
you
have
long
periods
where
nobody
comes
and
then
suddenly
you
have
like
people
who
just
showing
on
at
the
same
time,
and
so
it's
kind
of
inefficient
use
of
everyone's
time.
A
Okay,
next
I'm
going
to
talk
about
operation,
so
we
have
operations,
staff
on
site,
24,
7
every
day,
Christmas
4th
of
July
Sundays
at
2
A.M.
There
is
always
somebody
there,
our
operations
staff,
they
know
the
health
of
the
machines
and
they
can
help
you
with
some
tasks.
So
if
there's
a
job
that
you're
trying
to
kill,
it's
really
not
die,
they
could
probably
help
you
with
that.
A
If
it's
an
emergency
and
then
if
there
are
any
changes
that
you
want
to
make
to
running
to
a
running
reservation,
they
can
help
you
with
that
too.
But
otherwise,
please
avoid
contacting
our
operations
staff
except
for
under
urgent
circumstances,
because
they
are
very
busy
making
sure
that
that
everything
is
operating
well
and
so
they're
out
there
doing
a
lot
of
different
tasks.
A
Okay,
next
I'm
going
to
talk
about
the
nurse
user
group,
so
we
have
the
nurse
user
group
called
The
nug,
it's
our
community
and
nurse
users.
It's
a
great
source
of
advice
and
feedback
for
us.
There's
an
executive
committee
with
three
representatives
from
each
office
program
office
in
in
the
doe
office
of
science
plus
three
members
at
large.
We
have
monthly
teleconferences,
usually
the
third
Thursday
of
the
month
at
11
A.M
Pacific
time.
A
The
nug
also
has
a
slack
channel
that
you
can
join,
and
so,
if
you
click
on
that
link,
you'll
need
to
log
in
in
order
to
get
to
the
invitation
that
you
can.
Click
on
and
join
the
nug
slack
I
would
encourage
you
to
join
it.
It's
a
pretty
pretty
good
place
to
talk
to
your
peers
and
ask
them
questions
and
get
help
and
then
also
I'd
like
to
invite
you
to
join
us
for
the
nug
annual
meeting.
A
Another
thing
I
want
to
talk
about
is
Community
engagement,
so
we
do
have
some
special
interest
groups
through
nug
on
topics
including
Wharf
users
and
scientific
facility
users,
and
then
we
also
have
plans
and
works
for
building
a
cross-disciplinary
nurse
users
community
of
practice,
starting
in
2023.
Our
initial
Target
is
going
to
be
graduate
students
because
they
are
a
plurality
of
nurse
users,
but
for
more
information.
I
would
encourage
you
to
attend
the
nug
annual
meeting
on
the
Friday
session.
A
We
will
have
a
30-minute
talk
on
our
community
of
practice,
so
that
should
be
very
exciting
foreign.
Last
but
not
least,
I
have
one
minute,
but
I
think
I'm
gonna
make
it
use
the
responsibilities
and
expectations.
So
what
we
want
from
you
all
is
to
be
kind
to
your
neighbor
users.
Okay,
don't
abuse
the
shared
resources.
Sometimes
people
will
use
a
a
login
node
too
much
a
bit
too
excessively
and
that
kind
of
messes
things
up
for
your
neighbors
use
your
allocation,
smartly,
so
pick
the
right
resource
for
your
job
and
your
data.
A
So
you
know
so
that
you
use
your
allocation
well
back
up
your
stuff,
especially
from
scratch.
Where
there's
a
purge
policy
acknowledge
us
in
your
papers
when,
if
nobody
ever
acknowledged
us,
then
nobody
would
know
that
we
were
important
to
your
research
right.
So
so
please
acknowledge
us
so
that
we
can
stay
in
business
and
finally
pay
attention
to
security.