<s>
The	O
NASA	O
Advanced	O
Supercomputing	B-Architecture
(	O
NAS	O
)	O
Division	O
is	O
located	O
at	O
NASA	O
Ames	O
Research	O
Center	O
,	O
Moffett	O
Field	O
in	O
the	O
heart	O
of	O
Silicon	O
Valley	O
in	O
Mountain	O
View	O
,	O
California	O
.	O
</s>
<s>
It	O
has	O
been	O
the	O
major	O
supercomputing	B-Architecture
and	O
modeling	O
and	O
simulation	O
resource	O
for	O
NASA	O
missions	O
in	O
aerodynamics	O
,	O
space	O
exploration	O
,	O
studies	O
in	O
weather	O
patterns	O
and	O
ocean	O
currents	O
,	O
and	O
space	O
shuttle	O
and	O
aircraft	O
design	O
and	O
development	O
for	O
almost	O
forty	O
years	O
.	O
</s>
<s>
The	O
facility	O
currently	O
houses	O
the	O
petascale	B-General_Concept
Pleiades	B-Device
,	O
Aitken	O
,	O
and	O
Electra	O
supercomputers	B-Architecture
,	O
as	O
well	O
as	O
the	O
terascale	O
Endeavour	B-Device
supercomputer	B-Architecture
.	O
</s>
<s>
The	O
main	O
building	O
also	O
houses	O
disk	O
and	O
archival	O
tape	O
storage	O
systems	O
with	O
a	O
capacity	O
of	O
over	O
an	O
exabyte	O
of	O
data	O
,	O
the	O
hyperwall	O
visualization	O
system	O
,	O
and	O
one	O
of	O
the	O
largest	O
InfiniBand	B-Architecture
network	O
fabrics	O
in	O
the	O
world	O
.	O
</s>
<s>
In	O
the	O
mid-1970s	O
,	O
a	O
group	O
of	O
aerospace	O
engineers	O
at	O
Ames	O
Research	O
Center	O
began	O
to	O
look	O
into	O
transferring	O
aerospace	O
research	O
and	O
development	O
from	O
costly	O
and	O
time-consuming	O
wind	O
tunnel	O
testing	O
to	O
simulation-based	O
design	O
and	O
engineering	O
using	O
computational	O
fluid	O
dynamics	O
(	O
CFD	O
)	O
models	O
on	O
supercomputers	B-Architecture
more	O
powerful	O
than	O
those	O
commercially	O
available	O
at	O
the	O
time	O
.	O
</s>
<s>
Groundbreaking	O
on	O
a	O
state-of-the-art	O
supercomputing	B-Architecture
facility	O
took	O
place	O
on	O
March	O
14	O
,	O
1985	O
in	O
order	O
to	O
construct	O
a	O
building	O
where	O
CFD	O
experts	O
,	O
computer	O
scientists	O
,	O
visualization	O
specialists	O
,	O
and	O
network	O
and	O
storage	O
engineers	O
could	O
be	O
under	O
one	O
roof	O
in	O
a	O
collaborative	O
environment	O
.	O
</s>
<s>
In	O
1986	O
,	O
NAS	O
transitioned	O
into	O
a	O
full-fledged	O
NASA	O
division	O
and	O
in	O
1987	O
,	O
NAS	O
staff	O
and	O
equipment	O
,	O
including	O
a	O
second	O
supercomputer	B-Architecture
,	O
a	O
Cray-2	B-Device
named	O
Navier	O
,	O
were	O
relocated	O
to	O
the	O
new	O
facility	O
,	O
which	O
was	O
dedicated	O
on	O
March	O
9	O
,	O
1987	O
.	O
</s>
<s>
NAS	O
has	O
been	O
one	O
of	O
the	O
leading	O
innovators	O
in	O
the	O
supercomputing	B-Architecture
world	O
,	O
developing	O
many	O
tools	O
and	O
processes	O
that	O
became	O
widely	O
used	O
in	O
commercial	O
supercomputing	B-Architecture
.	O
</s>
<s>
NAS	O
develops	O
and	O
adapts	O
software	O
in	O
order	O
to	O
"	O
complement	O
and	O
enhance	O
the	O
work	O
performed	O
on	O
its	O
supercomputers	B-Architecture
,	O
including	O
software	O
for	O
systems	O
support	O
,	O
monitoring	O
systems	O
,	O
security	O
,	O
and	O
scientific	O
visualization	O
,	O
"	O
and	O
often	O
provides	O
this	O
software	O
to	O
its	O
users	O
through	O
the	O
NASA	O
Open	O
Source	O
Agreement	O
(	O
NOSA	O
)	O
.	O
</s>
<s>
NAS	B-Device
Parallel	I-Device
Benchmarks	I-Device
(	O
NPB	O
)	O
were	O
developed	O
to	O
evaluate	O
highly	O
parallel	O
supercomputers	B-Architecture
and	O
mimic	O
the	O
characteristics	O
of	O
large-scale	O
CFD	O
applications	O
.	O
</s>
<s>
Portable	B-Application
Batch	I-Application
System	I-Application
(	O
PBS	O
)	O
was	O
the	O
first	O
batch	O
queuing	O
software	O
for	O
parallel	O
and	O
distributed	O
systems	O
.	O
</s>
<s>
PLOT3D	O
was	O
created	O
in	O
1982	O
and	O
is	O
a	O
computer	O
graphics	O
program	O
still	O
used	O
today	O
to	O
visualize	O
the	O
grids	O
and	O
solutions	O
of	O
structured	O
CFD	O
datasets	B-General_Concept
.	O
</s>
<s>
Since	O
its	O
construction	O
in	O
1987	O
,	O
the	O
NASA	B-Device
Advanced	I-Device
Supercomputing	I-Device
Facility	I-Device
has	O
housed	O
and	O
operated	O
some	O
of	O
the	O
most	O
powerful	O
supercomputers	B-Architecture
in	O
the	O
world	O
.	O
</s>
<s>
In	O
1987	O
,	O
NAS	O
partnered	O
with	O
the	O
Defense	O
Advanced	O
Research	O
Projects	O
Agency	O
(	O
DARPA	O
)	O
and	O
the	O
University	O
of	O
California	O
,	O
Berkeley	O
in	O
the	O
Redundant	B-Architecture
Array	I-Architecture
of	I-Architecture
Inexpensive	I-Architecture
Disks	I-Architecture
(	O
RAID	B-Architecture
)	O
project	O
,	O
which	O
sought	O
to	O
create	O
a	O
storage	B-Application
technology	I-Application
that	O
combined	O
multiple	O
disk	O
drive	O
components	O
into	O
one	O
logical	O
unit	O
.	O
</s>
<s>
Completed	O
in	O
1992	O
,	O
the	O
RAID	B-Architecture
project	O
lead	O
to	O
the	O
distributed	O
data	O
storage	B-Application
technology	I-Application
used	O
today	O
.	O
</s>
<s>
The	O
NAS	O
facility	O
currently	O
houses	O
disk	O
mass	O
storage	O
on	O
an	O
SGI	O
parallel	O
DMF	O
cluster	O
with	O
high-availability	O
software	O
consisting	O
of	O
four	O
32-processor	O
front-end	O
systems	O
,	O
which	O
are	O
connected	O
to	O
the	O
supercomputers	B-Architecture
and	O
the	O
archival	O
tape	O
storage	O
system	O
.	O
</s>
<s>
Data	O
stored	O
on	O
disk	O
is	O
regularly	O
migrated	O
to	O
the	O
tape	O
archival	O
storage	O
systems	O
at	O
the	O
facility	O
to	O
free	O
up	O
space	O
for	O
other	O
user	O
projects	O
being	O
run	O
on	O
the	O
supercomputers	B-Architecture
.	O
</s>
<s>
In	O
1987	O
,	O
NAS	O
developed	O
the	O
first	O
UNIX-based	O
hierarchical	O
mass	O
storage	O
system	O
,	O
named	O
NAStore	O
.	O
</s>
<s>
It	O
contained	O
two	O
StorageTek	B-Application
4400	O
cartridge	O
tape	O
robots	O
,	O
each	O
with	O
a	O
storage	O
capacity	O
of	O
approximately	O
1.1	O
terabytes	O
,	O
cutting	O
tape	O
retrieval	O
time	O
from	O
4	O
minutes	O
to	O
15	O
seconds	O
.	O
</s>
<s>
With	O
the	O
installation	O
of	O
the	O
Pleiades	B-Device
supercomputer	I-Device
in	O
2008	O
,	O
the	O
StorageTek	B-Application
systems	O
that	O
NAS	O
had	O
been	O
using	O
for	O
20	O
years	O
were	O
unable	O
to	O
meet	O
the	O
needs	O
of	O
the	O
greater	O
number	O
of	O
users	O
and	O
increasing	O
file	O
sizes	O
of	O
each	O
project	O
's	O
datasets	B-General_Concept
.	O
</s>
<s>
In	O
2009	O
,	O
NAS	O
brought	O
in	O
Spectra	O
Logic	O
T950	O
robotic	O
tape	O
systems	O
which	O
increased	O
the	O
maximum	O
capacity	O
at	O
the	O
facility	O
to	O
16	O
petabytes	O
of	O
space	O
available	O
for	O
users	O
to	O
archive	O
their	O
data	O
from	O
the	O
supercomputers	B-Architecture
.	O
</s>
<s>
In	O
1984	O
,	O
NAS	O
purchased	O
25	O
SGI	O
IRIS	O
1000	O
graphics	O
terminals	O
,	O
the	O
beginning	O
of	O
their	O
long	O
partnership	O
with	O
the	O
Silicon	O
Valley-based	O
company	O
,	O
which	O
made	O
a	O
significant	O
impact	O
on	O
post-processing	O
and	O
visualization	O
of	O
CFD	O
results	O
run	O
on	O
the	O
supercomputers	B-Architecture
at	O
the	O
facility	O
.	O
</s>
<s>
Visualization	O
became	O
a	O
key	O
process	O
in	O
the	O
analysis	O
of	O
simulation	O
data	O
run	O
on	O
the	O
supercomputers	B-Architecture
,	O
allowing	O
engineers	O
and	O
scientists	O
to	O
view	O
their	O
results	O
spatially	O
and	O
in	O
ways	O
that	O
allowed	O
for	O
a	O
greater	O
understanding	O
of	O
the	O
CFD	O
forces	O
at	O
work	O
in	O
their	O
designs	O
.	O
</s>
<s>
In	O
2002	O
,	O
NAS	O
visualization	O
experts	O
developed	O
a	O
visualization	O
system	O
called	O
the	O
"	O
hyperwall	O
"	O
which	O
included	O
49	O
linked	O
LCD	B-Device
panels	I-Device
that	O
allowed	O
scientists	O
to	O
view	O
complex	O
datasets	B-General_Concept
on	O
a	O
large	O
,	O
dynamic	O
seven-by-seven	O
screen	O
array	O
.	O
</s>
<s>
Each	O
screen	O
had	O
its	O
own	O
processing	O
power	O
,	O
allowing	O
each	O
one	O
to	O
display	O
,	O
process	O
,	O
and	O
share	O
datasets	B-General_Concept
so	O
that	O
a	O
single	O
image	O
could	O
be	O
displayed	O
across	O
all	O
screens	O
or	O
configured	O
so	O
that	O
data	O
could	O
be	O
displayed	O
in	O
"	O
cells	O
"	O
like	O
a	O
giant	O
visual	O
spreadsheet	O
.	O
</s>
<s>
The	O
second	O
generation	O
"	O
hyperwall-2	O
"	O
was	O
developed	O
in	O
2008	O
by	O
NAS	O
in	O
partnership	O
with	O
Colfax	O
International	O
and	O
is	O
made	O
up	O
of	O
128	O
LCD	B-Device
screens	I-Device
arranged	O
in	O
an	O
8x16	O
grid	O
23	O
feet	O
wide	O
by	O
10	O
feet	O
tall	O
.	O
</s>
<s>
It	O
is	O
capable	O
of	O
rendering	O
one	O
quarter	O
billion	O
pixels	B-Algorithm
,	O
making	O
it	O
the	O
highest	O
resolution	O
scientific	O
visualization	O
system	O
in	O
the	O
world	O
.	O
</s>
<s>
It	O
contains	O
128	O
nodes	O
,	O
each	O
with	O
two	O
quad-core	O
AMD	B-General_Concept
Opteron	I-General_Concept
(	O
Barcelona	O
)	O
processors	O
and	O
a	O
Nvidia	O
GeForce	O
480	O
GTX	O
graphics	B-Architecture
processing	I-Architecture
unit	I-Architecture
(	O
GPU	B-Architecture
)	O
for	O
a	O
dedicated	O
peak	O
processing	O
power	O
of	O
128	O
teraflops	O
across	O
the	O
entire	O
system	O
—	O
100	O
times	O
more	O
powerful	O
than	O
the	O
original	O
hyperwall	O
.	O
</s>
<s>
The	O
hyperwall-2	O
is	O
directly	O
connected	O
to	O
the	O
Pleiades	B-Device
supercomputer	I-Device
's	O
filesystem	O
over	O
an	O
InfiniBand	B-Architecture
network	O
,	O
which	O
allows	O
the	O
system	O
to	O
read	O
data	O
directly	O
from	O
the	O
filesystem	O
without	O
needing	O
to	O
copy	O
files	O
onto	O
the	O
hyperwall-2	O
'	O
s	O
memory	O
.	O
</s>
<s>
In	O
2014	O
,	O
the	O
hyperwall	O
was	O
upgraded	O
with	O
new	O
hardware	O
:	O
256	O
Intel	O
Xeon	O
"	B-Device
Ivy	I-Device
Bridge	I-Device
"	I-Device
processors	O
and	O
128	O
NVIDIA	O
Geforce	O
780	O
Ti	O
GPUs	B-Architecture
.	O
</s>
<s>
In	O
2020	O
,	O
the	O
hyperwall	O
was	O
further	O
upgraded	O
with	O
new	O
hardware	O
:	O
256	O
Intel	O
Xeon	O
Platinum	O
8268	O
(	O
Cascade	B-Device
Lake	I-Device
)	O
processors	O
and	O
128	O
NVIDIA	O
Quadro	O
RTX	O
6000	O
GPUs	B-Architecture
with	O
a	O
total	O
of	O
3.1	O
terabytes	O
of	O
graphics	O
memory	O
.	O
</s>
<s>
An	O
important	O
feature	O
of	O
the	O
hyperwall	O
technology	O
developed	O
at	O
NAS	O
is	O
that	O
it	O
allows	O
for	O
"	O
concurrent	O
visualization	O
"	O
of	O
data	O
,	O
which	O
enables	O
scientists	O
and	O
engineers	O
to	O
analyze	O
and	O
interpret	O
data	O
while	O
the	O
calculations	O
are	O
running	O
on	O
the	O
supercomputers	B-Architecture
.	O
</s>
<s>
The	O
NAS	O
visualization	O
team	O
developed	O
a	O
configurable	O
concurrent	O
pipeline	B-General_Concept
for	O
use	O
with	O
a	O
massively	O
parallel	O
forecast	O
model	O
run	O
on	O
the	O
Columbia	B-General_Concept
supercomputer	I-General_Concept
in	O
2005	O
to	O
help	O
predict	O
the	O
Atlantic	O
hurricane	O
season	O
for	O
the	O
National	O
Hurricane	O
Center	O
.	O
</s>
