<s>
Pleiades	B-Device
(	O
)	O
is	O
a	O
petascale	B-General_Concept
supercomputer	B-Architecture
housed	O
at	O
the	O
NASA	O
Advanced	O
Supercomputing	B-Architecture
(	O
NAS	B-Device
)	O
facility	O
at	O
NASA	O
's	O
Ames	O
Research	O
Center	O
located	O
at	O
Moffett	O
Field	O
near	O
Mountain	O
View	O
,	O
California	O
.	O
</s>
<s>
As	O
of	O
November	O
2019	O
it	O
is	O
ranked	O
the	O
32nd	O
most	O
powerful	O
computer	O
on	O
the	O
TOP500	B-Operating_System
list	O
with	O
a	O
LINPACK	B-Device
rating	O
of	O
5.95	O
petaflops	O
(	O
5.95	O
quadrillion	O
floating	B-Algorithm
point	I-Algorithm
operations	O
per	O
second	O
)	O
and	O
a	O
peak	O
performance	O
of	O
7.09	O
petaflops	O
from	O
its	O
most	O
recent	O
hardware	O
upgrade	O
.	O
</s>
<s>
The	O
system	O
serves	O
as	O
NASA	O
's	O
largest	O
supercomputing	B-Architecture
resource	O
,	O
supporting	O
missions	O
in	O
aeronautics	O
,	O
human	O
spaceflight	O
,	O
astrophysics	O
,	O
and	O
Earth	O
science	O
.	O
</s>
<s>
Built	O
in	O
2008	O
and	O
named	O
for	O
the	O
Pleiades	B-Device
open	O
star	O
cluster	O
,	O
the	O
supercomputer	B-Architecture
debuted	O
as	O
the	O
third	O
most	O
powerful	O
supercomputer	B-Architecture
in	O
the	O
world	O
at	O
487	O
teraflops	O
.	O
</s>
<s>
It	O
originally	O
contained	O
100	O
SGI	B-Architecture
Altix	I-Architecture
ICE	O
8200EX	O
racks	O
with	O
12,800	O
Intel	B-Device
Xeon	I-Device
quad-core	O
E5472	O
Harpertown	O
processors	O
connected	O
with	O
more	O
than	O
20	O
miles	O
of	O
InfiniBand	B-Architecture
double	O
data	O
rate	O
(	O
DDR	O
)	O
cabling	O
.	O
</s>
<s>
With	O
the	O
addition	O
of	O
ten	O
more	O
racks	O
of	O
quad-core	O
X5570	O
Nehalem	O
processors	O
in	O
2009	O
,	O
Pleiades	B-Device
ranked	O
sixth	O
on	O
the	O
November	O
2009	O
TOP500	B-Operating_System
with	O
14,080	O
processors	O
running	O
at	O
544	O
teraflops	O
.	O
</s>
<s>
In	O
January	O
2010	O
,	O
the	O
scientists	O
and	O
engineers	O
at	O
NAS	B-Device
successfully	O
completed	O
a	O
"	O
live	O
integration	O
"	O
of	O
another	O
ICE	O
8200	O
rack	O
by	O
connecting	O
the	O
new	O
rack	O
's	O
InfiniBand	B-Architecture
dual	O
port	O
fabric	O
via	O
44	O
fibre	O
cables	O
while	O
the	O
supercomputer	B-Architecture
was	O
still	O
running	O
a	O
full	O
workload	O
,	O
saving	O
2	O
million	O
hours	O
in	O
productivity	O
that	O
would	O
otherwise	O
have	O
been	O
lost	O
.	O
</s>
<s>
Another	O
expansion	O
in	O
2010	O
added	O
32	O
new	O
SGI	B-Architecture
Altix	I-Architecture
ICE	O
8400	O
racks	O
with	O
Intel	B-Device
Xeon	I-Device
six-core	O
X5670	O
Westmere	O
processors	O
,	O
bringing	O
up	O
to	O
18,432	O
processors	O
(	O
81,920	O
cores	O
in	O
144	O
racks	O
)	O
at	O
a	O
theoretical	O
peak	O
of	O
973	O
teraflops	O
and	O
a	O
LINPACK	B-Device
rating	O
of	O
773	O
teraflops	O
.	O
</s>
<s>
NASA	O
also	O
put	O
an	O
emphasis	O
on	O
keeping	O
Pleiades	B-Device
energy	O
efficient	O
,	O
increasing	O
the	O
power	O
efficiency	O
with	O
each	O
expansion	O
so	O
that	O
in	O
2010	O
it	O
was	O
three	O
times	O
more	O
power-efficient	O
than	O
the	O
original	O
2008	O
components	O
,	O
which	O
were	O
the	O
most	O
power-efficient	O
at	O
the	O
time	O
.	O
</s>
<s>
The	O
integration	O
of	O
the	O
six-core	O
Westmere	O
nodes	B-Data_Structure
also	O
required	O
new	O
quad	O
data	O
rate	O
(	O
QDR	O
)	O
and	O
hybrid	O
DDR/QDR	O
InfiniBand	B-Architecture
cabling	O
,	O
making	O
the	O
world	O
's	O
largest	O
InfiniBand	B-Architecture
interconnect	O
network	O
with	O
more	O
than	O
65	O
miles	O
of	O
cable	O
.	O
</s>
<s>
After	O
another	O
14	O
ICE	O
8400	O
racks	O
containing	O
Westmere	O
processors	O
were	O
added	O
in	O
2011	O
,	O
Pleiades	B-Device
ranked	O
seventh	O
on	O
the	O
TOP500	B-Operating_System
list	O
in	O
June	O
of	O
that	O
year	O
at	O
a	O
LINPACK	B-Device
rating	O
of	O
1.09	O
petaflops	O
,	O
or	O
1.09	O
quadrillion	O
floating	B-Algorithm
point	I-Algorithm
operations	O
per	O
second	O
.	O
</s>
<s>
InfiniBand	B-Architecture
DDR	O
and	O
QDR	O
fiber	O
cables	O
are	O
used	O
to	O
connect	O
the	O
all	O
of	O
nodes	B-Data_Structure
to	O
each	O
other	O
,	O
as	O
well	O
as	O
to	O
the	O
mass	O
storage	O
systems	O
at	O
NAS	B-Device
and	O
the	O
hyperwall	O
visualization	O
system	O
,	O
creating	O
a	O
network	O
made	O
up	O
of	O
more	O
than	O
65	O
miles	O
of	O
InfiniBand	B-Architecture
fabric	O
,	O
the	O
largest	O
of	O
its	O
kind	O
in	O
the	O
world	O
.	O
</s>
<s>
Pleiades	B-Device
is	O
built	O
in	O
a	O
partial	O
11-D	O
hypercube	O
topology	O
,	O
where	O
each	O
node	O
has	O
eleven	O
connections	O
to	O
eleven	O
other	O
nodes	B-Data_Structure
,	O
with	O
some	O
making	O
up	O
to	O
twelve	O
connections	O
to	O
form	O
a	O
12-D	O
hypercube	O
.	O
</s>
<s>
In	O
2012	O
,	O
NASA	O
and	O
partners	O
SGI	O
and	O
Intel	O
began	O
working	O
on	O
the	O
integration	O
of	O
24	O
new	O
Altix	B-Architecture
ICE	O
X	O
racks	O
with	O
Intel	B-Device
Xeon	I-Device
eight-core	O
E5-2760	O
Sandy	O
Bridge	O
processors	O
to	O
replace	O
27	O
of	O
the	O
original	O
Alitx	O
8200	O
racks	O
containing	O
quad-core	O
Harpertown	O
processors	O
.	O
</s>
<s>
With	O
a	O
total	O
of	O
126,720	O
processor	O
cores	O
and	O
over	O
233	O
terabytes	O
of	O
RAM	O
across	O
182	O
racks	O
,	O
the	O
expansion	O
increased	O
Pleiades	B-Device
 '	O
available	O
computing	O
capacity	O
40	O
percent	O
.	O
</s>
<s>
Each	O
new	O
Sandy	O
Bridge	O
node	O
has	O
four	O
networking	O
links	O
using	O
fourteen	O
data	O
rate	O
(	O
FDR	O
)	O
InfiniBand	B-Architecture
cable	O
for	O
a	O
total	O
transfer	O
bandwidth	O
of	O
56	O
gigabits	O
(	O
about	O
7	O
gigabytes	O
)	O
per	O
second	O
.	O
</s>
<s>
In	O
early	O
2013	O
,	O
work	O
began	O
on	O
a	O
larger	O
hardware	O
refresh	O
for	O
Pleiades	B-Device
,	O
ultimately	O
removing	O
all	O
of	O
the	O
original	O
4-core	O
Harpertown	O
processors	O
and	O
adding	O
46	O
SGI	O
ICE	O
X	O
racks	O
with	O
10-core	O
Intel	B-Device
Xeon	I-Device
E5-2680V2	O
(	O
Ivy	O
Bridge	O
)	O
processors	O
.	O
</s>
<s>
The	O
system	O
was	O
slowly	O
upgraded	O
again	O
between	O
January	O
and	O
April	O
2014	O
,	O
adding	O
another	O
29	O
racks	O
of	O
Ivy	O
Bridge	O
nodes	B-Data_Structure
and	O
increasing	O
the	O
system	O
's	O
theoretical	O
computational	O
capability	O
to	O
3.59	O
petaflops	O
.	O
</s>
<s>
To	O
make	O
room	O
for	O
the	O
expansion	O
,	O
all	O
of	O
the	O
system	O
's	O
remaining	O
Nehalem	O
nodes	B-Data_Structure
and	O
12	O
Westmere	O
nodes	B-Data_Structure
were	O
removed	O
.	O
</s>
<s>
In	O
late	O
2014	O
,	O
more	O
Westmere	O
nodes	B-Data_Structure
were	O
removed	O
to	O
make	O
room	O
for	O
new	O
Intel	B-Device
Xeon	I-Device
Haswell	O
processors	O
,	O
increasing	O
the	O
theoretical	O
processing	O
power	O
by	O
one	O
petaflop	O
to	O
4.49	O
petaflops	O
.	O
</s>
<s>
In	O
January	O
2015	O
,	O
additional	O
Haswell	O
nodes	B-Data_Structure
were	O
installed	O
and	O
released	O
to	O
users	O
,	O
giving	O
Pleiades	B-Device
a	O
new	O
peak	O
theoretical	O
processing	O
capacity	O
of	O
5.35	O
petaflops	O
.	O
</s>
<s>
An	O
upgrade	O
,	O
completed	O
in	O
June	O
2016	O
,	O
replaced	O
all	O
remaining	O
racks	O
containing	O
nodes	B-Data_Structure
with	O
six-core	O
Intel	B-Device
Xeon	I-Device
X5670	O
(	O
Westmere	O
)	O
processors	O
with	O
racks	O
containing	O
nodes	B-Data_Structure
using	O
14-core	O
Intel	B-Device
Xeon	I-Device
E5-2680v4	O
(	O
Broadwell	O
)	O
processors	O
.	O
</s>
<s>
Pleiades	B-Device
is	O
part	O
of	O
NASA	O
's	O
High-End	O
Computing	O
Capability	O
(	O
HECC	O
)	O
Project	O
and	O
represents	O
NASA	O
's	O
state-of-the-art	O
technology	O
for	O
meeting	O
the	O
agency	O
's	O
supercomputing	B-Architecture
requirements	O
,	O
enabling	O
NASA	O
scientists	O
and	O
engineers	O
to	O
conduct	O
high-fidelity	O
modeling	O
and	O
simulation	O
for	O
NASA	O
missions	O
in	O
Earth	O
studies	O
,	O
space	O
science	O
,	O
aeronautics	O
research	O
,	O
as	O
well	O
as	O
human	O
and	O
robotic	O
space	O
exploration	O
.	O
</s>
<s>
Some	O
of	O
the	O
scientific	O
and	O
engineering	O
projects	O
run	O
on	O
Pleiades	B-Device
include	O
:	O
</s>
<s>
After	O
the	O
operations	O
center	O
gathers	O
this	O
data	O
,	O
it	O
is	O
pipelined	O
to	O
Pleiades	B-Device
in	O
order	O
to	O
calculate	O
the	O
size	O
,	O
orbit	O
,	O
and	O
location	O
of	O
the	O
planets	O
surrounding	O
these	O
stars	O
.	O
</s>
<s>
After	O
setbacks	O
following	O
the	O
failure	O
of	O
two	O
of	O
Kepler	O
's	O
four	O
reaction	O
wheels	O
,	O
responsible	O
for	O
keeping	O
the	O
spacecraft	O
pointed	O
in	O
the	O
correct	O
direction	O
,	O
in	O
2013	O
,	O
the	O
Kepler	O
team	O
moved	O
the	O
entire	O
data	O
pipeline	O
to	O
Pleiades	B-Device
,	O
which	O
continues	O
to	O
run	O
light	O
curve	O
analyses	O
from	O
the	O
existing	O
Kepler	O
data	O
.	O
</s>
<s>
Research	O
and	O
development	O
of	O
next	O
generation	O
space	O
launch	O
vehicles	O
is	O
done	O
on	O
Pleiades	B-Device
using	O
cutting-edge	O
analysis	O
tools	O
and	O
computational	O
fluid	O
dynamics	O
(	O
CFD	O
)	O
modeling	O
and	O
simulation	O
in	O
order	O
to	O
create	O
more	O
efficient	O
and	O
affordable	O
space	O
launch	O
system	O
and	O
vehicle	O
designs	O
.	O
</s>
<s>
Astrophysics	O
research	O
into	O
the	O
formation	O
of	O
galaxies	O
is	O
run	O
on	O
Pleiades	B-Device
to	O
create	O
simulations	O
of	O
how	O
our	O
own	O
Milky	O
Way	O
Galaxy	O
was	O
formed	O
and	O
what	O
forces	O
might	O
have	O
caused	O
it	O
to	O
form	O
in	O
its	O
signature	O
disk-shape	O
.	O
</s>
<s>
Pleiades	B-Device
has	O
also	O
been	O
the	O
supercomputing	B-Architecture
resource	O
for	O
dark	B-Application
matter	I-Application
research	O
and	O
simulation	O
,	O
helping	O
to	O
discover	O
gravitationally	O
bound	O
"	O
clumps	O
"	O
of	O
dark	B-Application
matter	I-Application
within	O
galaxies	O
in	O
one	O
of	O
the	O
largest	O
simulations	O
ever	O
done	O
,	O
in	O
terms	O
of	O
particle	O
numbers	O
.	O
</s>
<s>
In	O
the	O
2015	O
film	O
The	O
Martian	O
,	O
astrodynamicist	O
Rich	O
Purnell	O
uses	O
the	O
Pleiades	B-Device
supercomputer	I-Device
to	O
verify	O
calculations	O
for	O
a	O
gravity	O
assist	O
maneuver	O
for	O
a	O
spacecraft	O
,	O
in	O
order	O
to	O
rescue	O
an	O
astronaut	O
stranded	O
on	O
Mars	O
.	O
</s>
<s>
Unlike	O
what	O
is	O
shown	O
in	O
the	O
movie	O
,	O
one	O
need	O
not	O
be	O
physically	O
present	O
inside	O
the	O
racks	O
to	O
run	O
the	O
computations	O
;	O
to	O
submit	O
jobs	O
a	O
user	O
can	O
connect	O
from	O
a	O
remote	O
location	O
via	O
ssh	B-Protocol
,	O
while	O
employing	O
a	O
SecurID	O
.	O
</s>
