<s>
The	O
number	O
of	O
traversed	B-Operating_System
edges	I-Operating_System
per	I-Operating_System
second	I-Operating_System
(	O
TEPS	O
)	O
that	O
can	O
be	O
performed	O
by	O
a	O
supercomputer	B-Architecture
cluster	B-Architecture
is	O
a	O
measure	O
of	O
both	O
the	O
communications	O
capabilities	O
and	O
computational	O
power	O
of	O
the	O
machine	O
.	O
</s>
<s>
The	O
term	O
first	O
entered	O
usage	O
in	O
2010	O
with	O
the	O
advent	O
of	O
petascale	B-General_Concept
computing	I-General_Concept
,	O
and	O
has	O
since	O
been	O
measured	O
for	O
many	O
of	O
the	O
world	O
's	O
largest	O
supercomputers	B-Architecture
.	O
</s>
<s>
In	O
this	O
context	O
,	O
an	O
edge	O
is	O
a	O
connection	O
between	O
two	O
vertices	O
on	O
a	O
graph	B-Application
,	O
and	O
the	O
traversal	O
is	O
the	O
ability	O
of	O
the	O
machine	O
to	O
communicate	O
data	O
between	O
these	O
two	O
points	O
.	O
</s>
<s>
The	O
standardized	O
benchmark	O
associated	O
with	O
Graph500	B-Device
,	O
as	O
of	O
September	O
,	O
2011	O
,	O
calls	O
for	O
executing	O
graph	B-Application
generation	O
and	O
search	O
algorithms	O
on	O
graphs	O
as	O
large	O
as	O
1.1	O
Petabyte	O
.	O
</s>
<s>
The	O
ability	O
of	O
an	O
application	O
to	O
utilize	O
a	O
supercomputer	B-Architecture
cluster	B-Architecture
effectively	O
depends	O
not	O
only	O
on	O
the	O
raw	O
speed	O
of	O
each	O
processor	O
,	O
but	O
also	O
on	O
the	O
communication	O
network	O
.	O
</s>
<s>
The	O
importance	O
of	O
communication	O
capability	O
varies	O
from	O
application	O
to	O
application	O
,	O
but	O
it	O
is	O
clear	O
that	O
the	O
LINPACK	B-Device
benchmarks	I-Device
traditionally	O
used	O
for	O
rating	O
the	O
FLOPS	O
of	O
supercomputers	B-Architecture
do	O
not	O
require	O
the	O
same	O
communications	O
capability	O
as	O
many	O
scientific	O
applications	O
.	O
</s>
