<s>
In	O
computer	B-General_Concept
science	I-General_Concept
,	O
distributed	B-Operating_System
memory	I-Operating_System
refers	O
to	O
a	O
multiprocessor	B-Operating_System
computer	I-Operating_System
system	I-Operating_System
in	O
which	O
each	O
processor	B-General_Concept
has	O
its	O
own	O
private	O
memory	B-General_Concept
.	O
</s>
<s>
In	O
contrast	O
,	O
a	O
shared	B-Operating_System
memory	I-Operating_System
multiprocessor	B-Operating_System
offers	O
a	O
single	O
memory	B-General_Concept
space	O
used	O
by	O
all	O
processors	O
.	O
</s>
<s>
In	O
a	O
distributed	B-Operating_System
memory	I-Operating_System
system	O
there	O
is	O
typically	O
a	O
processor	B-General_Concept
,	O
a	O
memory	B-General_Concept
,	O
and	O
some	O
form	O
of	O
interconnection	O
that	O
allows	O
programs	O
on	O
each	O
processor	B-General_Concept
to	O
interact	O
with	O
each	O
other	O
.	O
</s>
<s>
The	O
network	B-Architecture
topology	I-Architecture
is	O
a	O
key	O
factor	O
in	O
determining	O
how	O
the	O
multiprocessor	B-Operating_System
machine	O
scales	B-Architecture
.	O
</s>
<s>
The	O
links	O
between	O
nodes	O
can	O
be	O
implemented	O
using	O
some	O
standard	O
network	O
protocol	O
(	O
for	O
example	O
Ethernet	O
)	O
,	O
using	O
bespoke	O
network	O
links	O
(	O
used	O
in	O
for	O
example	O
the	O
transputer	B-General_Concept
)	O
,	O
or	O
using	O
dual-ported	B-General_Concept
memories	I-General_Concept
.	O
</s>
<s>
The	O
key	O
issue	O
in	O
programming	O
distributed	B-Operating_System
memory	I-Operating_System
systems	O
is	O
how	O
to	O
distribute	O
the	O
data	O
over	O
the	O
memories	O
.	O
</s>
<s>
(	O
the	O
result	O
is	O
h(g(f(x )	O
)	O
)	O
)	O
,	O
then	O
this	O
can	O
be	O
expressed	O
as	O
a	O
distributed	B-Operating_System
memory	I-Operating_System
problem	O
where	O
the	O
data	O
is	O
transmitted	O
first	O
to	O
the	O
node	O
that	O
performs	O
f	O
that	O
passes	O
the	O
result	O
onto	O
the	O
second	O
node	O
that	O
computes	O
g	O
,	O
and	O
finally	O
to	O
the	O
third	O
node	O
that	O
computes	O
h	O
.	O
This	O
is	O
also	O
known	O
as	O
systolic	B-Architecture
computation	I-Architecture
.	O
</s>
<s>
Similarly	O
,	O
in	O
distributed	B-Operating_System
shared	I-Operating_System
memory	I-Operating_System
each	O
node	O
of	O
a	O
cluster	O
has	O
access	O
to	O
a	O
large	O
shared	B-Operating_System
memory	I-Operating_System
in	O
addition	O
to	O
each	O
node	O
's	O
limited	O
non-shared	O
private	O
memory	B-General_Concept
.	O
</s>
<s>
The	O
advantage	O
of	O
(	O
distributed	O
)	O
shared	B-Operating_System
memory	I-Operating_System
is	O
that	O
it	O
offers	O
a	O
unified	O
address	O
space	O
in	O
which	O
all	O
data	O
can	O
be	O
found	O
.	O
</s>
<s>
The	O
advantage	O
of	O
distributed	B-Operating_System
memory	I-Operating_System
is	O
that	O
it	O
excludes	O
race	O
conditions	O
,	O
and	O
that	O
it	O
forces	O
the	O
programmer	O
to	O
think	O
about	O
data	O
distribution	O
.	O
</s>
<s>
Distributed	B-Operating_System
shared	I-Operating_System
memory	I-Operating_System
hides	O
the	O
mechanism	O
of	O
communication	O
,	O
it	O
does	O
not	O
hide	O
the	O
latency	O
of	O
communication	O
.	O
</s>
