<s>
A	O
network	B-Algorithm
partition	I-Algorithm
is	O
a	O
division	O
of	O
a	O
computer	O
network	O
into	O
relatively	O
independent	O
subnets	B-Protocol
,	O
either	O
by	O
design	O
,	O
to	O
optimize	O
them	O
separately	O
,	O
or	O
due	O
to	O
the	O
failure	O
of	O
network	O
devices	O
.	O
</s>
<s>
For	O
example	O
,	O
in	O
a	O
network	O
with	O
multiple	O
subnets	B-Protocol
where	O
nodes	O
A	O
and	O
B	O
are	O
located	O
in	O
one	O
subnet	B-Protocol
and	O
nodes	O
C	O
and	O
D	O
are	O
in	O
another	O
,	O
a	O
partition	O
occurs	O
if	O
the	O
network	B-Protocol
switch	I-Protocol
device	O
between	O
the	O
two	O
subnets	B-Protocol
fails	O
.	O
</s>
<s>
The	O
CAP	B-Operating_System
Theorem	I-Operating_System
is	O
based	O
on	O
three	O
trade-offs	O
:	O
consistency	B-General_Concept
,	O
availability	O
,	O
and	O
partition	O
tolerance	O
.	O
</s>
<s>
Partition	O
tolerance	O
,	O
in	O
this	O
context	O
,	O
means	O
the	O
ability	O
of	O
a	O
data	O
processing	O
system	O
to	O
continue	O
processing	O
data	O
even	O
if	O
a	O
network	B-Algorithm
partition	I-Algorithm
causes	O
communication	O
errors	O
between	O
subsystems	O
.	O
</s>
