<s>
Recovery-oriented	B-Protocol
computing	I-Protocol
(	O
sometimes	O
abbreviated	O
to	O
ROC	O
)	O
is	O
a	O
method	O
constructed	O
at	O
Stanford	O
University	O
and	O
the	O
University	O
of	O
California	O
,	O
Berkeley	O
for	O
developing	O
reliable	O
Internet	O
services	O
.	O
</s>
<s>
Its	O
proponents	O
seek	O
to	O
recognize	O
computer	B-Error_Name
bugs	I-Error_Name
as	O
inevitable	O
,	O
and	O
then	O
reduce	O
their	O
harmful	O
effects	O
.	O
</s>
<s>
Isolation	O
must	O
be	O
failure	O
proof	O
for	O
all	O
types	O
of	O
failures	O
whether	O
they	O
be	O
software	B-Error_Name
or	O
human	O
caused	O
failures	O
.	O
</s>
<s>
One	O
potential	O
way	O
to	O
isolate	O
parts	O
of	O
a	O
system	O
is	O
using	O
virtual	O
machine	O
monitors	O
such	O
as	O
Xen	B-Operating_System
.	O
</s>
<s>
The	O
ability	O
to	O
undo	B-Application
across	O
different	O
programs	O
and	O
time	O
frames	O
is	O
necessary	O
in	O
this	O
type	O
of	O
system	O
because	O
human	O
error	O
is	O
the	O
cause	O
of	O
about	O
half	O
of	O
system	O
failures	O
.	O
</s>
<s>
Not	O
having	O
undo	B-Application
support	O
also	O
limits	O
testing	O
aspects	O
of	O
a	O
production	O
system	O
because	O
it	O
does	O
n’t	O
allow	O
for	O
trial	O
and	O
error	O
.	O
</s>
<s>
System-wide	O
undo	B-Application
support	O
should	O
cover	O
all	O
aspects	O
of	O
the	O
system	O
.	O
</s>
<s>
This	O
includes	O
hardware	O
and	O
software	B-Error_Name
upgrades	O
,	O
configuration	O
as	O
well	O
as	O
application	O
management	O
.	O
</s>
<s>
Software	B-General_Concept
aging	I-General_Concept
problems	O
are	O
best	O
resolved	O
by	O
restarting	O
the	O
component	O
that	O
is	O
affected	O
.	O
</s>
