<s>
Reliability	B-General_Concept
,	I-General_Concept
availability	I-General_Concept
and	I-General_Concept
serviceability	I-General_Concept
(	O
RAS	O
)	O
,	O
also	O
known	O
as	O
reliability	B-General_Concept
,	I-General_Concept
availability	I-General_Concept
,	I-General_Concept
and	I-General_Concept
maintainability	I-General_Concept
(	O
RAM	B-Architecture
)	O
,	O
is	O
a	O
computer	B-Architecture
hardware	I-Architecture
engineering	O
term	O
involving	O
reliability	O
engineering	O
,	O
high	O
availability	O
,	O
and	O
serviceability	O
design	O
.	O
</s>
<s>
The	O
phrase	O
was	O
originally	O
used	O
by	O
International	O
Business	O
Machines	O
(	O
IBM	O
)	O
as	O
a	O
term	O
to	O
describe	O
the	O
robustness	O
of	O
their	O
mainframe	B-Architecture
computers	I-Architecture
.	O
</s>
<s>
Computers	O
designed	O
with	O
higher	O
levels	O
of	O
RAS	O
have	O
many	O
features	O
that	O
protect	O
data	O
integrity	O
and	O
help	O
them	O
stay	O
available	O
for	O
long	O
periods	O
of	O
time	O
without	O
failure	B-General_Concept
This	O
data	O
integrity	O
and	O
uptime	B-Device
is	O
a	O
particular	O
selling	O
point	O
for	O
mainframes	B-Architecture
and	O
fault-tolerant	B-General_Concept
systems	I-General_Concept
.	O
</s>
<s>
While	O
RAS	O
originated	O
as	O
a	O
hardware-oriented	O
term	O
,	O
systems	O
thinking	O
has	O
extended	O
the	O
concept	O
of	O
reliability-availability-serviceability	O
to	O
systems	O
in	O
general	O
,	O
including	O
software	O
.	O
</s>
<s>
Reliability	O
can	O
be	O
defined	O
as	O
the	O
probability	O
that	O
a	O
system	O
will	O
produce	O
correct	O
outputs	O
up	O
to	O
some	O
given	O
time	O
t	O
.	O
Reliability	O
is	O
enhanced	O
by	O
features	O
that	O
help	O
to	O
avoid	O
,	O
detect	O
and	O
repair	O
hardware	B-Architecture
faults	O
.	O
</s>
<s>
Instead	O
,	O
it	O
detects	O
and	O
,	O
if	O
possible	O
,	O
corrects	O
the	O
corruption	O
,	O
for	O
example	O
:	O
by	O
retrying	O
an	O
operation	O
for	O
transient	O
(	O
soft	O
)	O
or	O
intermittent	O
errors	O
,	O
or	O
else	O
,	O
for	O
uncorrectable	O
errors	O
,	O
isolating	O
the	O
fault	O
and	O
reporting	O
it	O
to	O
higher-level	O
recovery	O
mechanisms	O
(	O
which	O
may	O
failover	B-General_Concept
to	O
redundant	O
replacement	O
hardware	B-Architecture
,	O
etc	O
.	O
</s>
<s>
High-availability	B-General_Concept
systems	O
may	O
report	O
availability	O
in	O
terms	O
of	O
minutes	O
or	O
hours	O
of	O
downtime	O
per	O
year	O
.	O
</s>
<s>
Permanent	O
faults	O
lead	O
to	O
a	O
continuing	O
error	O
and	O
are	O
typically	O
due	O
to	O
some	O
physical	O
failure	B-General_Concept
such	O
as	O
metal	O
electromigration	O
or	O
dielectric	O
breakdown	O
.	O
</s>
<s>
soft	O
)	O
faults	O
lead	O
to	O
independent	O
one-time	O
errors	O
and	O
are	O
not	O
due	O
to	O
permanent	O
hardware	B-Architecture
faults	O
:	O
examples	O
include	O
alpha	O
particles	O
flipping	O
a	O
memory	B-General_Concept
bit	O
,	O
electromagnetic	O
noise	O
,	O
or	O
power-supply	O
fluctuations	O
.	O
</s>
<s>
Transient	O
and	O
intermittent	O
faults	O
can	O
typically	O
be	O
handled	O
by	O
detection	O
and	O
correction	O
by	O
e.g.	O
,	O
ECC	B-Error_Name
codes	O
or	O
instruction	O
replay	O
(	O
see	O
below	O
)	O
.	O
</s>
<s>
Permanent	O
faults	O
will	O
lead	O
to	O
uncorrectable	O
errors	O
which	O
can	O
be	O
handled	O
by	O
replacement	O
by	O
duplicate	O
hardware	B-Architecture
,	O
e.g.	O
,	O
processor	B-General_Concept
sparing	O
,	O
or	O
by	O
the	O
passing	O
of	O
the	O
uncorrectable	O
error	O
to	O
high	O
level	O
recovery	O
mechanisms	O
.	O
</s>
<s>
A	O
successfully	O
corrected	O
intermittent	O
fault	O
can	O
also	O
be	O
reported	O
to	O
the	O
operating	B-General_Concept
system	I-General_Concept
(	O
OS	O
)	O
to	O
provide	O
information	O
for	O
predictive	B-Device
failure	I-Device
analysis	I-Device
.	O
</s>
<s>
Example	O
hardware	B-Architecture
features	O
for	O
improving	O
RAS	O
include	O
the	O
following	O
,	O
listed	O
by	O
subsystem	O
:	O
</s>
<s>
Processor	B-General_Concept
:	O
</s>
<s>
Processor	B-General_Concept
instruction	O
error	O
detection	O
(	O
e.g.	O
</s>
<s>
alternative	O
processor	B-General_Concept
recovery	O
in	O
IBM	O
mainframes	B-Architecture
,	O
or	O
"	O
Instruction	O
replay	O
technology	O
"	O
in	O
Itanium	B-General_Concept
systems	O
.	O
</s>
<s>
Processors	O
running	O
in	O
lock-step	B-General_Concept
to	O
perform	O
master-checker	B-Operating_System
or	O
voting	O
schemes	O
.	O
</s>
<s>
Machine	B-Device
Check	I-Device
Architecture	I-Device
and	O
ACPI	O
Platform	O
Error	O
Interface	O
to	O
report	O
errors	O
to	O
the	O
OS	O
.	O
</s>
<s>
Memory	B-General_Concept
:	O
</s>
<s>
Parity	B-Error_Name
or	O
ECC	B-Error_Name
(	O
including	O
single	B-Error_Name
device	I-Error_Name
correction	I-Error_Name
)	O
protection	O
of	O
memory	B-General_Concept
components	O
(	O
cache	B-General_Concept
and	O
main	O
memory	B-General_Concept
)	O
;	O
bad	O
cache	B-General_Concept
line	O
disabling	O
;	O
memory	B-General_Concept
scrubbing	I-General_Concept
;	O
memory	B-General_Concept
sparing	O
,	O
memory	B-General_Concept
mirroring	O
;	O
bad	O
page	O
offlining	O
;	O
redundant	B-Error_Name
bit	I-Error_Name
steering	I-Error_Name
;	O
redundant	B-Error_Name
array	I-Error_Name
of	I-Error_Name
independent	I-Error_Name
memory	I-Error_Name
(	O
RAIM	O
)	O
.	O
</s>
<s>
I/O	B-General_Concept
:	O
</s>
<s>
Cyclic	O
redundancy	O
check	O
checksums	B-Algorithm
for	O
data	O
transmission/retry	O
and	O
data	O
storage	B-General_Concept
,	O
e.g.	O
</s>
<s>
PCI	O
Express	O
(	O
PCIe	O
)	O
Advanced	O
Error	O
Reporting	O
(	O
AER	O
)	O
,	O
redundant	B-General_Concept
I/O	I-General_Concept
paths	I-General_Concept
.	O
</s>
<s>
Storage	B-General_Concept
:	O
</s>
<s>
RAID	B-Architecture
configurations	O
for	O
hard	B-Device
disk	I-Device
drive	I-Device
and	O
solid	B-Device
state	I-Device
drive	I-Device
storage	B-General_Concept
.	O
</s>
<s>
Journaling	B-Application
file	I-Application
systems	I-Application
for	O
file	O
repair	O
after	O
crashes	O
.	O
</s>
<s>
Checksums	B-Algorithm
on	O
both	O
data	O
and	O
metadata	O
,	O
and	O
background	O
scrubbing	B-Error_Name
.	O
</s>
<s>
Self-Monitoring	O
,	O
Analysis	O
,	O
and	O
Reporting	O
Technology	O
for	O
hard	B-Device
disk	I-Device
drive	I-Device
and	O
solid	B-Device
state	I-Device
drive	I-Device
.	O
</s>
<s>
Duplicating	O
components	O
to	O
avoid	O
single	B-Architecture
points	I-Architecture
of	I-Architecture
failure	I-Architecture
,	O
e.g.	O
,	O
power-supplies	O
.	O
</s>
<s>
Hot	B-Device
swapping	I-Device
of	O
components	O
:	O
CPUs	B-Device
,	O
RAMs	B-Architecture
,	O
hard	B-Device
disk	I-Device
drives	I-Device
and	O
solid	B-Device
state	I-Device
drives	I-Device
.	O
</s>
<s>
Predictive	B-Device
failure	I-Device
analysis	I-Device
to	O
predict	O
which	O
intermittent	O
correctable	O
errors	O
will	O
lead	O
eventually	O
to	O
hard	O
non-correctable	O
errors	O
.	O
</s>
<s>
Partitioning/domaining	B-Device
of	O
computer	B-Architecture
components	I-Architecture
to	O
allow	O
one	O
large	O
system	O
to	O
act	O
as	O
several	O
smaller	O
systems	O
.	O
</s>
<s>
Virtual	B-Architecture
machines	I-Architecture
to	O
decrease	O
the	O
severity	O
of	O
operating	B-General_Concept
system	I-General_Concept
software	O
faults	O
.	O
</s>
<s>
Redundant	O
I/O	B-General_Concept
domains	O
or	O
I/O	B-General_Concept
partitions	B-Device
for	O
providing	O
virtual	O
I/O	B-General_Concept
to	O
guest	O
virtual	B-Architecture
machines	I-Architecture
.	O
</s>
<s>
Computer	B-Architecture
clustering	I-Architecture
capability	O
with	O
failover	B-General_Concept
capability	O
,	O
for	O
complete	O
redundancy	O
of	O
hardware	B-Architecture
and	O
software	O
.	O
</s>
<s>
Dynamic	B-Architecture
software	I-Architecture
updating	I-Architecture
to	O
avoid	O
the	O
need	O
to	O
reboot	O
the	O
system	O
for	O
a	O
kernel	B-Operating_System
software	O
update	O
,	O
for	O
example	O
Ksplice	B-Application
under	O
Linux	O
.	O
</s>
<s>
Independent	O
management	O
processor	B-General_Concept
for	O
serviceability	O
:	O
remote	O
monitoring	O
,	O
alerting	O
and	O
control	O
.	O
</s>
<s>
Fault-tolerant	B-General_Concept
designs	I-General_Concept
extended	O
the	O
idea	O
by	O
making	O
RAS	O
to	O
be	O
the	O
defining	O
feature	O
of	O
their	O
computers	O
for	O
applications	O
like	O
stock	O
market	O
exchanges	O
or	O
air	B-Application
traffic	I-Application
control	I-Application
,	O
where	O
system	O
crashes	O
would	O
be	O
catastrophic	O
.	O
</s>
<s>
Fault-tolerant	B-General_Concept
computers	I-General_Concept
(	O
e.g.	O
,	O
see	O
Tandem	O
Computers	O
and	O
Stratus	B-General_Concept
Technologies	I-General_Concept
)	O
,	O
which	O
tend	O
to	O
have	O
duplicate	O
components	O
running	O
in	O
lock-step	B-General_Concept
for	O
reliability	O
,	O
have	O
become	O
less	O
popular	O
,	O
due	O
to	O
their	O
high	O
cost	O
.	O
</s>
<s>
High	B-General_Concept
availability	I-General_Concept
systems	I-General_Concept
,	O
using	O
distributed	B-Architecture
computing	I-Architecture
techniques	O
like	O
computer	B-Architecture
clusters	I-Architecture
,	O
are	O
often	O
used	O
as	O
cheaper	O
alternatives	O
.	O
</s>
