<s>
In	O
computer	B-General_Concept
storage	I-General_Concept
,	O
the	O
standard	B-General_Concept
RAID	I-General_Concept
levels	I-General_Concept
comprise	O
a	O
basic	O
set	O
of	O
RAID	B-Architecture
(	O
"	O
redundant	B-Architecture
array	I-Architecture
of	I-Architecture
independent	I-Architecture
disks	I-Architecture
"	O
or	O
"	O
redundant	B-Architecture
array	I-Architecture
of	I-Architecture
inexpensive	I-Architecture
disks	I-Architecture
"	O
)	O
configurations	O
that	O
employ	O
the	O
techniques	O
of	O
striping	B-General_Concept
,	O
mirroring	B-General_Concept
,	O
or	O
parity	B-Error_Name
to	O
create	O
large	O
reliable	O
data	O
stores	O
from	O
multiple	O
general-purpose	O
computer	O
hard	B-Device
disk	I-Device
drives	I-Device
(	O
HDDs	O
)	O
.	O
</s>
<s>
The	O
most	O
common	O
types	O
are	O
RAID0	O
(	O
striping	B-General_Concept
)	O
,	O
RAID1	O
(	O
mirroring	B-General_Concept
)	O
and	O
its	O
variants	O
,	O
RAID5	O
(	O
distributed	O
parity	B-Error_Name
)	O
,	O
and	O
RAID6	O
(	O
dual	O
parity	B-Error_Name
)	O
.	O
</s>
<s>
Multiple	O
RAID	B-General_Concept
levels	I-General_Concept
can	O
also	O
be	O
combined	O
or	O
nested	O
,	O
for	O
instance	O
RAID10	O
(	O
striping	B-General_Concept
of	O
mirrors	O
)	O
or	O
RAID01	O
(	O
mirroring	B-General_Concept
stripe	O
sets	O
)	O
.	O
</s>
<s>
RAID	B-General_Concept
levels	I-General_Concept
and	O
their	O
associated	O
data	O
formats	O
are	O
standardized	O
by	O
the	O
Storage	O
Networking	O
Industry	O
Association	O
(	O
SNIA	O
)	O
in	O
the	O
Common	O
RAID	B-Architecture
Disk	O
Drive	O
Format	O
(	O
DDF	O
)	O
standard	O
.	O
</s>
<s>
While	O
most	O
RAID	B-General_Concept
levels	I-General_Concept
can	O
provide	O
good	O
protection	O
against	O
and	O
recovery	O
from	O
hardware	O
defects	O
or	O
defective	O
sectors/read	O
errors	O
(	O
hard	O
errors	O
)	O
,	O
they	O
do	O
not	O
provide	O
any	O
protection	O
against	O
data	B-General_Concept
loss	I-General_Concept
due	O
to	O
catastrophic	O
failures	O
(	O
fire	O
,	O
water	O
)	O
or	O
soft	O
errors	O
such	O
as	O
user	O
error	O
,	O
software	O
malfunction	O
,	O
or	O
malware	O
infection	O
.	O
</s>
<s>
For	O
valuable	O
data	O
,	O
RAID	B-Architecture
is	O
only	O
one	O
building	O
block	B-General_Concept
of	O
a	O
larger	O
data	B-General_Concept
loss	I-General_Concept
prevention	O
and	O
recovery	O
scheme	O
it	O
cannot	O
replace	O
a	O
backup	B-Protocol
plan	O
.	O
</s>
<s>
RAID	B-Architecture
0	O
(	O
also	O
known	O
as	O
a	O
stripe	O
set	O
or	O
striped	O
volume	B-Application
)	O
splits	O
(	O
"	O
stripes	B-General_Concept
"	O
)	O
data	O
evenly	O
across	O
two	O
or	O
more	O
disks	O
,	O
without	O
parity	B-Error_Name
information	O
,	O
redundancy	O
,	O
or	O
fault	B-General_Concept
tolerance	I-General_Concept
.	O
</s>
<s>
Since	O
RAID0	O
provides	O
no	O
fault	B-General_Concept
tolerance	I-General_Concept
or	O
redundancy	O
,	O
the	O
failure	O
of	O
one	O
drive	O
will	O
cause	O
the	O
entire	O
array	O
to	O
fail	O
;	O
as	O
a	O
result	O
of	O
having	O
data	O
striped	O
across	O
all	O
disks	O
,	O
the	O
failure	O
will	O
result	O
in	O
total	O
data	B-General_Concept
loss	I-General_Concept
.	O
</s>
<s>
RAID0	O
is	O
normally	O
used	O
to	O
increase	O
performance	O
,	O
although	O
it	O
can	O
also	O
be	O
used	O
as	O
a	O
way	O
to	O
create	O
a	O
large	O
logical	O
volume	B-Application
out	O
of	O
two	O
or	O
more	O
physical	O
disks	O
.	O
</s>
<s>
However	O
,	O
some	O
RAID	B-Architecture
implementations	O
would	O
allow	O
the	O
remaining	O
200GB	O
to	O
be	O
used	O
for	O
other	O
purposes	O
.	O
</s>
<s>
The	O
diagram	O
in	O
this	O
section	O
shows	O
how	O
the	O
data	O
is	O
distributed	O
into	O
stripes	B-General_Concept
on	O
two	O
disks	O
,	O
with	O
A1:A2	O
as	O
the	O
first	O
stripe	O
,	O
A3:A4	O
as	O
the	O
second	O
one	O
,	O
etc	O
.	O
</s>
<s>
Since	O
the	O
stripes	B-General_Concept
are	O
accessed	O
in	O
parallel	O
,	O
an	O
-drive	O
RAID0	O
array	O
appears	O
as	O
a	O
single	O
large	O
disk	O
with	O
a	O
data	O
rate	O
times	O
higher	O
than	O
the	O
single-disk	O
rate	O
.	O
</s>
<s>
Another	O
article	O
examined	O
these	O
claims	O
and	O
concluded	O
that	O
"	O
striping	B-General_Concept
does	O
not	O
always	O
increase	O
performance	O
(	O
in	O
certain	O
situations	O
it	O
will	O
actually	O
be	O
slower	O
than	O
a	O
non-RAID	O
setup	O
)	O
,	O
but	O
in	O
most	O
situations	O
it	O
will	O
yield	O
a	O
significant	O
improvement	O
in	O
performance	O
"	O
.	O
</s>
<s>
RAID	B-Architecture
1	O
consists	O
of	O
an	O
exact	O
copy	O
(	O
or	O
mirror	B-General_Concept
)	O
of	O
a	O
set	O
of	O
data	O
on	O
two	O
or	O
more	O
disks	O
;	O
a	O
classic	O
RAID1	O
mirrored	O
pair	O
contains	O
two	O
disks	O
.	O
</s>
<s>
This	O
configuration	O
offers	O
no	O
parity	B-Error_Name
,	O
striping	B-General_Concept
,	O
or	O
spanning	O
of	O
disk	B-General_Concept
space	I-General_Concept
across	O
multiple	O
disks	O
,	O
since	O
the	O
data	O
is	O
mirrored	O
on	O
all	O
disks	O
belonging	O
to	O
the	O
array	O
,	O
and	O
the	O
array	O
can	O
only	O
be	O
as	O
big	O
as	O
the	O
smallest	O
member	O
disk	O
.	O
</s>
<s>
This	O
layout	O
is	O
useful	O
when	O
read	O
performance	O
or	O
reliability	O
is	O
more	O
important	O
than	O
write	O
performance	O
or	O
the	O
resulting	O
data	O
storage	B-General_Concept
capacity	I-General_Concept
.	O
</s>
<s>
Any	O
read	O
request	O
can	O
be	O
serviced	O
and	O
handled	O
by	O
any	O
drive	O
in	O
the	O
array	O
;	O
thus	O
,	O
depending	O
on	O
the	O
nature	O
of	O
I/O	B-General_Concept
load	O
,	O
random	O
read	O
performance	O
of	O
a	O
RAID1	O
array	O
may	O
equal	O
up	O
to	O
the	O
sum	O
of	O
each	O
member	O
's	O
performance	O
,	O
while	O
the	O
write	O
performance	O
remains	O
at	O
the	O
level	O
of	O
a	O
single	O
disk	O
.	O
</s>
<s>
RAID	B-Architecture
2	O
,	O
which	O
is	O
rarely	O
used	O
in	O
practice	O
,	O
stripes	B-General_Concept
data	O
at	O
the	O
bit	O
(	O
rather	O
than	O
block	B-General_Concept
)	O
level	O
,	O
and	O
uses	O
a	O
Hamming	B-Error_Name
code	I-Error_Name
for	O
error	B-Error_Name
correction	I-Error_Name
.	O
</s>
<s>
However	O
,	O
depending	O
with	O
a	O
high	O
rate	O
Hamming	B-Error_Name
code	I-Error_Name
,	O
many	O
spindles	O
would	O
operate	O
in	O
parallel	O
to	O
simultaneously	O
transfer	O
data	O
so	O
that	O
"	O
very	O
high	O
data	O
transfer	O
rates	O
"	O
are	O
possible	O
as	O
for	O
example	O
in	O
the	O
DataVault	B-Device
where	O
32	O
data	O
bits	O
were	O
transmitted	O
simultaneously	O
.	O
</s>
<s>
With	O
all	O
hard	B-Device
disk	I-Device
drives	I-Device
implementing	O
internal	O
error	B-Error_Name
correction	I-Error_Name
,	O
the	O
complexity	O
of	O
an	O
external	O
Hamming	B-Error_Name
code	I-Error_Name
offered	O
little	O
advantage	O
over	O
parity	B-Error_Name
so	O
RAID2	O
has	O
been	O
rarely	O
implemented	O
;	O
it	O
is	O
the	O
only	O
original	O
level	O
of	O
RAID	B-Architecture
that	O
is	O
not	O
currently	O
used	O
.	O
</s>
<s>
RAID	B-Architecture
3	O
,	O
which	O
is	O
rarely	O
used	O
in	O
practice	O
,	O
consists	O
of	O
byte-level	O
striping	B-General_Concept
with	O
a	O
dedicated	O
parity	B-Error_Name
disk	O
.	O
</s>
<s>
One	O
of	O
the	O
characteristics	O
of	O
RAID3	O
is	O
that	O
it	O
generally	O
cannot	O
service	O
multiple	O
requests	O
simultaneously	O
,	O
which	O
happens	O
because	O
any	O
single	O
block	B-General_Concept
of	O
data	O
will	O
,	O
by	O
definition	O
,	O
be	O
spread	O
across	O
all	O
members	O
of	O
the	O
set	O
and	O
will	O
reside	O
in	O
the	O
same	O
physical	O
location	O
on	O
each	O
disk	O
.	O
</s>
<s>
Therefore	O
,	O
any	O
I/O	B-General_Concept
operation	I-General_Concept
requires	O
activity	O
on	O
every	O
disk	O
and	O
usually	O
requires	O
synchronized	O
spindles	O
.	O
</s>
<s>
This	O
makes	O
it	O
suitable	O
for	O
applications	O
that	O
demand	O
the	O
highest	O
transfer	O
rates	O
in	O
long	O
sequential	O
reads	O
and	O
writes	O
,	O
for	O
example	O
uncompressed	B-Device
video	I-Device
editing	O
.	O
</s>
<s>
The	O
requirement	O
that	O
all	O
disks	O
spin	O
synchronously	O
(	O
in	O
a	O
lockstep	B-General_Concept
)	O
added	O
design	O
considerations	O
that	O
provided	O
no	O
significant	O
advantages	O
over	O
other	O
RAID	B-General_Concept
levels	I-General_Concept
.	O
</s>
<s>
RAID	B-Architecture
4	O
consists	O
of	O
block-level	O
striping	B-General_Concept
with	O
a	O
dedicated	O
parity	B-Error_Name
disk	O
.	O
</s>
<s>
As	O
a	O
result	O
of	O
its	O
layout	O
,	O
RAID4	O
provides	O
good	O
performance	O
of	O
random	O
reads	O
,	O
while	O
the	O
performance	O
of	O
random	O
writes	O
is	O
low	O
due	O
to	O
the	O
need	O
to	O
write	O
all	O
parity	B-Error_Name
data	O
to	O
a	O
single	O
disk	O
,	O
unless	O
the	O
filesystem	O
is	O
RAID-4-aware	O
and	O
compensates	O
for	O
that	O
.	O
</s>
<s>
An	O
advantage	O
of	O
RAID	B-Architecture
4	O
is	O
that	O
it	O
can	O
be	O
quickly	O
extended	O
online	O
,	O
without	O
parity	B-Error_Name
recomputation	O
,	O
as	O
long	O
as	O
the	O
newly	O
added	O
disks	O
are	O
completely	O
filled	O
with	O
0-bytes	O
.	O
</s>
<s>
In	O
diagram	O
1	O
,	O
a	O
read	O
request	O
for	O
block	B-General_Concept
A1	O
would	O
be	O
serviced	O
by	O
disk	O
0	O
.	O
</s>
<s>
A	O
simultaneous	O
read	O
request	O
for	O
block	B-General_Concept
B1	O
would	O
have	O
to	O
wait	O
,	O
but	O
a	O
read	O
request	O
for	O
B2	O
could	O
be	O
serviced	O
concurrently	O
by	O
disk	O
1	O
.	O
</s>
<s>
RAID	B-Architecture
5	O
consists	O
of	O
block-level	O
striping	B-General_Concept
with	O
distributed	O
parity	B-Error_Name
.	O
</s>
<s>
Unlike	O
in	O
RAID4	O
,	O
parity	B-Error_Name
information	O
is	O
distributed	O
among	O
the	O
drives	O
.	O
</s>
<s>
Upon	O
failure	O
of	O
a	O
single	O
drive	O
,	O
subsequent	O
reads	O
can	O
be	O
calculated	O
from	O
the	O
distributed	O
parity	B-Error_Name
such	O
that	O
no	O
data	O
is	O
lost	O
.	O
</s>
<s>
There	O
are	O
many	O
layouts	O
of	O
data	O
and	O
parity	B-Error_Name
in	O
a	O
RAID	B-Architecture
5	O
disk	O
drive	O
array	O
depending	O
upon	O
the	O
sequence	O
of	O
writing	O
across	O
the	O
disks	O
,	O
that	O
is	O
:	O
</s>
<s>
the	O
location	O
of	O
the	O
parity	B-Error_Name
block	I-Error_Name
at	O
the	O
beginning	O
or	O
end	O
of	O
the	O
stripe	O
.	O
</s>
<s>
the	O
location	O
of	O
the	O
first	O
block	B-General_Concept
of	O
a	O
stripe	O
with	O
respect	O
to	O
parity	B-Error_Name
of	O
the	O
previous	O
stripe	O
.	O
</s>
<s>
The	O
figure	O
to	O
the	O
right	O
shows	O
1	O
)	O
data	O
blocks	O
written	O
left	O
to	O
right	O
,	O
2	O
)	O
the	O
parity	B-Error_Name
block	I-Error_Name
at	O
the	O
end	O
of	O
the	O
stripe	O
and	O
3	O
)	O
the	O
first	O
block	B-General_Concept
of	O
the	O
next	O
stripe	O
not	O
on	O
the	O
same	O
disk	O
as	O
the	O
parity	B-Error_Name
block	I-Error_Name
of	O
the	O
previous	O
stripe	O
.	O
</s>
<s>
It	O
can	O
be	O
designated	O
as	O
a	O
Left	O
Asynchronous	O
RAID	B-Architecture
5	O
layout	O
and	O
this	O
is	O
the	O
only	O
layout	O
identified	O
in	O
the	O
last	O
edition	O
of	O
The	O
Raid	B-Architecture
Book	O
published	O
by	O
the	O
defunct	O
Raid	B-Architecture
Advisory	O
Board	O
.	O
</s>
<s>
In	O
a	O
Synchronous	O
layout	O
the	O
data	O
first	O
block	B-General_Concept
of	O
the	O
next	O
stripe	O
is	O
written	O
on	O
the	O
same	O
drive	O
as	O
the	O
parity	B-Error_Name
block	I-Error_Name
of	O
the	O
previous	O
stripe	O
.	O
</s>
<s>
In	O
comparison	O
to	O
RAID4	O
,	O
RAID5	O
's	O
distributed	O
parity	B-Error_Name
evens	O
out	O
the	O
stress	O
of	O
a	O
dedicated	O
parity	B-Error_Name
disk	O
among	O
all	O
RAID	B-Architecture
members	O
.	O
</s>
<s>
Additionally	O
,	O
write	O
performance	O
is	O
increased	O
since	O
all	O
RAID	B-Architecture
members	O
participate	O
in	O
the	O
serving	O
of	O
write	O
requests	O
.	O
</s>
<s>
Although	O
it	O
will	O
not	O
be	O
as	O
efficient	O
as	O
a	O
striping	B-General_Concept
(	O
RAID0	O
)	O
setup	O
,	O
because	O
parity	B-Error_Name
must	O
still	O
be	O
written	O
,	O
this	O
is	O
no	O
longer	O
a	O
bottleneck	O
.	O
</s>
<s>
Since	O
parity	B-Error_Name
calculation	O
is	O
performed	O
on	O
the	O
full	O
stripe	O
,	O
small	O
changes	O
to	O
the	O
array	O
experience	O
write	O
amplification	O
:	O
in	O
the	O
worst	O
case	O
when	O
a	O
single	O
,	O
logical	O
sector	O
is	O
to	O
be	O
written	O
,	O
the	O
original	O
sector	O
and	O
the	O
according	O
parity	B-Error_Name
sector	O
need	O
to	O
be	O
read	O
,	O
the	O
original	O
data	O
is	O
removed	O
from	O
the	O
parity	B-Error_Name
,	O
the	O
new	O
data	O
calculated	O
into	O
the	O
parity	B-Error_Name
and	O
both	O
the	O
new	O
data	O
sector	O
and	O
the	O
new	O
parity	B-Error_Name
sector	O
are	O
written	O
.	O
</s>
<s>
RAID	B-Architecture
6	O
extends	O
RAID5	O
by	O
adding	O
another	O
parity	B-Error_Name
block	I-Error_Name
;	O
thus	O
,	O
it	O
uses	O
block-level	O
striping	B-General_Concept
with	O
two	O
parity	B-Error_Name
blocks	I-Error_Name
distributed	O
across	O
all	O
member	O
disks	O
.	O
</s>
<s>
As	O
in	O
RAID	B-Architecture
5	O
,	O
there	O
are	O
many	O
layouts	O
of	O
RAID	B-Architecture
6	O
disk	O
arrays	O
depending	O
upon	O
the	O
direction	O
the	O
data	O
blocks	O
are	O
written	O
,	O
the	O
location	O
of	O
the	O
parity	B-Error_Name
blocks	I-Error_Name
with	O
respect	O
to	O
the	O
data	O
blocks	O
and	O
whether	O
or	O
not	O
the	O
first	O
data	O
block	B-General_Concept
of	O
a	O
subsequent	O
stripe	O
is	O
written	O
to	O
the	O
same	O
drive	O
as	O
the	O
last	O
parity	B-Error_Name
block	I-Error_Name
of	O
the	O
prior	O
stripe	O
.	O
</s>
<s>
According	O
to	O
the	O
Storage	O
Networking	O
Industry	O
Association	O
(	O
SNIA	O
)	O
,	O
the	O
definition	O
of	O
RAID6	O
is	O
:	O
"	O
Any	O
form	O
of	O
RAID	B-Architecture
that	O
can	O
continue	O
to	O
execute	O
read	O
and	O
write	O
requests	O
to	O
all	O
of	O
a	O
RAID	B-Architecture
array	I-Architecture
's	O
virtual	O
disks	O
in	O
the	O
presence	O
of	O
any	O
two	O
concurrent	O
disk	O
failures	O
.	O
</s>
<s>
Several	O
methods	O
,	O
including	O
dual	O
check	O
data	O
computations	O
(	O
parity	B-Error_Name
and	O
Reed	B-Error_Name
–	I-Error_Name
Solomon	I-Error_Name
)	O
,	O
orthogonal	O
dual	O
parity	B-Error_Name
check	I-Error_Name
data	O
and	O
diagonal	O
parity	B-Error_Name
,	O
have	O
been	O
used	O
to	O
implement	O
RAID	B-Architecture
Level	O
6.	O
"	O
</s>
<s>
RAID6	O
does	O
not	O
have	O
a	O
performance	O
penalty	O
for	O
read	O
operations	O
,	O
but	O
it	O
does	O
have	O
a	O
performance	O
penalty	O
on	O
write	O
operations	O
because	O
of	O
the	O
overhead	O
associated	O
with	O
parity	B-Error_Name
calculations	O
.	O
</s>
<s>
Performance	O
varies	O
greatly	O
depending	O
on	O
how	O
RAID6	O
is	O
implemented	O
in	O
the	O
manufacturer	O
's	O
storage	O
architecture	O
—	O
in	O
software	O
,	O
firmware	O
,	O
or	O
by	O
using	O
firmware	O
and	O
specialized	O
ASICs	O
for	O
intensive	O
parity	B-Error_Name
calculations	O
.	O
</s>
<s>
RAID	B-Architecture
6	O
can	O
read	O
up	O
to	O
the	O
same	O
speed	O
as	O
RAID	B-Architecture
5	O
with	O
the	O
same	O
number	O
of	O
physical	O
drives	O
.	O
</s>
<s>
When	O
either	O
diagonal	O
or	O
orthogonal	O
dual	O
parity	B-Error_Name
is	O
used	O
,	O
a	O
second	O
parity	B-Error_Name
calculation	O
is	O
necessary	O
for	O
write	O
operations	O
.	O
</s>
<s>
This	O
doubles	O
CPU	O
overhead	O
for	O
RAID-6	O
writes	O
,	O
versus	O
single-parity	O
RAID	B-General_Concept
levels	I-General_Concept
.	O
</s>
<s>
When	O
a	O
Reed	B-Error_Name
Solomon	I-Error_Name
code	O
is	O
used	O
,	O
the	O
second	O
parity	B-Error_Name
calculation	O
is	O
unnecessary	O
.	O
</s>
<s>
Reed	B-Error_Name
Solomon	I-Error_Name
has	O
the	O
advantage	O
of	O
allowing	O
all	O
redundancy	O
information	O
to	O
be	O
contained	O
within	O
a	O
given	O
stripe	O
.	O
</s>
<s>
It	O
is	O
possible	O
to	O
support	O
a	O
far	O
greater	O
number	O
of	O
drives	O
by	O
choosing	O
the	O
parity	B-Error_Name
function	O
more	O
carefully	O
.	O
</s>
<s>
Let	O
correspond	O
to	O
the	O
stripes	B-General_Concept
of	O
data	O
across	O
hard	B-Device
drives	I-Device
encoded	O
as	O
field	O
elements	O
in	O
this	O
manner	O
.	O
</s>
<s>
The	O
effect	O
of	O
can	O
be	O
thought	O
of	O
as	O
the	O
action	O
of	O
a	O
carefully	O
chosen	O
linear	B-Architecture
feedback	I-Architecture
shift	I-Architecture
register	I-Architecture
on	O
the	O
data	O
chunk	O
.	O
</s>
<s>
This	O
can	O
be	O
mitigated	O
with	O
a	O
hardware	O
implementation	O
or	O
by	O
using	O
an	O
FPGA	B-Architecture
.	O
</s>
<s>
The	O
following	O
table	O
provides	O
an	O
overview	O
of	O
some	O
considerations	O
for	O
standard	B-General_Concept
RAID	I-General_Concept
levels	I-General_Concept
.	O
</s>
<s>
Different	O
RAID	B-Architecture
configurations	O
can	O
also	O
detect	O
failure	O
during	O
so	O
called	O
data	B-Error_Name
scrubbing	I-Error_Name
.	O
</s>
<s>
Historically	O
disks	O
were	O
subject	O
to	O
lower	O
reliability	O
and	O
RAID	B-General_Concept
levels	I-General_Concept
were	O
also	O
used	O
to	O
detect	O
which	O
disk	O
in	O
the	O
array	O
had	O
failed	O
in	O
addition	O
to	O
that	O
a	O
disk	O
had	O
failed	O
.	O
</s>
<s>
even	O
at	O
the	O
inception	O
of	O
RAID	B-Architecture
many	O
(	O
though	O
not	O
all	O
)	O
disks	O
were	O
already	O
capable	O
of	O
finding	O
internal	O
errors	O
using	O
error	O
correcting	O
codes	O
.	O
</s>
<s>
Modern	O
RAID	B-Architecture
arrays	I-Architecture
depend	O
for	O
the	O
most	O
part	O
on	O
a	O
disk	O
's	O
ability	O
to	O
identify	O
itself	O
as	O
faulty	O
which	O
can	O
be	O
detected	O
as	O
part	O
of	O
a	O
scrub	O
.	O
</s>
<s>
Enterprise	O
drives	O
may	O
also	O
report	O
failure	O
in	O
far	O
fewer	O
tries	O
than	O
consumer	O
drives	O
as	O
part	O
of	O
TLER	B-Device
to	O
ensure	O
a	O
read	O
request	O
is	O
fulfilled	O
in	O
a	O
timely	O
manner	O
.	O
</s>
<s>
Level	O
Description	O
Minimum	O
number	O
of	O
drives	O
Space	O
efficiency	O
Fault	B-General_Concept
tolerance	I-General_Concept
Fault	O
isolation	O
Read	O
performance	O
Write	O
performance	O
as	O
factor	O
of	O
single	O
disk	O
RAID	B-Architecture
0	O
Block-level	O
striping	B-General_Concept
without	O
parity	B-Error_Name
or	O
mirroring	B-General_Concept
2	O
None	O
Drive	O
Firmware	O
Only	O
RAID	B-Architecture
1	O
Mirroring	B-General_Concept
without	O
parity	B-Error_Name
or	O
striping	B-General_Concept
2	O
drive	O
failures	O
Drive	O
Firmware	O
or	O
voting	O
if	O
RAID	B-Architecture
2	O
Bit-level	O
striping	B-General_Concept
with	O
Hamming	B-Error_Name
code	I-Error_Name
for	O
error	B-Error_Name
correction	I-Error_Name
3	O
One	O
drive	O
failure	O
Drive	O
Firmware	O
and	O
Parity	B-Error_Name
RAID	B-Architecture
3	O
Byte-level	O
striping	B-General_Concept
with	O
dedicated	O
parity	B-Error_Name
3	O
One	O
drive	O
failure	O
Drive	O
Firmware	O
and	O
Parity	B-Error_Name
RAID	B-Architecture
4	O
Block-level	O
striping	B-General_Concept
with	O
dedicated	O
parity	B-Error_Name
3	O
One	O
drive	O
failure	O
Drive	O
Firmware	O
and	O
Parity	B-Error_Name
RAID	B-Architecture
5	O
Block-level	O
striping	B-General_Concept
with	O
distributed	O
parity	B-Error_Name
3	O
One	O
drive	O
failure	O
Drive	O
Firmware	O
and	O
Parity	B-Error_Name
single	O
sector	O
:	O
full	O
stripe	O
:	O
RAID	B-Architecture
6	O
Block-level	O
striping	B-General_Concept
with	O
double	O
distributed	O
parity	B-Error_Name
4	O
Two	O
drive	O
failures	O
Drive	O
Firmware	O
and	O
Parity	B-Error_Name
single	O
sector	O
:	O
full	O
stripe	O
:	O
</s>
<s>
In	O
measurement	O
of	O
the	O
I/O	B-General_Concept
performance	O
of	O
five	O
filesystems	O
with	O
five	O
storage	O
configurations	O
—	O
single	O
SSD	O
,	O
RAID	B-Architecture
0	O
,	O
RAID	B-Architecture
1	O
,	O
RAID	B-Architecture
10	O
,	O
and	O
RAID	B-Architecture
5	O
it	O
was	O
shown	O
that	O
F2FS	B-Application
on	O
RAID	B-Architecture
0	O
and	O
RAID	B-Architecture
5	O
with	O
eight	O
SSDs	O
outperforms	O
EXT4	B-Application
by	O
5	O
times	O
and	O
50	O
times	O
,	O
respectively	O
.	O
</s>
<s>
The	O
measurements	O
also	O
suggest	O
that	O
the	O
RAID	B-Architecture
controller	O
can	O
be	O
a	O
significant	O
bottleneck	O
in	O
building	O
a	O
RAID	B-Architecture
system	O
with	O
high	O
speed	O
SSDs	O
.	O
</s>
<s>
Combinations	O
of	O
two	O
or	O
more	O
standard	B-General_Concept
RAID	I-General_Concept
levels	I-General_Concept
.	O
</s>
<s>
They	O
are	O
also	O
known	O
as	O
RAID	B-Architecture
0+1	O
or	O
RAID	B-Architecture
01	O
,	O
RAID	B-Architecture
0+3	O
or	O
RAID	B-Architecture
03	O
,	O
RAID	B-Architecture
1+0	O
or	O
RAID	B-Architecture
10	O
,	O
RAID	B-Architecture
5+0	O
or	O
RAID	B-Architecture
50	O
,	O
RAID	B-Architecture
6+0	O
or	O
RAID	B-Architecture
60	O
,	O
and	O
RAID	B-Architecture
10+0	O
or	O
RAID	B-Architecture
100	O
.	O
</s>
<s>
In	O
addition	O
to	O
standard	O
and	O
nested	O
RAID	B-General_Concept
levels	I-General_Concept
,	O
alternatives	O
include	O
non-standard	B-General_Concept
RAID	I-General_Concept
levels	I-General_Concept
,	O
and	O
non-RAID	B-General_Concept
drive	I-General_Concept
architectures	I-General_Concept
.	O
</s>
<s>
Non-RAID	B-General_Concept
drive	I-General_Concept
architectures	I-General_Concept
are	O
referred	O
to	O
by	O
similar	O
terms	O
and	O
acronyms	O
,	O
notably	O
JBOD	O
(	O
"	O
just	O
a	O
bunch	O
of	O
disks	O
"	O
)	O
,	O
SPAN/BIG	O
,	O
and	O
MAID	O
(	O
"	O
massive	O
array	O
of	O
idle	O
disks	O
"	O
)	O
.	O
</s>
