<s>
A	O
clustered	B-Application
file	I-Application
system	I-Application
is	O
a	O
file	B-Application
system	I-Application
which	O
is	O
shared	O
by	O
being	O
simultaneously	O
mounted	B-Application
on	O
multiple	O
servers	B-Application
.	O
</s>
<s>
There	O
are	O
several	O
approaches	O
to	O
clustering	B-Architecture
,	O
most	O
of	O
which	O
do	O
not	O
employ	O
a	O
clustered	B-Application
file	I-Application
system	I-Application
(	O
only	O
direct	B-General_Concept
attached	I-General_Concept
storage	I-General_Concept
for	O
each	O
node	O
)	O
.	O
</s>
<s>
Clustered	B-Application
file	I-Application
systems	I-Application
can	O
provide	O
features	O
like	O
location-independent	O
addressing	O
and	O
redundancy	O
which	O
improve	O
reliability	O
or	O
reduce	O
the	O
complexity	O
of	O
the	O
other	O
parts	O
of	O
the	O
cluster	O
.	O
</s>
<s>
Parallel	B-Application
file	I-Application
systems	I-Application
are	O
a	O
type	O
of	O
clustered	B-Application
file	I-Application
system	I-Application
that	O
spread	O
data	O
across	O
multiple	O
storage	O
nodes	O
,	O
usually	O
for	O
redundancy	O
or	O
performance	O
.	O
</s>
<s>
A	O
shared-disk	B-General_Concept
file	B-Application
system	I-Application
uses	O
a	O
storage	O
area	O
network	O
(	O
SAN	O
)	O
to	O
allow	O
multiple	O
computers	O
to	O
gain	O
direct	O
disk	O
access	O
at	O
the	O
block	B-General_Concept
level	I-General_Concept
.	O
</s>
<s>
The	O
most	O
common	O
type	O
of	O
clustered	B-Application
file	I-Application
system	I-Application
,	O
the	O
shared-disk	B-General_Concept
file	B-Application
system	I-Application
by	O
adding	O
mechanisms	O
for	O
concurrency	O
controlprovides	O
a	O
consistent	O
and	O
serializable	B-Application
view	O
of	O
the	O
file	B-Application
system	I-Application
,	O
avoiding	O
corruption	O
and	O
unintended	O
data	B-General_Concept
loss	I-General_Concept
even	O
when	O
multiple	O
clients	O
try	O
to	O
access	O
the	O
same	O
files	O
at	O
the	O
same	O
time	O
.	O
</s>
<s>
Shared-disk	B-General_Concept
file-systems	B-Application
commonly	O
employ	O
some	O
sort	O
of	O
fencing	B-Operating_System
mechanism	O
to	O
prevent	O
data	O
corruption	O
in	O
case	O
of	O
node	O
failures	O
,	O
because	O
an	O
unfenced	O
device	O
can	O
cause	O
data	O
corruption	O
if	O
it	O
loses	O
communication	O
with	O
its	O
sister	O
nodes	O
and	O
tries	O
to	O
access	O
the	O
same	O
information	O
other	O
nodes	O
are	O
accessing	O
.	O
</s>
<s>
The	O
underlying	O
storage	O
area	O
network	O
may	O
use	O
any	O
of	O
a	O
number	O
of	O
block-level	O
protocols	O
,	O
including	O
SCSI	B-Architecture
,	O
iSCSI	B-Protocol
,	O
HyperSCSI	B-Protocol
,	O
ATA	B-Protocol
over	I-Protocol
Ethernet	I-Protocol
(	O
AoE	O
)	O
,	O
Fibre	B-Architecture
Channel	I-Architecture
,	O
network	B-Application
block	I-Application
device	I-Application
,	O
and	O
InfiniBand	B-Architecture
.	O
</s>
<s>
There	O
are	O
different	O
architectural	O
approaches	O
to	O
a	O
shared-disk	B-General_Concept
filesystem	B-Application
.	O
</s>
<s>
Some	O
distribute	O
file	O
information	O
across	O
all	O
the	O
servers	B-Application
in	O
a	O
cluster	O
(	O
fully	O
distributed	O
)	O
.	O
</s>
<s>
Distributed	B-Protocol
file	I-Protocol
systems	I-Protocol
do	O
not	O
share	O
block	B-General_Concept
level	I-General_Concept
access	I-General_Concept
to	O
the	O
same	O
storage	O
but	O
use	O
a	O
network	O
protocol	O
.	O
</s>
<s>
These	O
are	O
commonly	O
known	O
as	O
network	B-Protocol
file	I-Protocol
systems	I-Protocol
,	O
even	O
though	O
they	O
are	O
not	O
the	O
only	O
file	B-Application
systems	I-Application
that	O
use	O
the	O
network	O
to	O
send	O
data	O
.	O
</s>
<s>
Distributed	B-Protocol
file	I-Protocol
systems	I-Protocol
can	O
restrict	O
access	O
to	O
the	O
file	B-Application
system	I-Application
depending	O
on	O
access	O
lists	O
or	O
capabilities	O
on	O
both	O
the	O
servers	B-Application
and	O
the	O
clients	O
,	O
depending	O
on	O
how	O
the	O
protocol	O
is	O
designed	O
.	O
</s>
<s>
The	O
difference	O
between	O
a	O
distributed	B-Protocol
file	I-Protocol
system	I-Protocol
and	O
a	O
distributed	O
data	O
store	O
is	O
that	O
a	O
distributed	B-Protocol
file	I-Protocol
system	I-Protocol
allows	O
files	O
to	O
be	O
accessed	O
using	O
the	O
same	O
interfaces	O
and	O
semantics	O
as	O
local	O
files	O
for	O
example	O
,	O
mounting/unmounting	O
,	O
listing	O
directories	O
,	O
read/write	O
at	O
byte	O
boundaries	O
,	O
system	O
's	O
native	O
permission	O
model	O
.	O
</s>
<s>
Distributed	B-Protocol
file	I-Protocol
systems	I-Protocol
may	O
aim	O
for	O
"	O
transparency	O
"	O
in	O
a	O
number	O
of	O
aspects	O
.	O
</s>
<s>
That	O
is	O
,	O
they	O
aim	O
to	O
be	O
"	O
invisible	O
"	O
to	O
client	O
programs	O
,	O
which	O
"	O
see	O
"	O
a	O
system	O
which	O
is	O
similar	O
to	O
a	O
local	O
file	B-Application
system	I-Application
.	O
</s>
<s>
Behind	O
the	O
scenes	O
,	O
the	O
distributed	B-Protocol
file	I-Protocol
system	I-Protocol
handles	O
locating	O
files	O
,	O
transporting	O
data	O
,	O
and	O
potentially	O
providing	O
other	O
features	O
listed	O
below	O
.	O
</s>
<s>
Concurrency	O
transparency	O
:	O
all	O
clients	O
have	O
the	O
same	O
view	O
of	O
the	O
state	O
of	O
the	O
file	B-Application
system	I-Application
.	O
</s>
<s>
Scalability	O
:	O
the	O
file	B-Application
system	I-Application
should	O
work	O
well	O
in	O
small	O
environments	O
(	O
1	O
machine	O
,	O
a	O
dozen	O
machines	O
)	O
and	O
also	O
scale	O
gracefully	O
to	O
bigger	O
ones	O
(	O
hundreds	O
through	O
tens	O
of	O
thousands	O
of	O
systems	O
)	O
.	O
</s>
<s>
Replication	B-Architecture
transparency	O
:	O
Clients	O
should	O
not	O
have	O
to	O
be	O
aware	O
of	O
the	O
file	O
replication	B-Architecture
performed	O
across	O
multiple	O
servers	B-Application
to	O
support	O
scalability	O
.	O
</s>
<s>
Migration	O
transparency	O
:	O
files	O
should	O
be	O
able	O
to	O
move	O
between	O
different	O
servers	B-Application
without	O
the	O
client	O
's	O
knowledge	O
.	O
</s>
<s>
The	O
Incompatible	B-Application
Timesharing	I-Application
System	I-Application
used	O
virtual	O
devices	O
for	O
transparent	O
inter-machine	O
file	B-Application
system	I-Application
access	O
in	O
the	O
1960s	O
.	O
</s>
<s>
More	O
file	O
servers	B-Application
were	O
developed	O
in	O
the	O
1970s	O
.	O
</s>
<s>
In	O
1976	O
Digital	O
Equipment	O
Corporation	O
created	O
the	O
File	B-Protocol
Access	I-Protocol
Listener	I-Protocol
(	O
FAL	O
)	O
,	O
an	O
implementation	O
of	O
the	O
Data	O
Access	O
Protocol	O
as	O
part	O
of	O
DECnet	B-Protocol
Phase	O
II	O
which	O
became	O
the	O
first	O
widely	O
used	O
network	B-Protocol
file	I-Protocol
system	I-Protocol
.	O
</s>
<s>
In	O
1985	O
Sun	O
Microsystems	O
created	O
the	O
file	B-Application
system	I-Application
called	O
"	O
Network	B-Protocol
File	I-Protocol
System	I-Protocol
"	O
(	O
NFS	B-Protocol
)	O
which	O
became	O
the	O
first	O
widely	O
used	O
Internet	B-Protocol
Protocol	I-Protocol
based	O
network	B-Protocol
file	I-Protocol
system	I-Protocol
.	O
</s>
<s>
Other	O
notable	O
network	B-Protocol
file	I-Protocol
systems	I-Protocol
are	O
Andrew	B-Application
File	I-Application
System	I-Application
(	O
AFS	B-Application
)	O
,	O
Apple	B-Protocol
Filing	I-Protocol
Protocol	I-Protocol
(	O
AFP	B-Protocol
)	O
,	O
NetWare	B-Protocol
Core	I-Protocol
Protocol	I-Protocol
(	O
NCP	O
)	O
,	O
and	O
Server	B-Operating_System
Message	I-Operating_System
Block	I-Operating_System
(	O
SMB	O
)	O
which	O
is	O
also	O
known	O
as	O
Common	B-Operating_System
Internet	I-Operating_System
File	I-Operating_System
System	I-Operating_System
(	O
CIFS	B-Operating_System
)	O
.	O
</s>
<s>
In	O
1986	O
,	O
IBM	O
announced	O
client	O
and	O
server	O
support	O
for	O
Distributed	O
Data	O
Management	O
Architecture	O
(	O
DDM	O
)	O
for	O
the	O
System/36	B-Device
,	O
System/38	B-Device
,	O
and	O
IBM	O
mainframe	O
computers	O
running	O
CICS	B-Application
.	O
</s>
<s>
This	O
was	O
followed	O
by	O
the	O
support	O
for	O
IBM	B-Device
Personal	I-Device
Computer	I-Device
,	O
AS/400	B-Device
,	O
IBM	O
mainframe	O
computers	O
under	O
the	O
MVS	B-Application
and	O
VSE	B-Application
operating	O
systems	O
,	O
and	O
FlexOS	B-Operating_System
.	O
</s>
<s>
DDM	O
also	O
became	O
the	O
foundation	O
for	O
Distributed	B-Language
Relational	I-Language
Database	I-Language
Architecture	I-Language
,	O
also	O
known	O
as	O
DRDA	B-Language
.	O
</s>
<s>
There	O
are	O
many	O
peer-to-peer	O
network	O
protocols	O
for	O
open-source	O
distributed	B-Protocol
file	I-Protocol
systems	I-Protocol
for	O
cloud	O
or	O
closed-source	O
clustered	B-Application
file	I-Application
systems	I-Application
,	O
e	O
.	O
g.	O
:	O
9P	B-Operating_System
,	O
AFS	B-Application
,	O
Coda	B-Application
,	O
CIFS/SMB	B-Operating_System
,	O
DCE/DFS	B-Protocol
,	O
,	O
Lustre	B-Application
,	O
,	O
Google	B-Application
File	I-Application
System	I-Application
,	O
Mnet	O
,	O
Chord	B-Application
Project	I-Application
.	O
</s>
<s>
GFS	B-Application
(	O
Google	O
Inc	O
.	O
)	O
</s>
<s>
Network-attached	O
storage	O
(	O
NAS	O
)	O
provides	O
both	O
storage	O
and	O
a	O
file	B-Application
system	I-Application
,	O
like	O
a	O
shared	B-General_Concept
disk	I-General_Concept
file	B-Application
system	I-Application
on	O
top	O
of	O
a	O
storage	O
area	O
network	O
(	O
SAN	O
)	O
.	O
</s>
<s>
NAS	O
typically	O
uses	O
file-based	O
protocols	O
(	O
as	O
opposed	O
to	O
block-based	O
protocols	O
a	O
SAN	O
would	O
use	O
)	O
such	O
as	O
NFS	B-Protocol
(	O
popular	O
on	O
UNIX	B-Application
systems	I-Application
)	O
,	O
SMB/CIFS	B-Operating_System
(	O
Server	B-Operating_System
Message	I-Operating_System
Block/Common	I-Operating_System
Internet	I-Operating_System
File	I-Operating_System
System	I-Operating_System
)	O
(	O
used	O
with	O
MS	O
Windows	O
systems	O
)	O
,	O
AFP	B-Protocol
(	O
used	O
with	O
Apple	B-Device
Macintosh	I-Device
computers	O
)	O
,	O
or	O
NCP	O
(	O
used	O
with	O
OES	B-Application
and	O
Novell	B-Operating_System
NetWare	I-Operating_System
)	O
.	O
</s>
<s>
The	O
failure	O
of	O
disk	O
hardware	O
or	O
a	O
given	O
storage	O
node	O
in	O
a	O
cluster	O
can	O
create	O
a	O
single	B-Architecture
point	I-Architecture
of	I-Architecture
failure	I-Architecture
that	O
can	O
result	O
in	O
data	B-General_Concept
loss	I-General_Concept
or	O
unavailability	O
.	O
</s>
<s>
Fault	B-General_Concept
tolerance	I-General_Concept
and	O
high	O
availability	O
can	O
be	O
provided	O
through	O
data	B-Architecture
replication	I-Architecture
of	O
one	O
sort	O
or	O
another	O
,	O
so	O
that	O
data	O
remains	O
intact	O
and	O
available	O
despite	O
the	O
failure	O
of	O
any	O
single	O
piece	O
of	O
equipment	O
.	O
</s>
<s>
For	O
examples	O
,	O
see	O
the	O
lists	O
of	O
distributed	O
fault-tolerant	B-General_Concept
file	B-Application
systems	I-Application
and	O
distributed	O
parallel	O
fault-tolerant	B-General_Concept
file	B-Application
systems	I-Application
.	O
</s>
<s>
A	O
common	O
performance	O
measurement	O
of	O
a	O
clustered	B-Application
file	I-Application
system	I-Application
is	O
the	O
amount	O
of	O
time	O
needed	O
to	O
satisfy	O
service	O
requests	O
.	O
</s>
<s>
In	O
conventional	O
systems	O
,	O
this	O
time	O
consists	O
of	O
a	O
disk-access	O
time	O
and	O
a	O
small	O
amount	O
of	O
CPU-processing	O
time	O
.	O
</s>
<s>
But	O
in	O
a	O
clustered	B-Application
file	I-Application
system	I-Application
,	O
a	O
remote	O
access	O
has	O
additional	O
overhead	O
due	O
to	O
the	O
distributed	O
structure	O
.	O
</s>
<s>
This	O
includes	O
the	O
time	O
to	O
deliver	O
the	O
request	O
to	O
a	O
server	O
,	O
the	O
time	O
to	O
deliver	O
the	O
response	O
to	O
the	O
client	O
,	O
and	O
for	O
each	O
direction	O
,	O
a	O
CPU	B-General_Concept
overhead	O
of	O
running	O
the	O
communication	O
protocol	O
software	O
.	O
</s>
<s>
Concurrency	B-Operating_System
control	I-Operating_System
becomes	O
an	O
issue	O
when	O
more	O
than	O
one	O
person	O
or	O
client	O
is	O
accessing	O
the	O
same	O
file	O
or	O
block	O
and	O
want	O
to	O
update	O
it	O
.	O
</s>
<s>
This	O
problem	O
is	O
more	O
complex	O
with	O
file	B-Application
systems	I-Application
due	O
to	O
concurrent	O
overlapping	O
writes	O
,	O
where	O
different	O
writers	O
write	O
to	O
overlapping	O
regions	O
of	O
the	O
file	O
concurrently	O
.	O
</s>
<s>
This	O
problem	O
is	O
usually	O
handled	O
by	O
concurrency	B-Operating_System
control	I-Operating_System
or	O
locking	B-Operating_System
which	O
may	O
either	O
be	O
built	O
into	O
the	O
file	B-Application
system	I-Application
or	O
provided	O
by	O
an	O
add-on	O
protocol	O
.	O
</s>
<s>
IBM	O
mainframes	O
in	O
the	O
1970s	O
could	O
share	O
physical	O
disks	O
and	O
file	B-Application
systems	I-Application
if	O
each	O
machine	O
had	O
its	O
own	O
channel	O
connection	O
to	O
the	O
drives	O
 '	O
control	O
units	O
.	O
</s>
<s>
In	O
the	O
1980s	O
,	O
Digital	O
Equipment	O
Corporation	O
's	O
TOPS-20	B-Operating_System
and	O
OpenVMS	B-Operating_System
clusters	O
(	O
VAX/ALPHA/IA64	O
)	O
included	O
shared	B-General_Concept
disk	I-General_Concept
file	B-Application
systems	I-Application
.	O
</s>
