<s>
A	O
diskless	B-Device
node	I-Device
(	O
or	O
diskless	B-Device
workstation	I-Device
)	O
is	O
a	O
workstation	B-Device
or	O
personal	B-Device
computer	I-Device
without	O
disk	B-Device
drives	I-Device
,	O
which	O
employs	O
network	B-Device
booting	I-Device
to	O
load	O
its	O
operating	B-General_Concept
system	I-General_Concept
from	O
a	O
server	B-Application
.	O
</s>
<s>
(	O
A	O
computer	O
may	O
also	O
be	O
said	O
to	O
act	O
as	O
a	O
diskless	B-Device
node	I-Device
,	O
if	O
its	O
disks	B-Device
are	O
unused	O
and	O
network	B-Device
booting	I-Device
is	O
used	O
.	O
)	O
</s>
<s>
Diskless	B-Device
nodes	I-Device
(	O
or	O
computers	O
acting	O
as	O
such	O
)	O
are	O
sometimes	O
known	O
as	O
network	B-Device
computers	I-Device
or	O
hybrid	B-Device
clients	I-Device
.	O
</s>
<s>
Hybrid	B-Device
client	I-Device
may	O
either	O
just	O
mean	O
diskless	B-Device
node	I-Device
,	O
or	O
it	O
may	O
be	O
used	O
in	O
a	O
more	O
particular	O
sense	O
to	O
mean	O
a	O
diskless	B-Device
node	I-Device
which	O
runs	O
some	O
,	O
but	O
not	O
all	O
,	O
applications	B-Application
remotely	O
,	O
as	O
in	O
the	O
thin	B-Device
client	I-Device
computing	O
architecture	O
.	O
</s>
<s>
Advantages	O
of	O
diskless	B-Device
nodes	I-Device
can	O
include	O
lower	O
production	O
cost	O
,	O
lower	O
running	O
costs	O
,	O
quieter	O
operation	O
,	O
and	O
manageability	O
advantages	O
(	O
for	O
example	O
,	O
centrally	O
managed	O
software	B-Application
installation	O
)	O
.	O
</s>
<s>
In	O
many	O
universities	O
and	O
in	O
some	O
large	O
organizations	O
,	O
PCs	B-Device
are	O
used	O
in	O
a	O
similar	O
configuration	O
,	O
with	O
some	O
or	O
all	O
applications	B-Application
stored	O
remotely	O
but	O
executed	B-General_Concept
locally	O
—	O
again	O
,	O
for	O
manageability	O
reasons	O
.	O
</s>
<s>
However	O
,	O
these	O
are	O
not	O
diskless	B-Device
nodes	I-Device
if	O
they	O
still	O
boot	B-Operating_System
from	O
a	O
local	O
hard	B-Device
drive	I-Device
.	O
</s>
<s>
Diskless	B-Device
nodes	I-Device
process	O
data	B-General_Concept
,	O
thus	O
using	O
their	O
own	O
CPU	B-General_Concept
and	O
RAM	B-Architecture
to	O
run	O
software	B-Application
,	O
but	O
do	O
not	O
store	O
data	B-General_Concept
persistently	O
—	O
that	O
task	O
is	O
handed	O
off	O
to	O
a	O
server	B-Application
.	O
</s>
<s>
This	O
is	O
distinct	O
from	O
thin	B-Device
clients	I-Device
,	O
in	O
which	O
all	O
significant	O
processing	O
happens	O
remotely	O
,	O
on	O
the	O
server	B-Application
—	O
the	O
only	O
software	B-Application
that	O
runs	O
on	O
a	O
thin	B-Device
client	I-Device
is	O
the	O
"	O
thin	O
"	O
(	O
i.e.	O
</s>
<s>
relatively	O
small	O
and	O
simple	O
)	O
client	O
software	B-Application
,	O
which	O
handles	O
simple	O
input/output	O
tasks	O
to	O
communicate	O
with	O
the	O
user	O
,	O
such	O
as	O
drawing	O
a	O
dialog	O
box	O
on	O
the	O
display	B-Device
or	O
waiting	O
for	O
user	O
input	O
.	O
</s>
<s>
A	O
collective	O
term	O
encompassing	O
both	O
thin	B-Device
client	I-Device
computing	O
,	O
and	O
its	O
technological	O
predecessor	O
,	O
text	B-General_Concept
terminals	I-General_Concept
(	O
which	O
are	O
text-only	O
)	O
,	O
is	O
centralized	B-General_Concept
computing	I-General_Concept
.	O
</s>
<s>
Thin	B-Device
clients	I-Device
and	O
text	B-General_Concept
terminals	I-General_Concept
can	O
both	O
require	O
powerful	O
central	O
processing	O
facilities	O
in	O
the	O
servers	O
,	O
in	O
order	O
to	O
perform	O
all	O
significant	O
processing	O
tasks	O
for	O
all	O
of	O
the	O
clients	O
.	O
</s>
<s>
Diskless	B-Device
nodes	I-Device
can	O
be	O
seen	O
as	O
a	O
compromise	O
between	O
fat	B-Protocol
clients	I-Protocol
(	O
such	O
as	O
ordinary	O
personal	B-Device
computers	I-Device
)	O
and	O
centralized	B-General_Concept
computing	I-General_Concept
,	O
using	O
central	O
storage	O
for	O
efficiency	O
,	O
but	O
not	O
requiring	O
centralized	O
processing	O
,	O
and	O
making	O
efficient	O
use	O
of	O
the	O
powerful	O
processing	O
power	O
of	O
even	O
the	O
slowest	O
of	O
contemporary	O
CPUs	O
,	O
which	O
would	O
tend	O
to	O
sit	O
idle	O
for	O
much	O
of	O
the	O
time	O
under	O
the	O
centralized	B-General_Concept
computing	I-General_Concept
model	O
.	O
</s>
<s>
The	O
operating	B-General_Concept
system	I-General_Concept
(	O
OS	O
)	O
for	O
a	O
diskless	B-Device
node	I-Device
is	O
loaded	O
from	O
a	O
server	B-Application
,	O
using	O
network	B-Device
booting	I-Device
.	O
</s>
<s>
In	O
some	O
cases	O
,	O
removable	O
storage	O
may	O
be	O
used	O
to	O
initiate	O
the	O
bootstrap	O
process	O
,	O
such	O
as	O
a	O
USB	B-Protocol
flash	O
drive	O
,	O
or	O
other	O
bootable	B-Operating_System
media	O
such	O
as	O
a	O
floppy	B-Device
disk	I-Device
,	O
CD	O
or	O
DVD	O
.	O
</s>
<s>
However	O
,	O
the	O
firmware	B-Application
in	O
many	O
modern	O
computers	O
can	O
be	O
configured	O
to	O
locate	O
a	O
server	B-Application
and	O
begin	O
the	O
bootup	B-Operating_System
process	O
automatically	O
,	O
without	O
the	O
need	O
to	O
insert	O
bootable	B-Operating_System
media	O
.	O
</s>
<s>
For	O
network	O
auto-booting	O
,	O
the	O
Preboot	B-Device
Execution	I-Device
Environment	I-Device
(	O
PXE	O
)	O
or	O
Bootstrap	B-Device
Protocol	I-Device
(	O
BOOTP	B-Device
)	O
network	O
protocols	O
are	O
commonly	O
used	O
to	O
find	O
a	O
server	B-Application
with	O
files	O
for	O
booting	B-Operating_System
the	O
device	O
.	O
</s>
<s>
Standard	O
full-size	O
desktop	O
PCs	B-Device
are	O
able	O
to	O
be	O
network-booted	O
in	O
this	O
manner	O
with	O
an	O
add-on	O
network	O
card	O
that	O
includes	O
a	O
UNDI	B-Device
boot	B-Operating_System
ROM	O
.	O
</s>
<s>
Diskless	B-Device
network	B-Device
booting	I-Device
is	O
commonly	O
a	O
built-in	O
feature	O
of	O
desktop	O
and	O
laptop	O
PCs	B-Device
intended	O
for	O
business	O
use	O
,	O
since	O
it	O
can	O
be	O
used	O
on	O
an	O
otherwise	O
disk-booted	O
standard	O
desktop	O
computer	O
to	O
remotely	O
run	O
diagnostics	O
,	O
to	O
install	O
software	B-Application
,	O
or	O
to	O
apply	O
a	O
disk	B-Application
image	I-Application
to	O
the	O
local	O
hard	B-Device
drive	I-Device
.	O
</s>
<s>
In	O
the	O
first	O
approach	O
(	O
used	O
,	O
for	O
example	O
,	O
by	O
the	O
Linux	B-Application
Terminal	I-Application
Server	I-Application
Project	I-Application
)	O
,	O
the	O
kernel	B-Operating_System
is	O
loaded	O
into	O
memory	O
and	O
then	O
the	O
rest	O
of	O
the	O
operating	B-General_Concept
system	I-General_Concept
is	O
accessed	O
via	O
a	O
network	B-Application
filesystem	I-Application
connection	O
to	O
the	O
server	B-Application
.	O
</s>
<s>
(	O
A	O
small	O
RAM	B-Application
disk	I-Application
may	O
be	O
created	O
to	O
store	O
temporary	O
files	O
locally	O
.	O
)	O
</s>
<s>
This	O
approach	O
is	O
sometimes	O
called	O
the	O
"	O
NFS	B-Protocol
root	O
"	O
technique	O
when	O
used	O
with	O
Linux	B-Application
or	O
Unix	O
client	O
operating	B-General_Concept
systems	I-General_Concept
.	O
</s>
<s>
In	O
the	O
second	O
approach	O
,	O
the	O
kernel	B-Operating_System
of	O
the	O
OS	O
is	O
loaded	O
,	O
and	O
part	O
of	O
the	O
system	O
's	O
memory	O
is	O
configured	O
as	O
a	O
large	O
RAM	B-Application
disk	I-Application
,	O
and	O
then	O
the	O
remainder	O
of	O
the	O
OS	O
image	O
is	O
fetched	O
and	O
loaded	O
into	O
the	O
RAM	B-Application
disk	I-Application
.	O
</s>
<s>
This	O
is	O
the	O
implementation	O
that	O
Microsoft	O
has	O
chosen	O
for	O
its	O
Windows	B-Application
XP	O
Embedded	O
remote	O
boot	B-Operating_System
feature	O
.	O
</s>
<s>
In	O
the	O
third	O
approach	O
,	O
disk	B-Device
operations	O
are	O
virtualized	O
and	O
are	O
actually	O
translated	O
into	O
a	O
network	O
protocol	O
.	O
</s>
<s>
The	O
data	B-General_Concept
that	O
is	O
usually	O
stored	O
in	O
a	O
disk	B-Device
drive	I-Device
are	O
then	O
stored	O
in	O
virtual	O
disks	B-Device
files	O
homed	O
on	O
a	O
server	B-Application
.	O
</s>
<s>
The	O
disk	B-Device
operations	O
such	O
as	O
requests	O
to	O
read/write	O
disk	B-Device
sectors	O
are	O
translated	O
into	O
corresponding	O
network	O
requests	O
and	O
processed	O
by	O
a	O
service	O
or	O
daemon	O
running	O
on	O
the	O
server	B-Application
side	O
.	O
</s>
<s>
This	O
is	O
the	O
implementation	O
that	O
is	O
used	O
by	O
Neoware	O
Image	O
Manager	O
,	O
Ardence	O
,	O
VHD	O
Central	O
Management	O
System	O
and	O
various	O
"	O
boot	B-Operating_System
over	O
iSCSI	O
"	O
products	O
.	O
</s>
<s>
This	O
third	O
approach	O
differs	O
from	O
the	O
first	O
approach	O
because	O
what	O
is	O
remote	O
is	O
not	O
a	O
file	B-Application
system	I-Application
but	O
actually	O
a	O
disk	B-Device
device	O
(	O
or	O
raw	B-Application
device	I-Application
)	O
and	O
that	O
the	O
client	O
OS	O
is	O
not	O
aware	O
that	O
it	O
is	O
not	O
running	O
off	O
a	O
hard	B-Device
disk	I-Device
.	O
</s>
<s>
This	O
is	O
why	O
this	O
approach	O
is	O
sometimes	O
named	O
"	O
Virtual	O
Hard	B-Device
Disk	I-Device
"	O
or	O
"	O
Network	O
Virtual	O
Disk	B-Device
"	O
.	O
</s>
<s>
This	O
third	O
approach	O
makes	O
it	O
easier	O
to	O
use	O
client	O
OS	O
than	O
having	O
a	O
complete	O
disk	B-Application
image	I-Application
in	O
RAM	B-Architecture
or	O
using	O
a	O
read-only	O
file	B-Application
system	I-Application
.	O
</s>
<s>
In	O
this	O
approach	O
,	O
the	O
system	O
uses	O
some	O
"	O
write	O
cache	B-General_Concept
"	O
that	O
stores	O
every	O
data	B-General_Concept
that	O
a	O
diskless	B-Device
node	I-Device
has	O
written	O
.	O
</s>
<s>
This	O
write	O
cache	B-General_Concept
is	O
usually	O
a	O
file	O
,	O
stored	O
on	O
a	O
server	B-Application
(	O
or	O
on	O
the	O
client	O
storage	O
if	O
any	O
)	O
.	O
</s>
<s>
It	O
can	O
also	O
be	O
a	O
portion	O
of	O
the	O
client	O
RAM	B-Architecture
.	O
</s>
<s>
This	O
write	O
cache	B-General_Concept
can	O
be	O
persistent	O
or	O
volatile	O
.	O
</s>
<s>
When	O
volatile	O
,	O
all	O
the	O
data	B-General_Concept
that	O
has	O
been	O
written	O
by	O
a	O
specific	O
client	O
to	O
the	O
virtual	O
disk	B-Device
are	O
dismissed	O
when	O
said	O
client	O
is	O
rebooted	O
,	O
and	O
yet	O
,	O
user	O
data	B-General_Concept
can	O
remain	O
persistent	O
if	O
recorded	O
in	O
user	O
(	O
roaming	O
)	O
profiles	O
or	O
home	O
folders	O
(	O
that	O
are	O
stored	O
on	O
remote	O
servers	O
)	O
.	O
</s>
<s>
The	O
two	O
major	O
commercial	O
products	O
(	O
the	O
one	O
from	O
Hewlett-Packard	O
,	O
and	O
the	O
other	O
one	O
from	O
Citrix	B-Application
Systems	I-Application
)	O
that	O
allow	O
the	O
deployment	O
of	O
Diskless	B-Device
Nodes	I-Device
that	O
can	O
boot	B-Operating_System
Microsoft	B-Application
Windows	I-Application
or	O
Linux	B-Application
client	O
OS	O
use	O
such	O
write	O
caches	B-General_Concept
.	O
</s>
<s>
The	O
Citrix	B-Application
product	O
cannot	O
use	O
persistent	O
write	O
cache	B-General_Concept
,	O
but	O
VHD	O
and	O
HP	O
product	O
can	O
.	O
</s>
<s>
Windows	B-Application
3.x	O
and	O
Windows	B-Application
95	O
OSR1	O
supported	O
Remote	O
Boot	B-Operating_System
operations	O
,	O
from	O
NetWare	B-Operating_System
servers	O
,	O
Windows	B-Application
NT	O
Servers	O
and	O
even	O
DEC	O
Pathworks	B-Application
servers	O
.	O
</s>
<s>
Third	O
party	O
software	B-Application
vendors	O
such	O
as	O
Qualystem	O
(	O
acquired	O
by	O
Neoware	O
)	O
,	O
LanWorks	O
(	O
acquired	O
by	O
3Com	O
)	O
,	O
Ardence	O
(	O
acquired	O
by	O
Citrix	B-Application
Systems	I-Application
)	O
,	O
APCT	O
and	O
Xtreamining	O
Technology	O
have	O
developed	O
and	O
marketed	O
software	B-Application
products	I-Application
aimed	O
to	O
remote-boot	O
newer	O
versions	O
of	O
the	O
Windows	B-Application
product	O
line	O
:	O
Windows	B-Application
95	O
OSR2	O
and	O
Windows	B-Application
98	O
were	O
supported	O
by	O
Qualystem	O
and	O
Lanworks	O
,	O
Windows	B-Application
NT	O
was	O
supported	O
by	O
APCT	O
and	O
Ardence	O
(	O
called	O
VenturCom	O
at	O
that	O
time	O
)	O
,	O
and	O
Windows	B-Application
2000/XP/2003/Vista/Windows	O
7	O
are	O
supported	O
by	O
Hewlett-Packard	O
(	O
which	O
acquired	O
Neoware	O
which	O
had	O
previously	O
acquired	O
Qualystem	O
)	O
and	O
Citrix	B-Application
Systems	I-Application
(	O
which	O
acquired	O
Ardence	O
)	O
.	O
</s>
<s>
With	O
essentially	O
a	O
single	O
OS	O
image	O
for	O
an	O
array	O
of	O
machines	O
(	O
with	O
perhaps	O
some	O
customizations	O
for	O
differences	O
in	O
hardware	O
configurations	O
among	O
the	O
nodes	O
)	O
,	O
installing	O
software	B-Application
and	O
maintaining	O
installed	O
software	B-Application
can	O
be	O
more	O
efficient	O
.	O
</s>
<s>
can	O
be	O
either	O
wiped	O
out	O
when	O
the	O
power	O
is	O
removed	O
(	O
if	O
the	O
image	O
is	O
copied	O
to	O
a	O
local	O
RAM	B-Application
disk	I-Application
)	O
such	O
as	O
Windows	B-Application
XP	O
Embedded	O
remote	O
boot	B-Operating_System
or	O
prohibited	O
entirely	O
(	O
if	O
the	O
image	O
is	O
a	O
network	B-Application
filesystem	I-Application
)	O
.	O
</s>
<s>
However	O
,	O
it	O
is	O
not	O
necessary	O
to	O
implement	O
network	B-Device
booting	I-Device
to	O
achieve	O
either	O
of	O
the	O
above	O
advantages	O
-	O
ordinary	O
PCs	B-Device
(	O
with	O
the	O
help	O
of	O
appropriate	O
software	B-Application
)	O
can	O
be	O
configured	O
to	O
download	O
and	O
reinstall	O
their	O
operating	B-General_Concept
systems	I-General_Concept
on	O
(	O
e.g.	O
)	O
</s>
<s>
a	O
nightly	O
basis	O
,	O
with	O
extra	O
work	O
compared	O
to	O
using	O
shared	O
disk	B-Application
image	I-Application
that	O
diskless	B-Device
nodes	I-Device
boot	B-Operating_System
off	O
.	O
</s>
<s>
Modern	O
diskless	B-Device
nodes	I-Device
can	O
share	O
the	O
very	O
same	O
disk	B-Application
image	I-Application
,	O
using	O
a	O
1:N	O
relationship	O
(	O
1	O
disk	B-Application
image	I-Application
used	O
simultaneously	O
by	O
N	O
diskless	B-Device
nodes	I-Device
)	O
.	O
</s>
<s>
This	O
makes	O
it	O
very	O
easy	O
to	O
install	O
and	O
maintain	O
software	B-Application
applications	I-Application
:	O
The	O
administrator	O
needs	O
to	O
install	O
or	O
maintain	O
the	O
application	O
only	O
once	O
,	O
and	O
the	O
clients	O
can	O
get	O
the	O
new	O
application	O
as	O
soon	O
as	O
they	O
boot	B-Operating_System
off	O
the	O
updated	O
image	O
.	O
</s>
<s>
Disk	B-Application
image	I-Application
sharing	O
is	O
made	O
possible	O
because	O
they	O
use	O
the	O
write	O
cache	B-General_Concept
:	O
No	O
client	O
competes	O
for	O
any	O
writing	O
in	O
a	O
shared	O
disk	B-Application
image	I-Application
,	O
because	O
each	O
client	O
writes	O
to	O
its	O
own	O
cache	B-General_Concept
.	O
</s>
<s>
All	O
the	O
modern	O
diskless	B-Device
nodes	I-Device
systems	O
can	O
also	O
use	O
a	O
1:1	O
Client-to-DiskImage	O
relationship	O
,	O
where	O
one	O
client	O
"	O
owns	O
"	O
one	O
disk	B-Application
image	I-Application
and	O
writes	O
directly	O
into	O
said	O
disk	B-Application
image	I-Application
.	O
</s>
<s>
No	O
write	O
cache	B-General_Concept
is	O
used	O
then	O
.	O
</s>
<s>
Making	O
a	O
modification	O
in	O
a	O
shared	O
disk	B-Application
image	I-Application
is	O
usually	O
made	O
this	O
way	O
:	O
</s>
<s>
The	O
diskless	B-Device
nodes	I-Device
use	O
the	O
shared	O
disk	B-Application
image	I-Application
(	O
1:N	O
)	O
as	O
soon	O
as	O
they	O
are	O
rebooted	O
.	O
</s>
<s>
The	O
use	O
of	O
central	O
disk	B-Device
storage	I-Device
also	O
makes	O
more	O
efficient	O
use	O
of	O
disk	B-Device
storage	I-Device
.	O
</s>
<s>
This	O
can	O
cut	O
storage	O
costs	O
,	O
freeing	O
up	O
capital	O
to	O
invest	O
in	O
more	O
reliable	O
,	O
modern	O
storage	O
technologies	O
,	O
such	O
as	O
RAID	B-Architecture
arrays	I-Architecture
which	O
support	O
redundant	O
operation	O
,	O
and	O
storage	O
area	O
networks	O
which	O
allow	O
hot-adding	O
of	O
storage	O
without	O
any	O
interruption	O
.	O
</s>
<s>
Further	O
,	O
it	O
means	O
that	O
losses	O
of	O
disk	B-Device
drives	I-Device
to	O
mechanical	O
or	O
electrical	O
failure	O
—	O
which	O
are	O
statistically	O
highly	O
probable	O
events	O
over	O
a	O
timeframe	O
of	O
years	O
,	O
with	O
a	O
large	O
number	O
of	O
disks	B-Device
involved	O
—	O
are	O
often	O
both	O
less	O
likely	O
to	O
happen	O
(	O
because	O
there	O
are	O
typically	O
fewer	O
disk	B-Device
drives	I-Device
that	O
can	O
fail	O
)	O
and	O
less	O
likely	O
to	O
cause	O
interruption	O
(	O
because	O
they	O
would	O
likely	O
be	O
part	O
of	O
RAID	B-Architecture
arrays	I-Architecture
)	O
.	O
</s>
<s>
This	O
also	O
means	O
that	O
the	O
nodes	O
themselves	O
are	O
less	O
likely	O
to	O
have	O
hardware	O
failures	O
than	O
fat	B-Protocol
clients	I-Protocol
.	O
</s>
<s>
Diskless	B-Device
nodes	I-Device
share	O
these	O
advantages	O
with	O
thin	B-Device
clients	I-Device
.	O
</s>
<s>
Large	O
numbers	O
of	O
nodes	O
making	O
demands	O
on	O
the	O
same	O
server	B-Application
simultaneously	O
can	O
slow	O
down	O
everyone	O
's	O
experience	O
.	O
</s>
<s>
However	O
,	O
this	O
can	O
be	O
mitigated	O
by	O
installing	O
large	O
amounts	O
of	O
RAM	B-Architecture
on	O
the	O
server	B-Application
(	O
which	O
speeds	O
up	O
read	O
operations	O
by	O
improving	O
caching	B-General_Concept
performance	O
)	O
,	O
by	O
adding	O
more	O
servers	O
(	O
which	O
distributes	O
the	O
I/O	O
workload	O
)	O
,	O
or	O
by	O
adding	O
more	O
disks	B-Device
to	O
a	O
RAID	B-Architecture
array	I-Architecture
(	O
which	O
distributes	O
the	O
physical	O
I/O	O
workload	O
)	O
.	O
</s>
<s>
In	O
any	O
case	O
this	O
is	O
also	O
a	O
problem	O
which	O
can	O
affect	O
any	O
client-server	O
network	O
to	O
some	O
extent	O
,	O
since	O
,	O
of	O
course	O
,	O
fat	B-Protocol
clients	I-Protocol
also	O
use	O
servers	O
to	O
store	O
user	O
data	B-General_Concept
.	O
</s>
<s>
Indeed	O
,	O
user	O
data	B-General_Concept
may	O
be	O
much	O
more	O
significant	O
in	O
size	O
and	O
may	O
be	O
accessed	O
far	O
more	O
frequently	O
than	O
operating	B-General_Concept
systems	I-General_Concept
and	O
programs	B-Application
in	O
some	O
environments	O
,	O
so	O
moving	O
to	O
a	O
diskless	B-Device
model	O
will	O
not	O
necessarily	O
cause	O
a	O
noticeable	O
degradation	O
in	O
performance	O
.	O
</s>
<s>
capacity	O
)	O
will	O
also	O
be	O
used	O
in	O
a	O
diskless	B-Device
model	O
,	O
compared	O
to	O
a	O
fat	B-Protocol
client	I-Protocol
model	O
.	O
</s>
<s>
Finally	O
,	O
the	O
combination	O
of	O
network	O
data	B-General_Concept
transfer	O
latencies	O
(	O
physically	O
transferring	O
the	O
data	B-General_Concept
over	O
the	O
network	O
)	O
and	O
contention	O
latencies	O
(	O
waiting	O
for	O
the	O
server	B-Application
to	O
process	O
other	O
nodes	O
 '	O
requests	O
before	O
yours	O
)	O
can	O
lead	O
to	O
an	O
unacceptable	O
degradation	O
in	O
performance	O
compared	O
to	O
using	O
local	O
drives	O
,	O
depending	O
on	O
the	O
nature	O
of	O
the	O
application	O
and	O
the	O
capacity	O
of	O
the	O
network	O
infrastructure	O
and	O
the	O
server	B-Application
.	O
</s>
<s>
Another	O
example	O
of	O
a	O
situation	O
where	O
a	O
diskless	B-Device
node	I-Device
would	O
be	O
useful	O
is	O
in	O
a	O
possibly	O
hazardous	O
environment	O
where	O
computers	O
are	O
likely	O
to	O
be	O
damaged	O
or	O
destroyed	O
,	O
thus	O
making	O
the	O
need	O
for	O
inexpensive	O
nodes	O
,	O
and	O
minimal	O
hardware	O
a	O
benefit	O
.	O
</s>
<s>
Again	O
,	O
thin	B-Device
clients	I-Device
can	O
also	O
be	O
used	O
here	O
.	O
</s>
<s>
Diskless	B-Device
machines	O
may	O
also	O
consume	O
little	O
power	O
and	O
make	O
little	O
noise	O
,	O
which	O
implies	O
potential	O
environmental	O
benefits	O
and	O
makes	O
them	O
ideal	O
for	O
some	O
computer	B-Architecture
cluster	I-Architecture
applications	B-Application
.	O
</s>
<s>
Major	O
corporations	O
tend	O
to	O
instead	O
implement	O
thin	B-Device
clients	I-Device
(	O
using	O
Microsoft	B-General_Concept
Windows	I-General_Concept
Terminal	I-General_Concept
Server	I-General_Concept
or	O
other	O
such	O
software	B-Application
)	O
,	O
since	O
much	O
lower	O
specification	O
hardware	O
can	O
be	O
used	O
for	O
the	O
client	O
(	O
which	O
essentially	O
acts	O
as	O
a	O
simple	O
"	O
window	O
"	O
into	O
the	O
central	O
server	B-Application
which	O
is	O
actually	O
running	O
the	O
users	O
operating	B-General_Concept
system	I-General_Concept
as	O
a	O
login	B-Operating_System
session	I-Operating_System
)	O
.	O
</s>
<s>
Of	O
course	O
,	O
diskless	B-Device
nodes	I-Device
can	O
also	O
be	O
used	O
as	O
thin	B-Device
clients	I-Device
.	O
</s>
<s>
Moreover	O
,	O
thin	B-Device
client	I-Device
computers	O
are	O
increasing	O
in	O
power	O
to	O
the	O
point	O
where	O
they	O
are	O
becoming	O
suitable	O
as	O
fully-fledged	O
diskless	B-Device
workstations	I-Device
for	O
some	O
applications	B-Application
.	O
</s>
<s>
Both	O
thin	B-Device
client	I-Device
and	O
diskless	B-Device
node	I-Device
architectures	O
employ	O
diskless	B-Device
clients	I-Device
which	O
have	O
advantages	O
over	O
fat	B-Protocol
clients	I-Protocol
(	O
see	O
above	O
)	O
,	O
but	O
differ	O
with	O
regard	O
to	O
the	O
location	O
of	O
processing	O
.	O
</s>
<s>
Distributed	O
load	O
The	O
processing	O
load	O
of	O
diskless	B-Device
nodes	I-Device
is	O
distributed	O
.	O
</s>
<s>
Each	O
user	O
gets	O
its	O
own	O
processing	O
isolated	O
environment	O
,	O
barely	O
affecting	O
other	O
users	O
in	O
the	O
network	O
,	O
as	O
long	O
as	O
their	O
workload	O
is	O
not	O
filesystem-intensive	O
.	O
</s>
<s>
Thin	B-Device
clients	I-Device
rely	O
on	O
the	O
central	O
server	B-Application
for	O
the	O
processing	O
and	O
thus	O
require	O
a	O
fast	O
server	B-Application
.	O
</s>
<s>
When	O
the	O
central	O
server	B-Application
is	O
busy	O
and	O
slow	O
,	O
both	O
kinds	O
of	O
clients	O
will	O
be	O
affected	O
,	O
but	O
thin	B-Device
clients	I-Device
will	O
be	O
slowed	O
completely	O
,	O
whereas	O
diskless	B-Device
nodes	I-Device
will	O
only	O
be	O
slowed	O
when	O
accessing	O
data	B-General_Concept
on	O
the	O
server	B-Application
.	O
</s>
<s>
Diskless	B-Device
nodes	I-Device
have	O
advantages	O
over	O
thin	B-Device
clients	I-Device
in	O
multimedia-rich	O
applications	B-Application
that	O
would	O
be	O
bandwidth	O
intensive	O
if	O
fully	O
served	O
.	O
</s>
<s>
For	O
example	O
,	O
diskless	B-Device
nodes	I-Device
are	O
well	O
suited	O
for	O
video	O
gaming	O
because	O
the	O
rendering	O
is	O
local	O
,	O
lowering	O
the	O
latency	O
.	O
</s>
<s>
Peripheral	O
support	O
Diskless	B-Device
nodes	I-Device
are	O
typically	O
ordinary	O
personal	B-Device
computers	I-Device
or	O
workstations	B-Device
with	O
no	O
hard	B-Device
drives	I-Device
supplied	O
,	O
which	O
means	O
the	O
usual	O
large	O
variety	O
of	O
peripherals	O
can	O
be	O
added	O
.	O
</s>
<s>
By	O
contrast	O
,	O
thin	B-Device
clients	I-Device
are	O
typically	O
very	O
small	O
,	O
sealed	O
boxes	O
with	O
no	O
possibility	O
for	O
internal	O
expansion	O
,	O
and	O
limited	O
or	O
non-existent	O
possibility	O
for	O
external	O
expansion	O
.	O
</s>
<s>
a	O
USB	B-Protocol
device	O
can	O
be	O
physically	O
attached	O
to	O
a	O
thin	B-Device
client	I-Device
,	O
the	O
thin	B-Device
client	I-Device
software	B-Application
might	O
not	O
support	O
peripherals	O
beyond	O
the	O
basic	O
input	O
and	O
output	O
devices	O
-	O
for	O
example	O
,	O
it	O
may	O
not	O
be	O
compatible	O
with	O
graphics	B-Device
tablets	I-Device
,	O
digital	B-Device
cameras	I-Device
or	O
scanners	O
.	O
</s>
<s>
The	O
hardware	O
is	O
cheaper	O
on	O
thin	B-Device
clients	I-Device
,	O
since	O
processing	O
requirements	O
on	O
the	O
client	O
are	O
minimal	O
,	O
and	O
3D	B-Architecture
acceleration	I-Architecture
and	O
elaborate	O
audio	O
support	O
are	O
not	O
usually	O
provided	O
.	O
</s>
<s>
Of	O
course	O
,	O
a	O
diskless	B-Device
node	I-Device
can	O
also	O
be	O
purchased	O
with	O
a	O
cheap	O
CPU	B-General_Concept
and	O
minimal	O
multimedia	O
support	O
,	O
if	O
suitable	O
.	O
</s>
<s>
However	O
,	O
many	O
large	O
organizations	O
habitually	O
buy	O
hardware	O
with	O
a	O
higher	O
than	O
necessary	O
specification	O
to	O
meet	O
the	O
needs	O
of	O
particular	O
applications	B-Application
and	O
uses	O
,	O
or	O
to	O
ensure	O
future	O
proofing	O
(	O
see	O
next	O
point	O
)	O
.	O
</s>
<s>
There	O
are	O
also	O
less	O
"	O
rational	O
"	O
reasons	O
for	O
overspecifying	O
hardware	O
which	O
quite	O
often	O
come	O
into	O
play	O
:	O
departments	O
wastefully	O
using	O
up	O
budgets	O
in	O
order	O
to	O
retain	O
their	O
current	O
budget	O
levels	O
for	O
next	O
year	O
;	O
and	O
uncertainty	O
about	O
the	O
future	O
,	O
or	O
lack	O
of	O
technical	O
knowledge	O
,	O
or	O
lack	O
of	O
care	O
and	O
attention	O
,	O
when	O
choosing	O
PC	B-Device
specifications	O
.	O
</s>
<s>
Taking	O
all	O
these	O
factors	O
into	O
account	O
,	O
thin	B-Device
clients	I-Device
may	O
bring	O
the	O
most	O
substantial	O
savings	O
,	O
as	O
only	O
the	O
servers	O
are	O
likely	O
to	O
be	O
substantially	O
"	O
gold-plated	O
"	O
and/or	O
"	O
future-proofed	O
"	O
in	O
the	O
thin	B-Device
client	I-Device
model	O
.	O
</s>
<s>
Future	O
proofing	O
is	O
not	O
much	O
of	O
an	O
issue	O
for	O
thin	B-Device
clients	I-Device
,	O
which	O
are	O
likely	O
to	O
remain	O
useful	O
for	O
the	O
entirety	O
of	O
their	O
replacement	O
cycle	O
-	O
one	O
to	O
four	O
years	O
,	O
or	O
even	O
longer	O
-	O
as	O
the	O
burden	O
is	O
on	O
the	O
servers	O
.	O
</s>
<s>
There	O
are	O
issues	O
when	O
it	O
comes	O
to	O
diskless	B-Device
nodes	I-Device
,	O
as	O
the	O
processing	O
load	O
is	O
potentially	O
much	O
higher	O
,	O
thus	O
meaning	O
more	O
consideration	O
is	O
required	O
when	O
purchasing	O
.	O
</s>
<s>
Thin	B-Device
client	I-Device
networks	O
may	O
require	O
significantly	O
more	O
powerful	O
servers	O
in	O
the	O
future	O
,	O
whereas	O
a	O
diskless	B-Device
nodes	I-Device
network	O
may	O
in	O
future	O
need	O
a	O
server	B-Application
upgrade	O
,	O
a	O
client	O
upgrade	O
,	O
or	O
both	O
.	O
</s>
<s>
Thin	B-Device
client	I-Device
networks	O
have	O
less	O
network	O
bandwidth	O
consumption	O
potentially	O
,	O
since	O
much	O
data	B-General_Concept
is	O
simply	O
read	O
by	O
the	O
server	B-Application
and	O
processed	O
there	O
,	O
and	O
only	O
transferred	O
to	O
the	O
client	O
in	O
small	O
pieces	O
,	O
as	O
and	O
when	O
needed	O
for	O
display	B-Device
.	O
</s>
<s>
Also	O
,	O
transferring	O
graphical	O
data	B-General_Concept
to	O
the	O
display	B-Device
is	O
usually	O
more	O
suited	O
for	O
efficient	O
data	B-General_Concept
compression	I-General_Concept
and	O
optimisation	O
technologies	O
(	O
see	O
e.g.	O
</s>
<s>
NX	B-Application
technology	I-Application
)	O
than	O
transferring	O
arbitrary	O
programs	B-Application
,	O
or	O
user	O
data	B-General_Concept
.	O
</s>
<s>
In	O
many	O
typical	O
application	O
scenarios	O
,	O
both	O
total	O
bandwidth	O
consumption	O
and	O
"	O
burst	O
"	O
consumption	O
would	O
be	O
expected	O
to	O
be	O
less	O
for	O
an	O
efficient	O
thin	B-Device
client	I-Device
,	O
than	O
for	O
a	O
diskless	B-Device
node	I-Device
.	O
</s>
