<s>
Computation	B-Operating_System
offloading	I-Operating_System
is	O
the	O
transfer	O
of	O
resource	O
intensive	O
computational	O
tasks	O
to	O
a	O
separate	O
processor	O
,	O
such	O
as	O
a	O
hardware	B-General_Concept
accelerator	I-General_Concept
,	O
or	O
an	O
external	O
platform	O
,	O
such	O
as	O
a	O
cluster	B-Architecture
,	O
grid	B-Architecture
,	O
or	O
a	O
cloud	B-Architecture
.	O
</s>
<s>
Offloading	O
to	O
a	O
coprocessor	B-General_Concept
can	O
be	O
used	O
to	O
accelerate	O
applications	O
including	O
:	O
image	O
rendering	O
and	O
mathematical	O
calculations	O
.	O
</s>
<s>
The	O
first	O
concepts	O
of	O
stored-program	O
computers	O
were	O
developed	O
in	O
the	O
design	O
of	O
the	O
ENIAC	B-Device
,	O
the	O
first	O
general-purpose	O
digital	O
computer	O
.	O
</s>
<s>
The	O
ENIAC	B-Device
was	O
limited	O
in	O
performance	O
to	O
single	O
tasks	O
which	O
led	O
to	O
the	O
development	O
of	O
the	O
EDVAC	B-Device
which	O
would	O
become	O
the	O
first	O
computer	O
designed	O
to	O
perform	O
instructions	O
of	O
various	O
types	O
.	O
</s>
<s>
The	O
first	O
instances	O
of	O
computation	B-Operating_System
offloading	I-Operating_System
were	O
the	O
use	O
of	O
simple	O
sub-processors	O
to	O
handle	O
Input/output	O
processing	O
through	O
a	O
separate	O
system	O
called	O
Channel	B-Device
I/O	I-Device
.	O
</s>
<s>
During	O
the	O
1970s	O
,	O
coprocessors	B-General_Concept
began	O
being	O
used	O
to	O
accelerate	O
floating-point	B-Algorithm
arithmetic	I-Algorithm
faster	O
than	O
earlier	O
8-bit	O
and	O
16-bit	O
processors	O
which	O
used	O
software	O
.	O
</s>
<s>
As	O
a	O
result	O
,	O
math	O
coprocessors	B-General_Concept
became	O
common	O
for	O
scientific	O
and	O
engineering	O
calculations	O
.	O
</s>
<s>
Another	O
form	O
of	O
coprocessor	B-General_Concept
was	O
the	O
graphics	B-Architecture
coprocessor	I-Architecture
.	O
</s>
<s>
Coprocessors	B-General_Concept
were	O
common	O
in	O
most	O
computers	O
,	O
however	O
,	O
declined	O
in	O
usage	O
due	O
to	O
the	O
development	O
in	O
microprocessor	O
technologies	O
which	O
integrated	O
many	O
coprocessor	B-General_Concept
functions	O
.	O
</s>
<s>
Dedicated	O
graphics	B-Architecture
processing	I-Architecture
units	I-Architecture
,	O
however	O
,	O
are	O
still	O
widely	O
used	O
for	O
their	O
effectiveness	O
in	O
many	O
tasks	O
including	O
;	O
image	O
processing	O
,	O
machine	O
learning	O
,	O
parallel	B-Operating_System
computing	I-Operating_System
,	O
computer	O
vision	O
,	O
and	O
physical	O
simulation	O
.	O
</s>
<s>
In	O
the	O
1990s	O
telecommunications	O
companies	O
started	O
offering	O
virtual	B-Application
private	I-Application
network	I-Application
(	O
VPN	B-Application
)	O
services	O
.	O
</s>
<s>
The	O
cloud	B-Architecture
symbol	O
became	O
synonymous	O
with	O
the	O
interaction	O
between	O
providers	O
and	O
users	O
.	O
</s>
<s>
This	O
concept	O
known	O
as	O
grid	B-Architecture
computing	I-Architecture
was	O
associated	O
with	O
cloud	B-Architecture
computing	I-Architecture
systems	O
.	O
</s>
<s>
The	O
first	O
concept	O
of	O
linking	O
large	O
mainframes	O
to	O
provide	O
an	O
effective	O
form	O
of	O
parallelism	B-Operating_System
was	O
developed	O
in	O
the	O
1960s	O
by	O
IBM	O
.	O
</s>
<s>
Cluster	B-Architecture
computing	I-Architecture
was	O
used	O
by	O
IBM	O
was	O
to	O
increase	O
hardware	O
,	O
operating	O
system	O
,	O
and	O
software	O
performance	O
while	O
allowing	O
users	O
to	O
run	O
existing	O
applications	O
.	O
</s>
<s>
Clusters	O
could	O
efficiently	O
split	O
and	O
offload	O
computation	O
to	O
individual	O
nodes	O
for	O
increased	O
performance	O
while	O
also	O
gaining	O
scalability	B-Architecture
.	O
</s>
<s>
Computational	O
tasks	O
are	O
handled	O
by	O
a	O
central	B-General_Concept
processor	I-General_Concept
which	O
executes	O
instructions	O
by	O
carrying	O
out	O
rudimentary	O
arithmetic	O
,	O
control	O
logic	O
,	O
and	O
input/output	O
operations	O
.	O
</s>
<s>
Certain	O
application	O
processes	O
can	O
be	O
accelerated	O
by	O
offloading	O
tasks	O
from	O
the	O
main	O
processor	O
to	O
a	O
coprocessor	B-General_Concept
while	O
other	O
processes	O
might	O
require	O
an	O
external	O
processing	O
platform	O
.	O
</s>
<s>
For	O
example	O
,	O
a	O
graphics	B-Architecture
processing	I-Architecture
unit	I-Architecture
(	O
GPU	B-Architecture
)	O
,	O
which	O
consists	O
of	O
numerous	O
low-performance	O
cores	O
,	O
is	O
more	O
efficient	O
at	O
graphical	O
computation	O
than	O
a	O
CPU	O
which	O
features	O
fewer	O
,	O
high	O
power	O
,	O
cores	O
.	O
</s>
<s>
Hardware	B-General_Concept
accelerators	I-General_Concept
,	O
however	O
,	O
are	O
less	O
versatile	O
when	O
compared	O
to	O
a	O
CPU	O
.	O
</s>
<s>
Cloud	B-Architecture
Computing	I-Architecture
refers	O
to	O
both	O
the	O
applications	O
transported	O
over	O
the	O
Internet	O
and	O
the	O
hardware	O
and	O
software	O
in	O
the	O
data	O
centers	O
that	O
provide	O
services	O
;	O
which	O
include	O
data	O
storage	O
and	O
computing	O
.	O
</s>
<s>
Through	O
network	O
access	O
,	O
a	O
computer	O
can	O
migrate	O
part	O
of	O
its	O
computing	O
to	O
the	O
cloud	B-Architecture
.	O
</s>
<s>
Cluster	B-Architecture
computing	I-Architecture
is	O
a	O
type	O
of	O
parallel	B-Operating_System
processing	I-Operating_System
system	O
which	O
combines	O
interconnected	O
stand-alone	O
computers	O
to	O
work	O
as	O
a	O
single	O
computing	O
resource	O
.	O
</s>
<s>
Clusters	O
employ	O
a	O
parallel	B-Operating_System
programming	I-Operating_System
model	O
which	O
requires	O
fast	O
interconnection	O
technologies	O
to	O
support	O
high-bandwidth	O
and	O
low	O
latency	O
for	O
communication	O
between	O
nodes	O
.	O
</s>
<s>
Grid	B-Architecture
computing	I-Architecture
is	O
a	O
group	O
of	O
networked	O
computers	O
that	O
work	O
together	O
as	O
a	O
virtual	O
supercomputer	O
to	O
perform	O
intensive	O
computational	O
tasks	O
,	O
such	O
as	O
analyzing	O
huge	O
sets	O
of	O
data	O
.	O
</s>
<s>
Through	O
the	O
cloud	B-Architecture
,	O
it	O
is	O
possible	O
to	O
create	O
and	O
use	O
computer	O
grids	O
for	O
purposes	O
and	O
specific	O
periods	O
.	O
</s>
<s>
Unlike	O
with	O
parallel	B-Operating_System
computing	I-Operating_System
,	O
grid	B-Architecture
computing	I-Architecture
tasks	O
usually	O
have	O
no	O
time	O
dependency	O
associated	O
with	O
them	O
,	O
but	O
instead	O
,	O
they	O
use	O
computers	O
that	O
are	O
part	O
of	O
the	O
grid	B-Architecture
only	O
when	O
idle	O
,	O
and	O
users	O
can	O
perform	O
tasks	O
unrelated	O
to	O
the	O
grid	B-Architecture
at	O
any	O
time	O
.	O
</s>
<s>
Cluster	B-Architecture
computing	I-Architecture
is	O
more	O
cost-effective	O
than	O
a	O
single	O
computer	O
and	O
is	O
much	O
more	O
flexible	O
in	O
terms	O
of	O
adding	O
more	O
processing	O
units	O
.	O
</s>
<s>
Cloud	B-Architecture
computing	I-Architecture
services	O
have	O
issues	O
regarding	O
downtime	O
,	O
as	O
outages	O
can	O
occur	O
when	O
service	O
providers	O
are	O
overloaded	O
with	O
tasks	O
.	O
</s>
<s>
Cloud	B-Architecture
Providers	O
are	O
not	O
always	O
secure	O
,	O
and	O
in	O
an	O
event	O
of	O
a	O
security	O
breach	O
,	O
important	O
information	O
could	O
be	O
disclosed	O
.	O
</s>
<s>
Modern	O
computers	O
integrate	O
various	O
technologies	O
including	O
;	O
graphics	O
,	O
sound	O
hardware	O
,	O
and	O
networking	O
within	O
the	O
main	O
circuit	O
board	O
which	O
makes	O
many	O
hardware	B-General_Concept
accelerators	I-General_Concept
irrelevant	O
.	O
</s>
<s>
Using	O
cloud	B-Architecture
services	I-Architecture
removes	O
individual	O
control	O
over	O
the	O
hardware	O
and	O
forces	O
users	O
to	O
trust	O
that	O
cloud	B-Architecture
providers	O
will	O
maintain	O
infrastructure	O
and	O
properly	O
secure	O
their	O
data	O
.	O
</s>
<s>
Cloud	B-Architecture
services	I-Architecture
can	O
be	O
described	O
by	O
3	O
main	O
cloud	B-Architecture
service	I-Architecture
models	O
:	O
SaaS	B-Architecture
,	O
PaaS	B-Application
,	O
and	O
IaaS	B-General_Concept
.	O
</s>
<s>
Software	B-Architecture
as	I-Architecture
a	I-Architecture
service	I-Architecture
(	O
SaaS	B-Architecture
)	O
is	O
an	O
externally	O
hosted	O
service	O
which	O
can	O
be	O
easily	O
accessed	O
by	O
a	O
consumer	O
through	O
a	O
web	O
browser	O
.	O
</s>
<s>
Platform	B-Application
as	I-Application
a	I-Application
service	I-Application
(	O
PaaS	B-Application
)	O
is	O
a	O
development	O
environment	O
where	O
software	O
can	O
be	O
built	O
,	O
tested	O
,	O
and	O
deployed	O
without	O
the	O
user	O
having	O
to	O
focus	O
on	O
building	O
and	O
maintaining	O
computational	O
infrastructure	O
.	O
</s>
<s>
Infrastructure	B-General_Concept
as	I-General_Concept
a	I-General_Concept
service	I-General_Concept
(	O
IaaS	B-General_Concept
)	O
is	O
access	O
to	O
an	O
infrastructure	O
's	O
resources	O
,	O
network	O
technology	O
,	O
and	O
security	O
compliance	O
which	O
can	O
be	O
used	O
by	O
enterprises	O
to	O
build	O
software	O
.	O
</s>
<s>
Cloud	B-Architecture
computing	I-Architecture
services	O
provide	O
users	O
with	O
access	O
to	O
large	O
amounts	O
of	O
computational	O
power	O
and	O
storage	O
not	O
viable	O
on	O
local	O
computers	O
without	O
having	O
significant	O
expenditure	O
.	O
</s>
<s>
Despite	O
the	O
constant	O
development	O
in	O
key	O
components	O
including	O
;	O
CPU	O
,	O
GPU	B-Architecture
,	O
memory	O
,	O
and	O
wireless	O
access	O
technologies	O
;	O
mobile	O
devices	O
need	O
to	O
be	O
portable	O
and	O
energy	O
efficient	O
.	O
</s>
<s>
Mobile	B-General_Concept
cloud	I-General_Concept
computing	I-General_Concept
is	O
the	O
combination	O
of	O
cloud	B-Architecture
computing	I-Architecture
and	O
mobile	O
computing	O
,	O
in	O
which	O
mobile	O
devices	O
perform	O
computation	B-Operating_System
offloading	I-Operating_System
to	O
balance	O
the	O
power	O
of	O
the	O
cloud	B-Architecture
for	O
accelerating	O
application	O
execution	O
and	O
saving	O
energy	O
consumption	O
.	O
</s>
<s>
In	O
this	O
computation	B-Operating_System
offloading	I-Operating_System
,	O
a	O
mobile	O
device	O
migrates	O
part	O
of	O
its	O
computing	O
to	O
the	O
cloud	B-Architecture
.	O
</s>
<s>
These	O
input/output	O
operations	O
rely	O
on	O
a	O
computer	O
and	O
its	O
components	O
,	O
including	O
the	O
CPU	O
,	O
GPU	B-Architecture
,	O
RAM	O
,	O
and	O
storage	O
.	O
</s>
<s>
The	O
CPU	O
is	O
responsible	O
for	O
processing	O
input	O
from	O
the	O
user	O
and	O
passing	O
information	O
to	O
the	O
GPU	B-Architecture
.	O
</s>
<s>
The	O
GPU	B-Architecture
does	O
not	O
have	O
access	O
to	O
a	O
computer	O
's	O
main	O
memory	O
so	O
instead	O
,	O
graphical	O
assets	O
have	O
to	O
be	O
loaded	O
into	O
VRAM	O
which	O
is	O
the	O
GPU	B-Architecture
's	O
memory	O
.	O
</s>
<s>
The	O
CPU	O
is	O
responsible	O
for	O
instructing	O
the	O
GPU	B-Architecture
while	O
the	O
GPU	B-Architecture
uses	O
the	O
information	O
to	O
render	O
an	O
image	O
on	O
to	O
an	O
output	O
device	O
.	O
</s>
<s>
CPU	O
's	O
are	O
able	O
to	O
run	O
games	O
without	O
a	O
GPU	B-Architecture
through	O
software	O
rendering	O
,	O
however	O
,	O
offloading	O
rendering	O
to	O
a	O
GPU	B-Architecture
which	O
has	O
specialized	O
hardware	O
results	O
in	O
improved	O
performance	O
.	O
</s>
