<s>
In	O
distributed	B-Architecture
system	I-Architecture
and	O
system	B-General_Concept
resource	I-General_Concept
,	O
elasticity	B-General_Concept
is	O
defined	O
as	O
"	O
the	O
degree	O
to	O
which	O
a	O
system	O
is	O
able	O
to	O
adapt	O
to	O
workload	O
changes	O
by	O
provisioning	O
and	O
de-provisioning	O
resources	O
in	O
an	O
autonomic	O
manner	O
,	O
such	O
that	O
at	O
each	O
point	O
in	O
time	O
the	O
available	O
resources	O
match	O
the	O
current	O
demand	O
as	O
closely	O
as	O
possible	O
"	O
.	O
</s>
<s>
Elasticity	B-General_Concept
is	O
a	O
defining	O
characteristic	O
that	O
differentiates	O
cloud	B-Architecture
computing	I-Architecture
from	O
previously	O
proposed	O
computing	O
paradigms	O
,	O
such	O
as	O
grid	B-Architecture
computing	I-Architecture
.	O
</s>
<s>
The	O
dynamic	O
adaptation	O
of	O
capacity	O
,	O
e.g.	O
,	O
by	O
altering	O
the	O
use	O
of	O
computing	B-General_Concept
resources	I-General_Concept
,	O
to	O
meet	O
a	O
varying	O
workload	O
is	O
called	O
"	O
elastic	B-General_Concept
computing	I-General_Concept
"	O
.	O
</s>
<s>
In	O
the	O
world	O
of	O
distributed	B-Architecture
systems	I-Architecture
,	O
there	O
are	O
several	O
definitions	O
according	O
to	O
the	O
authors	O
,	O
some	O
considering	O
the	O
concepts	O
of	O
scalability	B-Architecture
a	O
sub-part	O
of	O
elasticity	B-General_Concept
,	O
others	O
as	O
being	O
distinct	O
.	O
</s>
<s>
Let	O
us	O
illustrate	O
elasticity	B-General_Concept
through	O
a	O
simple	O
example	O
of	O
a	O
service	O
provider	O
who	O
wants	O
to	O
run	O
a	O
website	O
on	O
an	O
IaaS	B-General_Concept
cloud	O
.	O
</s>
<s>
At	O
moment	O
,	O
the	O
website	O
is	O
unpopular	O
and	O
a	O
single	O
machine	O
(	O
most	O
commonly	O
a	O
virtual	B-Architecture
machine	I-Architecture
)	O
is	O
sufficient	O
to	O
serve	O
all	O
web	O
users	O
.	O
</s>
<s>
At	O
moment	O
,	O
the	O
website	O
suddenly	O
becomes	O
popular	O
,	O
for	O
example	O
,	O
as	O
a	O
result	O
of	O
a	O
flash	B-General_Concept
crowd	I-General_Concept
,	O
and	O
a	O
single	O
machine	O
is	O
no	O
longer	O
sufficient	O
to	O
serve	O
all	O
users	O
.	O
</s>
<s>
Based	O
on	O
the	O
number	O
of	O
web	O
users	O
simultaneously	O
accessing	O
the	O
website	O
and	O
the	O
resource	O
requirements	O
of	O
the	O
web	B-Application
server	I-Application
,	O
it	O
might	O
be	O
that	O
ten	O
machines	O
are	O
needed	O
.	O
</s>
<s>
Elasticity	B-General_Concept
aims	O
at	O
matching	O
the	O
amount	O
of	O
resource	O
allocated	O
to	O
a	O
service	O
with	O
the	O
amount	O
of	O
resource	O
it	O
actually	O
requires	O
,	O
avoiding	O
over	O
-	O
or	O
under-provisioning	O
.	O
</s>
<s>
For	O
example	O
,	O
an	O
Amazon	B-General_Concept
EC2	I-General_Concept
M4	O
extra-large	O
instance	O
costs	O
US	O
$0.239	O
/hour	O
.	O
</s>
<s>
If	O
a	O
service	O
has	O
allocated	O
two	O
virtual	B-Architecture
machines	I-Architecture
when	O
only	O
one	O
is	O
required	O
,	O
the	O
service	O
provider	O
wastes	O
$	O
2,095	O
every	O
year	O
.	O
</s>
<s>
One	O
potential	O
problem	O
is	O
that	O
elasticity	B-General_Concept
takes	O
time	O
.	O
</s>
<s>
A	O
cloud	O
virtual	B-Architecture
machine	I-Architecture
(	O
VM	O
)	O
can	O
be	O
acquired	O
at	O
any	O
time	O
by	O
the	O
user	O
;	O
however	O
,	O
it	O
may	O
take	O
up	O
to	O
several	O
minutes	O
for	O
the	O
acquired	O
VM	O
to	O
be	O
ready	O
to	O
use	O
.	O
</s>
<s>
This	O
implies	O
any	O
control	O
mechanism	O
designed	O
for	O
elastic	O
applications	O
must	O
consider	O
in	O
its	O
decision	O
process	O
the	O
time	O
needed	O
for	O
the	O
elasticity	B-General_Concept
actions	O
to	O
take	O
effect	O
,	O
such	O
as	O
provisioning	O
another	O
VM	O
for	O
a	O
specific	O
application	O
component	O
.	O
</s>
<s>
VM	O
)	O
,	O
such	O
as	O
Ganglia	B-Language
or	O
Nagios	B-Application
,	O
are	O
no	O
longer	O
suitable	O
for	O
monitoring	O
the	O
behavior	O
of	O
elastic	O
applications	O
.	O
</s>
<s>
When	O
deploying	O
applications	O
in	O
cloud	O
infrastructures	O
(	O
IaaS/PaaS	O
)	O
,	O
requirements	O
of	O
the	O
stakeholder	O
need	O
to	O
be	O
considered	O
in	O
order	O
to	O
ensure	O
proper	O
elasticity	B-General_Concept
behavior	O
.	O
</s>
<s>
Even	O
though	O
traditionally	O
one	O
would	O
try	O
to	O
find	O
the	O
optimal	O
trade-off	O
between	O
cost	O
and	O
quality	O
or	O
performance	O
,	O
for	O
real	O
world	O
cloud	B-Architecture
users	I-Architecture
requirements	O
regarding	O
the	O
behavior	O
are	O
more	O
complex	O
and	O
target	O
multiple	O
dimensions	O
of	O
elasticity	B-General_Concept
(	O
e.g.	O
,	O
SYBL	O
)	O
.	O
</s>
<s>
Cloud	B-Architecture
applications	I-Architecture
can	O
be	O
of	O
varying	O
types	O
and	O
complexities	O
,	O
with	O
multiple	O
levels	O
of	O
artifacts	O
deployed	O
in	O
layers	O
.	O
</s>
<s>
For	O
multi-level	O
control	O
,	O
control	O
systems	O
need	O
to	O
consider	O
the	O
impact	O
lower	O
level	O
control	O
has	O
upon	O
higher	O
level	O
ones	O
and	O
vice	O
versa	O
(	O
e.g.	O
,	O
controlling	O
virtual	B-Architecture
machines	I-Architecture
,	O
web	O
containers	O
,	O
or	O
web	O
services	O
in	O
the	O
same	O
time	O
)	O
,	O
as	O
well	O
as	O
conflicts	O
which	O
may	O
appear	O
between	O
various	O
control	O
strategies	O
from	O
various	O
levels	O
.	O
</s>
