Image Processing Reference
In-Depth Information
Table 7.2 Dynamic resource migration algorithm
1: Inputs:
MN
m
, Location(
MN
m
8
[
S
t
,
P
d
]), uploaded resources
S
aj
2: find the
r
k
that hosts F
a
3: for all
a
j
claims
4: find an optimal
C
r
k
,
a
j
such that
C
r
k
,
a
j
1
5: Estimate the [
S
t
,
P
d
] for each
r
k
,
C
r
k
,
a
j
of the claim of
MN
m
6:
if
r
k
has not enough resources satisfying the
MN
m
8
[
S
t
,
P
d
]
7:
then
use the partitionable notation of
a
nearby rack satisfying
X
n
8:
s
þ
pxðÞ
z
P
, for each
r
k
,
C
r
k
,
a
j
i¼
1
9: Assign
a
j
to the
r
k
terminals
10:
end for
to respond to the performance requirements [
12
]. A significant measure in the
system is the capacity of the rack that was not utilized in order to service an
application resource/task, enabling a reallocation decision on that proportion. The
capacity metric is used to measure the capacity loss, as in Eq. (
7.4
), where
r
k
is the
rack,
a
j
is an application and
s
j
k
is the number of terminals
P
in rack
r
k
(
P
) hosting
application
a
j
, and
S
a
j
k
ðÞ
is the number of servers hosting application across all
different racks
r
k
(P)
.
s
k
C
r
k
,
a
j
¼
X
ð
7
:
4
Þ
S
a
j
k
ðÞ
k
Equation (
7.4
) shows that if there is minimal loss in the capacity utilization, i.e.,
C
r
k
,
a
j
is optimally utilized. The latter is shown
through the conducted simulation experiments in the next section. The dynamic
resource migration algorithm is shown in Table
7.2
with the basic steps for
obtaining an efficient execution for a partitionable resource that cannot be handled
by the existing cloud rack and therefore the migration policy is used to ensure that it
will be continuing the execution. The continuation is based on the migrated policy
of the partitionable processes that are split, in order to be handled by other cloud
rack terminals and thus omit any potential failures.
1 then the sequence of racks
S
a
j
k
ðÞ
7.4 Performance Evaluation Analysis, Experimental
Results, and Discussion
The performance evaluation encompasses the extracted results conducted by
simulation experiments, using the network model—as per the mentioned scenario
above. The scenario was simulated using discrete event simulation model using
