Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
9
, No
.
3
,
J
un
e
201
9
, pp.
1957
~
19
67
IS
S
N:
20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v
9
i
3
.
pp1957
-
19
67
1957
Journ
al h
om
e
page
:
http:
//
ia
es
core
.c
om/
journa
ls
/i
ndex.
ph
p/IJECE
ERMO
2
a
lgo
rithm:
a
n
e
nerg
y
e
ffi
cient
m
ob
ility
m
an
agemen
t in
m
obile
c
loud
c
om
pu
ting
s
ystem
for 5
G
h
etero
ge
n
eous
n
etw
or
ks
L. Pa
ll
avi
1
,
A. J
agan
2
,
B. Th
ir
uma
la
Ra
o
3
1
,3
Depa
rt
m
ent
of
Com
pute
r
Sci
en
ce
and
Engi
ne
ering,
KL
Univer
sit
y
,
Indi
a
2
Depa
rt
m
ent
of C
om
pute
r
Scie
n
ce
and
Engi
ne
ering,
BVRIT
,
Indi
a
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
J
un
10
, 2
01
8
Re
vised N
ov
1
0,
201
8
Accepte
d
Dec
5
, 2
01
8
Rec
en
tly
,
m
obile
device
s
are
b
e
coming
the
primar
y
p
la
t
form
s
for
eve
r
y
user
who
al
wa
y
s
ro
am
aro
und
and
acce
ss
the
c
lo
ud
computing
a
ppli
c
at
ions.
Mobile
cl
oud
c
om
puti
ng
(MC
C)
combines
the
both
m
obil
e
and
cl
oud
computing,
whi
ch
provide
s
opti
m
al
servic
es
to
the
m
obil
e
users.
In
next
-
gene
ra
ti
on
m
obi
le
envi
ronm
ent
s,
m
ai
nl
y
du
e
to
t
he
huge
num
ber
of
m
obile
users
in
con
junction
with
th
e
sm
al
l
cell
si
ze
and
the
ir
porta
b
le
inf
orm
at
ion’s,
the
inf
lue
nc
e
of
m
obil
ity
on
th
e
net
work
per
form
anc
e
is
strengt
h
e
ned.
In
thi
s
pape
r,
we
propo
se
an
ene
rg
y
eff
i
ci
en
t
m
obil
ity
m
ana
gement
in
m
obil
e
c
loud
computing
(E2
M2M
C2)
sy
ste
m
for
5G
het
ero
gene
ous
ne
t
works
.
The
proposed
E2M2MC2
sy
st
em
use
el
e
ctive
r
epeat
m
ult
i
-
objecti
v
e
opti
m
iz
ation
(ERMO2)
al
gorit
hm
to
det
ermi
ne
the
best
cl
o
uds
base
d
on
the
select
i
on
m
et
ric
s
ar
e
d
el
a
y
,
j
it
t
er,
b
it
err
or
rate
(BER)
,
pac
ke
t
loss,
co
m
m
unic
at
ion
cost,
response
tim
e,
and
net
work
loa
d.
ERMO2
al
gorit
hm
provide
s
ene
r
g
y
eff
icient
m
ana
g
ement
of
user
m
obil
ity
as
we
l
l
as
ne
twork
re
source
s.
Th
e
sim
ula
ti
on
resul
ts
show
s
tha
t
t
he
proposed
E
2M2M
C2
sy
ste
m
hel
ps
i
n
m
ini
m
iz
ing
dela
y
,
p
ac
ke
t
los
s
rat
e
and
e
ner
g
y
consum
pti
on
in
a
het
ero
g
ene
ous n
et
work.
Ke
yw
or
d
s
:
Be
st C
loud
Energy E
ff
ic
ie
nt
Heter
og
e
ne
ou
s
N
et
w
ork
Mob
il
e Cl
oud
Com
pu
ti
ng
Mob
il
it
y M
anag
em
ent
Copyright
©
201
9
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
L. P
al
la
vi,
Dep
a
rtm
ent o
f C
om
pu
te
r
Scie
nce a
nd E
ng
i
ne
erin
g,
KL Un
i
versi
ty
,
Vaddes
war
am
, Gun
t
ur,
522502, A.
P, India
.
Em
a
il
:
palla
vi5
03@
gm
ail.co
m
1.
INTROD
U
CTION
Cl
oud
com
pu
ti
ng
is g
e
ne
rall
y e
m
plo
ys fo
r
t
he
co
m
pu
ti
ng
r
e
so
urces
wh
ic
h
are d
ist
rib
uted by the h
el
p
of
netw
ork.
S
uppose,
if
t
he
functi
ons
an
d
data
are
m
ai
nt
ai
ned
on
t
he
inter
net
then
it
offer
i
ng
on
-
dem
and
acce
ss
[
1].
A
pp
li
cat
io
ns
a
re
norm
al
ly
execu
te
on
a
r
e
m
ote
serv
er.
Af
te
r
wa
rd,
it
is
transm
it
te
d
to
the
consum
er.
Mo
bile
-
le
ar
ning
is
m
os
tl
y
ta
ke
place
if
the
po
pu
la
ces
a
re
fa
r
from
their
off
ic
es
or
cl
assr
oom
s.
Af
te
r
the
sch
oo
l
or
office,
seve
ral
popula
ces
are
ch
ose
n
li
ste
n
m
us
ic
,
ra
dio
ne
ws
,
or
s
ports
pro
gr
am
s.
More
ov
e
r,
the
y
desire
to
le
ar
n
m
ob
il
e
de
vic
es,
DVD/C
D
P
la
ye
rs,
vid
e
otapes,
com
pu
te
rs
instal
le
d
by
m
eans
of
le
a
rn
i
ng
s
oft
war
e
or
com
pu
te
rs
with
high
s
peed
acce
ss
to
the
I
nter
ne
t
after
they
get
hom
e
wh
ic
h
are
al
l
base
d
on
e
-
le
arn
i
ng
[
2].
G
ener
al
ly
,
the
m
o
bile
ph
on
e
s
are
e
m
plo
ys
fo
r
c
omm
un
i
cat
ion
s
with
f
ur
t
he
r
popula
ce,
not
f
or
le
ar
ning
int
ention.
M
-
le
ar
ning
is
norm
al
l
y
an
innova
ti
ve
te
chnolo
gy
f
or
le
ar
ning
bu
t
it
do
es
no
t
s
wa
p
c
onve
ntion
al
le
a
rn
i
ng
[
3]. H
ere
,
t
he
m
ob
il
e
phones
a
re
c
on
ta
in
ing
a
dim
inu
ti
ve
m
e
m
or
y
and
stor
a
ge
capab
il
it
y,
wh
i
ch
are
restrict
s
the
le
ng
th
of
e
m
ai
ls,
the
op
enin
g
of
at
ta
chm
ents,
and
th
e
e
m
plo
y
of
ta
bs
or
nu
m
erous win
dows [
4].
Re
centl
y,
the
a
ccess
of
cl
oud
serv
e
rs
thr
ough
m
ob
il
e
ph
one
is
c
on
si
der
e
d
as
a
ge
ne
ral
a
nd
tre
nd
y.
I
n
2018,
t
he
cl
oud
a
pp
li
cat
io
ns
are
e
ncou
ntere
d
90%
of
m
obil
e
data
traf
fic.
Ge
ner
al
ly
,
the
m
ob
il
e
dev
ic
e
s
ar
e
con
ta
ini
ng
nu
m
ero
us
obta
in
able
sto
rag
e
s
erv
ic
es
li
ke
D
rop
box,
iC
lo
ud,
G
oogle
Dr
i
ve,
a
nd
S
ky
dri
v
e
[
5
].
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
3
,
June
201
9
:
1957
-
1967
1958
The
f
undam
en
ta
l
te
chn
ol
og
y
of
cl
oud
c
ompu
ti
ng
is
us
e
d
to
centrali
ze
t
he
com
pu
ti
ng,
serv
ic
es
,
an
d
pr
eci
se
app
li
cat
io
ns
by
the
help
of
se
rv
ic
e
wh
ic
h
is
distrib
uted
li
ke
water,
ga
s
or
el
ect
rici
ty
to
con
s
um
er.
T
herefo
re
,
the
m
ixtur
e
of
ub
i
qu
it
ie
s
m
ob
il
e
netwo
r
k
an
d
cl
oud
com
puti
ng
is
us
ed
to
eng
e
nder
a
n
inn
ovat
ive
com
pu
ti
ng
appr
oach,
wh
i
ch
is
known
as
MC
C
[
6
]
.
In
m
od
ern
com
pute
r
reg
io
ns,
the
m
ob
il
it
y
is
tur
n
out
to
be
a
trend
y
m
et
rics.
Norm
al
ly
,
the
m
ob
il
e
de
vices
li
ke
Sm
artph
one,
P
DA,
GP
S
Na
vi
gation
an
d
la
pt
op
s
are
ex
pa
nded
by
var
i
ou
s
m
ob
il
e
com
pu
ti
ng
,
netw
ork
i
ng
a
nd
sec
ur
it
y
te
chnolo
gies
[
7
]
.
Her
e
,
ex
pa
ns
ion
rate
of
co
m
po
un
d
pro
gr
am
is
qu
ic
ker
tha
n
the
e
xp
a
ns
i
on
rate
of
com
pu
ti
ng
r
eso
ur
ces
of
m
ob
il
e
dev
ic
es
w
hich
are
based
on
the
exp
a
ns
i
on
of
c
om
pu
ti
ng
-
inte
ns
ive
a
pp
li
cat
ion
s
[
8
]
.
Additi
on
al
ly
,
the
sur
fin
g
of
In
te
rn
e
t
is
turn
ed
ou
t
to
be
easi
er
by
the
e
xp
a
ns
i
on
of
wi
reless
te
ch
nolo
gy
li
ke
WiM
ax
,
A
d
H
oc
Netw
ork,
an
d
WIFI
,
but
it
not
restr
ic
te
d
by
m
eans
of
c
ables.
T
heref
ore,
th
os
e
m
ob
il
e
de
vices
are
e
sta
blished
as
t
heir
forem
os
t
op
ti
on
of
f
unct
ion
i
ng
and
e
ntertai
nm
ent
in
pe
op
le
li
ves
[
9
]
.
Ba
se
d
on
the
an
al
ysi
s
of
J
un
i
per
,
t
he
cl
oud
c
om
pu
ti
ng
relat
ed
m
ob
il
e
so
ft
war
e
a
nd
a
pp
li
cat
io
n
are
antic
ipate
d
to
increase
88%
per
a
nnum
fro
m
20
09
t
o
2014.
T
his
e
xp
a
nsi
on
is
al
so
gen
e
rati
ng
U
S
9
.5 bill
ion
do
ll
ars
in 2
014 [
10
],
[
11
].
The
f
or
em
os
t
ben
e
fit
of
MC
C
is
to
con
quer
diff
ic
ulti
es
of
m
ob
il
e
co
m
pu
ti
ng
w
hich
a
re
associat
ed
t
o
the prese
ntati
on
, atm
os
ph
e
re,
and protect
io
n [
1
2
]
. Mobil
e c
om
pu
ti
ng
is
ge
ner
al
ly
d
e
rive
d from
thr
ee f
ore
m
os
t
per
ce
ptio
ns
li
ke
hard
war
e
,
s
oft
war
e
an
d
c
om
m
un
ic
at
ion
.
The
per
ce
ptio
n
of
ha
rdwar
e
is
ind
ic
at
in
g
m
ob
il
e
dev
ic
es
as
sm
a
rt
phone
a
nd
l
apto
p,
or
t
heir
m
ob
il
e
co
m
ponen
ts
.
The
dif
f
ic
ulty
of
com
m
un
ic
at
ion
is
m
ai
nl
y
enco
m
passing
the
c
omm
un
ic
at
ion
s
of
m
ob
il
e
netw
ork
s,
pr
oto
c
ols
a
nd
data
deliv
eran
ce
.
S
om
e
of
th
e
at
tribu
te
s
of
m
ob
il
e
com
pu
ti
ng
are
s
pecified
as
fo
ll
ows
[
1
3
]
:
In
m
ob
il
e
com
pu
ti
ng
netw
ork
,
the
m
ob
il
e
node
s
are
us
e
d
to
m
a
ke
ass
ociat
io
n
with
oth
e
rs.
M
or
e
over,
the
pr
edeterm
ined
node
s
in
wi
red
netw
ork
a
re
ta
ke
place
by m
ob
il
e sup
port stat
ion (M
SS)
f
or
t
h
e
per
i
od of m
ov
in
g.
Norm
al
l
y,
the
e
m
plo
yi
ng
of
m
ob
il
e
no
de
s
in
netw
ork
is
not
excl
us
ive
.
T
her
e
fore,
t
his
netw
ork
ca
n
be
a
wired
ne
twork
t
hroug
h
high
-
ba
ndwidt
h,
lo
w
-
ba
ndw
idth,
or
e
ve
n
in
posit
ion
of
disco
nnect
ed
[1
4
].
Accor
ding
to
r
est
rict
ion
of
ba
tt
ery
p
ower
,
charge
of
wire
le
ss
com
m
un
ic
at
ion
,
netw
ork
co
nd
it
io
ns
a
nd
et
c,
the
m
ob
il
e
node
s
are
not
c
ons
ta
ntly
m
ai
ntaining
t
he
a
sso
ci
a
ti
on
,
but
detach
a
nd
reli
a
ble
thr
ough
t
he
wir
el
ess
netw
ork
is
ta
ke
place
as
i
na
ct
ively
or
act
i
vely
[1
5
]
.
Here,
the
se
r
ver
s
,
acce
s
s
points
,
and
f
ur
t
her
M
SS
are
facil
it
at
ing
a
s
tur
dy
sen
d/rec
ei
ve
capa
bili
ty.
This
ca
pa
bili
ty
in
m
ob
il
e
node
s
is
so
m
ewh
at
wea
k.
The
refor
e
,
the
com
m
un
ic
at
ion
band
widt
h
a
nd
over
hea
d
am
on
g
dow
nlink
an
d
up
li
nk
are
i
nconsi
ste
ncy
[
1
6
]
.
A
m
ob
il
e
com
pu
ti
ng
net
work
syst
em
is
gen
erall
y
reco
gniz
ed
fro
m
te
r
m
inals,
netw
orks,
database
platf
or
m
s,
and
app
li
cat
io
ns
e
xpan
sio
n
f
or
defense
by r
ea
son
of sig
nal is
vu
l
ner
a
ble to
obst
ru
ct
io
n
a
nd s
noopin
g.
In
co
ntrast
t
o
the
c
onve
ntio
nal
wi
red
net
work,
t
he
m
ob
il
e
com
pu
ti
ng
netw
ork
i
s
enco
m
passing
div
e
rse
dif
ficult
ie
s
and
ch
al
le
ng
es
in
di
ssi
m
il
ar
featur
es
li
ke
sign
al
interrupti
on,
pr
otect
io
n,
ha
nd
-
off
po
st
ponem
ent,
restrict
ed
po
wer
a
nd
l
ow
com
pu
ti
ng
c
apab
il
it
y
[1
7
]
.
Thr
ee
forem
os
t
possessi
ons
ar
e
consi
der
e
d
f
or
the
lim
it
ed
at
t
rib
utes
of
m
obil
e
netwo
r
ks
:
there
is
a
re
qu
i
rem
ent
fo
r
co
m
bin
ing
posse
ssion
s
from
nu
m
ero
us
dev
ic
es
beca
us
e
m
ob
il
e
dev
ic
e
is
enco
m
passing
re
stric
te
d
res
ource
th
an
co
nventio
na
l
wire
d
netw
orks
[
1
8
]
.
Ph
ysi
cal
un
iq
uen
e
ss
of
netw
ork
is
necessa
r
y
to
incorporat
ed
li
ke
sta
ble
r
ou
te
s
,
an
d
et
c
[1
9
]
.
Energy
li
m
i
ta
t
ion
s
are
sig
nificant
in
m
ob
il
e
networ
k
be
cause
t
he
re
cent
ap
plica
ti
on
s
are
nec
es
sit
at
ing
add
it
io
nal
batte
ry
[
20
]
.
O
n
the
oth
e
r
ha
nd,
the
at
tri
bu
te
s
of
m
ob
il
it
y
is
al
so
c
on
ta
ini
ng
t
he
disad
va
nt
age
of
rap
i
dly
energy
reducti
on
an
d
unbalan
ced
m
et
hod
[
15]
-
[
20
].
Fo
r
t
his
rea
so
n,
the
pro
po
sed
e
nergy
ef
f
ic
ie
nt
m
ob
il
i
ty
m
ana
gem
ent
in
m
ob
il
e
cl
oud
c
om
pu
ti
ng
(E
2M
2MC2)
syst
em
is
require
d
for
ne
xt
gen
e
rati
on
5G
heter
og
e
ne
ou
s
netw
orks
.
T
he
m
a
in
of
pro
po
s
ed
E2M
2M
C2
syst
em
i
s
to
com
pu
te
th
e
optim
al
best
cl
oud
a
m
on
g
ot
her
s
,
wh
ic
h
prov
i
de
s
co
nn
ect
io
n
le
ss
ser
vice
to
e
nd
m
ob
il
e
us
er
s;
and
the
us
e
r
m
ob
il
ity
is
m
a
nag
e
d
by the
bac
k
tra
ck
sea
rch
i
ng (B
TS)
al
gorith
m
w
it
h
congest
ion
c
ontr
ol als
o.
The
r
em
ai
nd
er
of
t
his
pa
pe
r
is
orga
niz
ed
as
fo
ll
ows
.
Se
ct
ion
2
disc
usse
s
rece
nt
rel
at
ed
w
ork.
In
Sect
io
n
3,
we
present
the
prob
le
m
m
e
tho
dolo
gy
an
d
s
yst
e
m
m
od
el
of
pro
posed
work.
Sect
ion
4
de
scribes
the
wor
king
f
unct
ion
of
pro
pose
d
E2M
2MC2
syst
e
m
in
detai
l
with
pro
per
m
at
he
m
at
i
cal
m
od
el
s.
Se
ct
i
on
5
pro
vid
es
t
he
si
m
ula
ti
on
res
ult
s
an
d
perf
or
m
a
nce
a
naly
sis
w
it
h
pro
per
te
st
scenari
os
.
The
pa
per
co
nclu
de
s
in
Sect
ion
6.
2.
RELATE
D
W
ORKS
Sar
dis
et
al
.
[21]
ha
ve
offe
re
d
a
possi
ble
ci
rcu
m
sta
nce
f
or
to
ge
ner
at
e
t
r
avel
c
ongestio
n
dif
ficult
ie
s
on
t
he
I
nter
n
et
by
reas
on
of
el
evated
band
width
m
edia
serv
ic
es
a
nd
co
ns
um
er
m
ob
il
i
ty
.
Her
e,
t
he
l
og
ic
al
structu
re
is
em
plo
ys
to
exam
ine
the
featur
e
w
hich
co
nc
ern
the
Qu
al
it
y
of
Ex
per
ie
nc
e
(QoE)
a
nd
QoS
for
VoD
ser
vices
in
a
m
ob
il
e
at
m
os
ph
ere.
T
he
cl
oud
-
relat
ed
serv
ic
es
a
re
ge
ner
al
ly
pro
vid
in
g
reas
onabl
e
and
centrali
zed
co
m
pu
ti
ng
resou
r
ces
and
m
ob
il
e
dev
ic
es.
Mo
re
ov
e
r,
it
is
al
so
requirin
g
cent
r
al
iz
ed
resour
ce
s
fo
r
to
gen
e
rate
the
ir
need
of
pr
oc
essing
powe
r.
Af
te
r
ward,
th
e
exam
inati
on
of
se
rv
ic
e
deli
ver
a
nce
str
uctur
e
is
us
e
d
to
con
qu
e
r
a
dif
ficult
y t
hro
ugh t
he
em
plo
y o
f
se
r
vice popula
ti
ng pr
oc
edure a
nd Cl
ou
d
se
rv
ic
es
.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
ERMO
2 alg
or
i
thm:
an en
er
gy
eff
ic
ie
nt m
obil
it
y m
anage
men
t i
n
m
obil
e clo
ud
co
m
pu
ti
ng
… (
L.
Pall
avi
)
1959
Qi
et
al
.
[
22
]
ha
ve
antic
ipate
d
a
serv
ic
e
-
res
pons
i
ve
po
sit
io
n
m
et
ho
d
wh
ic
h
is
e
m
plo
ys
to
identify
the
occurre
nce
an
d
po
sit
ion
o
f
m
ob
il
e
dev
ic
e
de
void
of
co
n
ve
ntion
al
cy
cl
ic
regi
strat
ion
updat
e.
In
te
r
net
pro
tocol
(I
P
)
in
m
ultimed
ia
subsyst
em
(I
MS)
is
m
ai
nly
us
ed
to
acq
uire
the
po
sit
io
n
an
d
ci
rc
um
st
ances
in
form
ation
of
m
ob
il
e
dev
ic
es
by
the
hel
p
of
re
gistrati
on
pr
ogressi
on.
H
er
e,
a
bindin
g
w
as
f
or
m
ed
by
m
eans
of
se
rv
i
ng
-
cal
l
session
c
on
t
ro
l
f
un
ct
io
n
(S
-
C
SCF)
w
hich
is
de
rive
d
from
pr
el
im
inary
re
gistrati
on
am
on
g
pu
blic
co
nsum
er
identit
y
an
d
I
P
ad
dr
ess
of
m
ob
il
e
de
vice.
T
he
S
-
CSC
F
an
d
pro
xy
cal
l
ses
sion
co
ntr
ol
f
unct
ion
(
P
-
CSC
F)
is
gen
e
rall
y
us
ed
to
m
a
intai
n
the
c
on
s
um
er
reg
i
strat
ion
po
sit
io
n
li
ke
the
tim
e
r
w
hich
is
sp
ec
ify
ing
the
‘e
xp
ires’
lim
it
at
ion
.
Gan
i
et
al
.
[
23
]
hav
e
antic
ipa
te
d
a
them
atic
ta
xono
m
y
wh
ic
h
is
us
e
d
f
or
t
he
cl
assifi
cat
io
n
of
existi
ng
interw
orkin
g
a
nd
m
ob
il
it
y
pr
ocedu
re
of
wir
el
ess
data
netw
orks.
Her
e
,
the
propositi
on
a
nd
sig
nificant
fe
at
ur
es
of
e
xisti
ng
fa
ultl
ess
connect
ivit
y
pr
oce
dur
es
are
exam
ined
as
ef
fecti
ve
ly
.
In
hete
rogen
e
ous
wi
reless
data
netw
orks,
t
he
consi
der
e
d
t
wo
si
gnific
ant
un
it
s
of
fau
l
tl
ess
connecti
vity
are
inter
work
i
ng
an
d
m
ob
il
i
ty
proce
dures.
H
ere,
the
inte
r
work
i
ng
proce
dure
is
us
e
d
to
inc
orporate
var
i
ou
s
wirele
ss
data
netw
orks
for
dissim
il
ar
OS
I
la
ye
rs
wh
ic
h
are
facil
it
at
in
g
node
m
ob
il
it
y
in
hetero
ge
neous
wireless
data
netw
ork
s
for
retai
ning
c
onne
ct
ivit
y.
In
the p
r
ogressi
on o
f
m
ob
il
i
ty
,
nu
m
erous
proce
dure
s
are
em
plo
yi
ng
p
r
oacti
ve
m
ob
il
it
y
to
cal
culat
e
the
po
ssi
ble
reli
nquis
h
f
or
to
e
va
de
pac
ka
ge
fa
il
ur
e
an
d
ass
oc
ia
ti
on
fail
ur
e
.
More
ov
e
r,
the
furthe
r
proce
dures
are
of
fe
rin
g
reacti
ve
m
ob
il
ity
fo
r
le
sser
sign
al
in
g
trans
par
e
ncy
handove
r
.
Th
e
fo
rem
os
t
intenti
on
of
m
et
hod
is
t
o
dim
inish
the
la
te
ncy,
pa
ck
age
fail
ur
e,
a
nd
m
ultim
od
al
.
Her
e
,
Q
oS
is
m
ai
ntaining
si
gn
al
in
g
trans
par
e
ncy a
nd m
ob
il
it
y fo
r
the
op
ti
m
iz
ati
on d
i
ff
ic
ulty
.
Chen
et
al
.
[
24
]
ha
ve
antic
ipate
d
m
ob
il
i
ty
-
dr
ive
n
net
w
ork
(MD
N
)
w
hich
is
der
i
ve
d
from
the
exam
inati
on
of
dev
el
op
m
ent
ph
a
ses.
Her
e
,
the
m
ob
il
it
y
or
gan
iz
at
io
n
is
c
la
ssifie
d
as
th
r
ee
segm
ents
su
ch
a
s
on
e
at
trib
ute
in
a
preci
se
net
work,
m
ob
il
it
y
in
a
c
omm
on
com
petence
of
netw
orks
an
d
m
ob
il
i
ty
.
Su
pp
os
e,
i
f
the
m
ob
il
it
y
is
su
pp
or
ti
ng
t
he
env
i
ronm
ent
then
ob
ta
in
able
m
ob
il
i
ty
te
chnolo
gy
is
il
lustr
at
ing
th
e
prob
l
e
m
of
ta
sk
re
dunda
nc
y,
syst
e
m
com
plica
ti
on
,
ina
de
qu
acy
,
a
nd
sc
al
abili
ty
.
MDN
is
m
ai
nly
em
plo
ys
the
de
sign
of
view
po
i
nt,
ta
s
k
ori
entat
io
n
r
epr
ese
ntati
on,
and
prot
oco
l
ori
e
ntati
on
re
presentat
ion
f
or
a
high
-
le
vel
co
ncep
t
of
key task
, net
w
ork
e
ntit
ie
s,
an
d
sci
e
ntific
v
al
ues.
Hu
et
al
.
[
25
]
hav
e
m
ai
nly
fo
cusin
g
the
m
ob
il
it
y
pr
oble
m
in
the
cl
oud
da
ta
center
w
hich
is
us
e
d
t
o
conve
ne
div
er
se
necessit
ie
s
of
In
te
rn
et
se
r
vices
to
s
up
pl
y
enh
a
nce
d
co
ns
um
er
ex
per
i
ence.
Ge
ner
al
ly
,
the
virtu
al
m
achines
are
com
pet
ent
to
transfer
fr
om
on
e
po
sit
ion
to
ano
t
her.
The
f
or
e
m
os
t
diff
ic
ult
y
of
IP
m
ob
il
i
ty
was
a
relocati
on
of vi
rtual
m
achine (
VM)
in
IP
s
ubnets. I
t
is dem
on
strat
in
g
t
he
m
ob
il
it
y
-
base
d
c
loud
data
center
ne
twork
at
str
uctu
ral
desig
n
wh
ic
h
is
de
rive
d
from
the
desig
n
of
MDN.
Her
e
,
the
Entit
y
-
identit
y/
Locati
on
-
i
den
t
ifie
r
is
decoup
l
ing
to
est
ablis
h
a
po
sit
ion
e
xe
cutive
to
pla
n
the
Entit
y
-
ide
ntit
y
a
t
the
e
xisti
ng
po
sit
ion
of
VM.
More
ov
e
r,
the
m
app
in
g
ta
ble
is
us
e
d
t
o
rear
range
t
he
VM
for
fau
lt
le
ss
se
rv
ic
e
wh
ic
h
is
offer
e
d by m
eans o
f
cl
oud data ce
nter.
Ju
ni
or
et
al
.
[
26]
ha
ve
offere
d
the
m
ob
il
e
offloa
ding
syst
e
m
(MOSys
)
w
hich
is
ge
ne
rall
y
e
m
plo
ys
to
retai
n
an
d
m
a
i
ntain
the
f
un
c
ti
on
of
m
ob
il
e
nodes.
T
he
ba
sic
syst
e
m
st
ru
ct
ur
al
desi
gn
is
m
ai
nly
us
ed
to
su
pp
or
t
the
sof
tware
def
ine
d
netw
ork
(
SDN
)
f
or
t
he
m
ob
il
it
y
or
ga
nizat
ion
ta
sk.
The
op
erati
on
al
sit
uat
ion
of
cache
relat
ed
orga
nizat
ion
sys
tem
is
reli
ant
on
the
off
-
loa
d
reacti
on
ti
m
e.
The
progressi
on
of
m
idd
le
w
are
is
e
m
plo
ys
to
pr
eserv
e
t
he
dat
a
off
-
l
oad,
pr
of
il
in
g
ser
vice
s,
cl
oud
detec
ti
on
,
a
nd
a
ppli
cat
ion
co
nsu
m
pt
ion
.
The
syst
em
was
est
i
m
at
ed
by
m
eans
of
m
ob
il
i
ty
i
m
pa
ct
li
m
it
ation
on
the
off
-
l
oad
pr
ese
ntati
on
at
the
i
m
ple
m
entat
io
n
arr
a
ngem
ent
in
the
su
bse
qu
ent
producti
on
m
ob
il
e
dev
ic
es.
The
MOSys m
ob
il
i
ty
or
gani
zat
ion
syst
e
m
is
m
a
i
nly
us
ed
to
c
on
t
ro
l
the
el
e
vated
den
sit
y
of
m
ob
il
e
nodes
thr
ough
t
he
colle
ct
ion
of
lo
w
exp
e
ndi
ture
and e
nergy c
om
petent co
nversi
on
from
o
ne pos
it
ion
to
a
nothe
r
posit
ion.
Kim
et
al
.
[27]
ha
ve
offere
d
a
tra
ff
ic
orga
nizat
ion
pr
ocess
by
m
ea
ns
of
m
ob
il
e
edg
e
cl
oud.
The
m
ob
il
e
ed
ge
cl
ou
d
is
ge
ne
rall
y
sit
uated
in
the
m
ob
il
e
edg
e
net
work
f
or
to
obser
ve
th
e
posit
ion
of
m
ob
il
e
te
rm
inals.
The
r
efore,
it
tur
ns
ou
t
to
be
pro
ba
ble
f
or
c
on
t
ro
l
the
traf
fic
of
m
ob
il
e
te
rm
inals
eff
ic
ie
ntly
thr
ough
the
net
wor
k.
T
his
pr
ocess
is
norm
al
l
y
con
tr
ols
the
vide
o
t
raffic
f
ro
m
co
ntent
se
rv
e
r
i
n
the
I
nter
net
w
hich
is
base
d
on
t
he
e
dg
e
net
work
posit
ion
a
nd
m
ob
il
it
y
of
a
m
obil
e
te
r
m
inal.
More
over,
it
offe
rs
the
vid
e
o
t
ra
ff
ic
to
the
m
ob
il
e
te
rm
inals.
This
pr
oces
s
is
e
m
pl
oying
the
m
obil
e
edg
e
cl
oud
fo
r
to
de
vel
op
the
Q
oE
of
m
ob
il
e
vid
e
o
c
onsu
m
er.
Seo
et
al
.
[
28]
hav
e
a
ntici
pated
the
c
lo
ud
c
omm
un
ic
at
ion
s
for
ubiq
uito
us
com
pu
ti
ng
e
nv
i
ronm
ent
m
ob
il
e
app
li
cat
ion
(CI
-
UCE
MA),
w
hich
e
nco
m
passes
t
hree
la
ye
rs
li
ke
cl
oud
ser
vice
la
ye
r
(CS
L),
M2M
serv
ic
e
la
ye
r
(
MSL)
an
d
ubi
qu
it
ous
se
rv
ic
e
la
ye
r
(U
SL).
The
M2M
is
including
I
oT
se
rv
ic
es
la
ye
r
(M
SL)
f
or
to
dim
inish
th
e
com
plica
t
ion
of
bo
t
h
the
enh
a
ncem
ent
and
m
ai
ntenan
ce
of
Io
T
s
yst
e
m
s.
It
is
m
ai
nly
dem
on
strat
ed
by
m
eans
of
e
nh
a
nce
d
inter
opera
bili
ty
and
the
em
plo
y
of
sta
nd
a
rd
proce
ss.
Prote
ct
io
n
is
al
so
accom
plished
by
m
eans
of
de
vice
m
anag
e
m
ent
and
pr
ovisi
on
in
g.
The
MSL
was
de
pe
nd
a
ble
for
preven
ti
ng
the
con
s
um
er
fr
om
fu
ndam
ental
co
m
plica
ti
o
n
an
d
inco
ns
ist
ency
in
sel
f
-
m
od
i
ficat
ion
at
m
os
phe
re
by
the
help
of m
ob
il
it
y and
a
dap
ta
ti
on at
CDPS
a
nd CI
MS.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
3
,
June
201
9
:
1957
-
1967
1960
3.
PROBLE
M
METHO
DOL
OGY
A
N
D
S
YS
TE
M
MO
DEL
3.1. Pr
ob
le
m
m
eth
odolo
gy
Aissio
ui
et
al
.
[29]
ha
ve
an
ti
ci
pated
an
el
ast
ic
m
et
ho
d
wh
ic
h
is
de
riv
ed
f
ro
m
SD
N/
Op
e
n
-
Flo
w
structu
ral
de
sign
a
nd
a
c
on
t
ro
l
plane
st
ru
c
ture.
They
a
re
generall
y
m
o
dified
f
or
m
obil
e
cl
ou
d
c
ompu
ti
ng
syst
e
m
s
and
Fo
ll
ow
Me
Cl
oud
(F
MC
)
-
rela
te
d
syst
e
m
s
.
T
he
ci
rcu
m
sta
nce
of
5G
m
ob
il
e
networ
ks
are
m
ai
nly
e
m
plo
ys
to
m
anag
e
the
pla
ne
struct
ur
e
f
or
MC
C
-
relat
ed
s
yst
e
m
s
wh
ere
m
ob
il
e
no
de
s
and
net
wor
k
s
erv
ic
es
are
in
dicat
ing
lim
i
ta
ti
on
of
act
ivit
ie
s
and
relocati
ons.
M
or
e
over,
the
di
ssi
m
i
la
rity
of
centrali
zed
m
e
thod
thr
ough
so
li
ta
r
y
SD
N
re
gula
tor
is
facil
it
at
ing
to
al
locat
e
the
SDN/O
pe
n
-
Flow
c
on
t
ro
l
pl
ane
on
a
two
-
le
vel
hierar
c
hical
str
uctu
ral
desig
n
wh
ic
h
are
co
ntainin
g
first
s
ta
ge
am
on
g
a
global
regulat
or
of
G
-
FMC
C
and
seco
nd
sta
ge
a
m
on
g
num
ero
us
l
ocal
re
gula
to
r.
T
he
ass
essm
ent
con
sequ
e
nces
a
re
a
cqu
i
red
by
m
eans
of
inv
est
igati
on.
Additi
on
al
ly
,
this
ex
planati
on
is
m
aking
su
re
t
he
en
ha
nced
c
ontr
ol
plane
orga
nizat
io
n,
pr
ese
ntati
on
m
ai
ntenance
,
a
nd n
et
w
ork
res
ource
co
ns
e
rv
at
i
on.
MC
C
is
a
c
onve
rg
e
nt
te
ch
no
l
og
y com
pr
i
sed
of
three
co
r
ner
st
one
heter
ogene
ous
te
ch
no
l
og
ie
s,
nam
el
y
m
ob
il
e
co
m
pu
ti
ng,
cl
oud
com
pu
ti
ng,
an
d
net
wor
king.
The
upc
om
ing
heteroge
neou
s
5G
netw
ork
e
m
ph
asi
zes
on
a
dr
am
at
ic
i
ncr
ease
i
n
the
transm
issi
on
rate
of
MC
C
traff
ic
.
W
it
h
m
or
e
us
e
rs
ope
rati
ng
at
hig
h
rates
,
the
ty
pe
of
data
s
ha
red
over
t
he
ne
twork
is
goin
g
to
be
com
plex
and
a
m
ajo
rity
of
it
will
include
vid
eo
tra
ff
ic
.
Such
com
plex
str
uctu
re
of
traf
fic
and
hea
vy
lo
ad
ov
e
r
the
com
po
ne
nts
of
the
net
wor
k
are
dif
ficult
to
co
ntr
ol.
Further
,
the
m
ob
il
i
ty
of
us
ers
a
dds
up
to
this
iss
ue
an
d
m
akes
it
diff
ic
ult
to
m
anag
e
and
operate
t
he
netw
ork
with
ou
t
a
ny
brea
kdown.
Th
us,
it
is
i
m
po
rta
nt
to
con
t
ro
l
traff
ic
as
w
el
l
as
m
anag
e
the
m
ob
il
i
ty
of
use
rs
to
pr
ov
i
de
eff
ic
ie
nt
com
m
un
ic
at
io
n,
wh
i
ch
can
s
upport
vid
e
o
traff
i
c
at
high
delivery
rates.
Additi
on
al
ly
,
t
he
basic
par
am
et
ers
af
fect
the
perf
or
m
ance
of
net
w
ork
as
t
he
vas
t
var
ie
ty
of
m
ob
il
e
dev
ic
es
with
dif
fer
e
nt
op
e
rati
ng
syst
em
s
(O
S
s),
platfo
r
m
s,
and
wirele
ss
netw
ork
sta
nd
a
r
ds
.
The
upc
om
ing
5G
netw
orks
aim
at
pr
ov
i
din
g
hi
gh
s
pe
ed
com
m
un
ic
at
ion
s
to
us
er
s
irresp
ect
i
ve
of
their
m
ov
e
m
ent.
W
i
th
an
inc
rease
in
the
num
ber
of
de
vices
a
nd
the
netw
ork
a
tt
ai
nin
g
it
s
pea
k
siz
e,
due
to
dense
dep
l
oym
ent,
i
t
beco
m
es
im
po
rtant
to
m
anag
e
a
nd
con
t
ro
l
m
ob
il
it
y
fo
r
e
ff
ic
ie
nt
com
m
un
ic
at
ion
.
Mob
il
it
y
m
anag
em
ent
requi
res
m
ulti
ple
operati
ons
at
the
sam
e
insta
nce,
w
hich
in
cl
ud
e
opti
m
a
l
r
ou
t
e
sel
ect
ion
, m
ob
i
le
an
ch
or s
upport,
u
se
r patt
er
n
ide
ntific
at
ion, an
d ser
vice ha
ndoffs
.
An
e
ne
rg
y
e
ff
i
ci
ent
m
ob
il
ity
m
anag
em
ent
i
n
m
ob
il
e
cl
oud
com
pu
ti
ng
(
E2
M2MC
2)
sy
stem
util
iz
es
the
opti
m
a
l
clo
ud
sel
ect
io
n
to
obta
in
e
nergy
ef
fici
ent
m
ob
il
it
y
m
anage
m
ent.
The
m
ai
n
co
ntri
bu
ti
ons
of
pro
po
se
d E2M
2MC2 syste
m
are
s
umm
arized as
foll
ows:
a.
In
E
2M
2MC2
syst
e
m
,
an
el
ect
ive
rep
eat
m
ulti
-
obj
ect
ive
opti
m
iz
a
ti
on
(ERM
O
2
)
a
lgor
it
hm
us
ed
to
com
pu
te
the
be
st cl
oud
am
ong othe
rs
i
n
the
net
work.
b.
Var
i
ou
s
par
a
m
et
ers
us
ed
for
best
cl
ou
d
el
ect
ion
pr
ocess
are:
de
la
y,
j
it
te
r,
BER
,
pack
et
loss,
com
m
un
ic
at
ion
c
os
t,
res
ponse
tim
e, an
d net
work loa
d.
c.
Finall
y,
the
per
f
or
m
ance
of
pro
po
se
d
E
2M2
MC
2
syst
em
is
co
m
par
ed
with
the
existi
ng
si
m
ulation
resu
lt
s
sho
ws
that
the
propo
sed
E2M
2MC
2
syst
e
m
helps
in
m
ini
m
iz
i
ng
hand
ov
e
r
de
la
y,
pack
et
loss
,
aver
a
ge q
ueu
i
ng
delay
, and de
vice li
fetim
e in a
hete
roge
neous netw
ork.
3.2. S
ys
te
m m
od
el
o
f
p
r
opos
ed E2
M
2
M
C
2 s
ystem
The
syst
em
m
od
el
of
propos
ed
E
2M2
MC
2
syst
e
m
is
sh
ows
i
n
Fi
gure
1,
w
hich
use
s
t
he
fo
ll
ow
m
e
cl
oud
(F
MC
)
con
ce
pt,
w
hic
h
al
lo
ws
t
he
relocati
on
of
serv
ic
es
pro
vi
ded
to
us
e
rs
dep
e
ndin
g
on
their
m
ov
e
m
ents.
Serv
ic
es
a
re
there
fore
al
ways
pr
ov
i
ded
f
r
om
da
ta
center
locat
ion
s
t
hat
are o
pt
i
m
a
l
fo
r
the
c
urre
nt
locat
ion
s
of
t
he
us
e
rs.
T
his p
r
ov
i
des
us
e
rs
w
it
h
im
pr
ov
e
d
Q
oS
/Q
oE
,
at
t
he
sam
e
t
i
m
e
and
it
al
lows
p
re
ser
ving
op
e
rato
rs’
net
work
res
ources
by
escapin
g
ne
twork
tra
ff
ic
t
o
data
ce
nters
t
hro
ugh
the
nea
rest
p
oin
ts
co
m
par
ed
with
us
e
rs’
l
oc
at
ion
s.
Anothe
r
ad
van
ta
ge
of
FMC
te
chn
ol
og
y
is
that
m
i
gr
at
io
n
of
ser
vi
ces
is
sea
m
le
s
s
and
trans
par
e
nt
to
us
ers
.
MC
C
fr
a
m
ewo
r
k
us
e
bo
t
h
the
inf
orm
at
ion
stock
pili
ng
an
d
the
inf
or
m
at
ion
ha
nd
li
ng
happe
n
ou
tsi
de
of
the
cel
l
ph
on
e
.
C
on
ce
rn
i
ng
de
finiti
on,
m
ob
il
e
app
li
cat
ion
s
m
ov
e
t
he
proces
sin
g
f
or
ce
an
d
capaci
ty
fr
om
t
he
cel
l
te
le
ph
ones
to
the
Cl
oud.
It
m
igh
t
be
tho
ug
ht
the
co
ns
oli
dation
of
t
he
cl
oud
c
om
pu
ti
ng
and m
ob
il
e en
vir
on
m
ent.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
ERMO
2 alg
or
i
thm:
an en
er
gy
eff
ic
ie
nt m
obil
it
y m
anage
men
t i
n
m
obil
e clo
ud
co
m
pu
ti
ng
… (
L.
Pall
avi
)
1961
Figure
1
.
Syst
em
m
od
el
o
f propose
d
E
2M
2M
C2 syst
em
4.
ENERG
Y
E
FFICIE
NT
MOBILIT
Y
MAN
AGE
M
ENT
I
N
MO
BIL
E
CLO
U
D
COMP
UTING
(E2M2
M
C2) SYSTE
M
The
m
ov
em
ent
is
te
naciou
s
and
cha
nged
cha
ng
e
i
nf
l
ue
nced
by
the
creatu
re'
s
own
pa
rtic
ula
r
locom
oto
rs
at
t
e
m
pts
passing
on
t
hem
to
new
ty
pical
cond
it
ion
.
Creat
ur
e
m
ov
e
m
ent
is
t
he
fairly
lo
ng
-
confine
change
of
pe
ople
,
by
an
d
la
rg
e
ta
lking
on
a
ty
pical
beg
ins.
It
is
an
inevita
ble
ponder
th
at
can
be
found
in
al
l
enorm
ou
s
crea
ture
par
ti
es,
for
exam
ple,
bir
ds
,
m
a
m
m
al
s,
fish,
re
ptil
es,
am
ph
ibians,
in
s
ect
s,
an
d
cr
us
t
acea
ns.
The
m
ulti
-
ob
j
e
ct
ive
op
ti
m
iz
a
t
ion
(M
OO)
al
gorithm
[3
0]
can
be
div
ide
d
i
nto
Dias
pora
proces
s
an
d
upda
ti
ng
process
.
Th
e
a
nim
a
l
Diaspor
a
process
c
ompu
te
s
t
he
gr
oups
of
a
nim
al
mo
ve
from
on
e
posit
ion
t
o
ne
xt
and
t
he
anim
a
l
up
da
te
process
c
ompu
te
s
posit
ion
change
d
ani
m
al
s
are
upda
te
d
by
the
pro
bab
il
ist
ic
m
et
ho
d.
MOO
al
gorith
m
with
the
Diasp
or
a
proc
ess
and
popul
at
ion
upda
ti
ng
process
are
us
e
d
to
com
pu
te
a
sat
isfact
or
y
s
olu
ti
on.
T
he
pro
po
s
ed
el
ect
ive
rep
eat
m
ulti
-
ob
j
ect
iv
e
opti
m
i
zat
ion
(
ERM
O
2
)
al
gorithm
insp
ir
e
d
from
con
ve
nt
i
on
al
M
O
O
al
gorithm
[30]
wi
th
a
new
m
igrati
on
process
by
est
ablishi
ng
a
li
ving
a
rea
by
the
best
fitness
va
lue
owne
d
a
nim
al
and
anim
al
s
m
igrate
from
cur
ren
t
lo
cat
ion
s
int
o
th
is
new
li
ving
area
to
si
m
ulate
ani
m
al
Diasp
or
a
pr
ocess.
W
e
co
nsi
der
N
a
nim
als
that
li
ve
in
li
vin
g
area
,
s
om
e
ind
ivid
uals
m
ov
e
rand
om
l
y
and
their
posit
io
n
updated
,
a
nd
t
hen
we
cal
c
ulate
the
best
po
s
it
ion
of
anim
als
by
fitnes
s
f
unct
io
n
and
rec
ord
it
.
But
the
am
ount
of
f
ood
or
w
at
er
grad
ually
dim
inished
as
the
tim
e
wore
on,
a
nd
so
m
e
anim
a
ls
m
igrate from
t
he
c
urren
t a
rea
s which
h
a
ve n
o foo
d
a
nd w
at
er to a
new are
a w
it
h
a
bunda
nt
f
oo
d
a
nd w
at
er.
ERM
O
2
al
gorithm
beg
ins
wi
th
an
init
ia
li
zat
ion
proc
ess,
f
or
this
wor
k,
we
co
ns
ide
r
th
e
anim
a
l
as
sens
or
nodes
and
t
heir
posi
ti
on
as
net
wor
k
posit
ion.
Le
t
set
of
N
se
ns
or
no
des
a
nd
their
po
sit
io
ns
are
N
P
P
P
P
,
,
,
,
3
2
1
;
each
a
nim
al
posit
ion
i
P
is
a
d
n
1
-
dim
ension
al
vecto
r,
w
here
n
the
nu
m
ber
of
cl
us
te
rs
an
d
d
is
the
di
m
ension
of
the
te
st
set
.
The
cl
us
te
rs
d
i
i
i
i
p
p
p
P
2
1
,
*
,
wh
e
re
n
i
,
,
2
,
1
;
each
cl
us
te
r
is
1
×
D
-
dim
ension
al
vector,
an
d
th
e
lower
bound
of
the
cente
rs
is
the
m
ini
m
um
of
each
col
um
n
in
te
st
set
d
n
t
,d
e
no
te
s
d
l
t
t
t
t
,
,
,
m
i
n
2
1
,
and
the
upper
bound
of
the
centers
is
n
u
t
t
t
t
,
,
,
m
a
x
2
1
.
Senso
r
node
s
are
rand
om
ly
and
un
i
form
l
y
distribu
te
d
be
tween
the
pr
e
-
sp
eci
fied
lowe
r
init
ia
l
p
aram
et
er
bo
und
l
an
d
the
uppe
r
init
ia
l
par
am
et
er
bo
und
u
.
On
ce
the
nei
ghbo
rho
od
to
pology
has
been
co
ns
t
ru
ct
e
d,
we
sel
e
ct
on
e
nei
ghbo
r
ra
ndom
ly
and
update
t
he
pos
it
ion
of
the
in
di
vid
ual
acc
ordi
ng
to
this nei
ghbor
a
s foll
ow
s:
j
j
j
l
u
j
i
l
j
i
t
t
r
t
p
1
,
0
,
,
(1)
wh
e
re
n
i
,
,
2
,
1
and
d
n
j
,
,
2
,
1
,
1
,
0
,
j
i
r
rep
rese
nts
the
rand
om
n
um
ber
s
betwe
en
0
to
1.
Durin
g
the
Dia
sp
ora
process, becau
se of no
de
s m
ov
e
m
ent i
n
the
netw
ork a
rea,
s
om
e p
arts o
f
t
he
net
wor
ks
are
la
cking
du
e
to
the
c
onditi
on
cha
nge,
a
nd
so
m
e
nodes
m
ov
i
ng
f
r
om
the
cu
rr
e
nt
posit
i
on
to
ne
w
po
sit
ion
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
3
,
June
201
9
:
1957
-
1967
1962
dep
e
nds
on
m
ov
in
g
act
io
n.
D
ur
i
ng
t
he
updat
ing
process
,
th
e
al
gorithm
com
pu
te
s
how
node
s
le
ave
t
he
cl
us
te
r
and
s
om
e
j
oin
in
the
new
po
pula
ti
on
. I
ndivi
du
al
s
will
be
r
eplace
d
by
som
e
new
nodes
with
a
prob
a
bili
ty
n
e
w
P
.Th
e
pr
obabili
ty
is
us
ed
acc
ordi
ng
t
o
the
qual
it
y
of
the
fitness.
We
s
or
t
fitness
in
des
cend
i
ng
orde
r,
so
th
e
pro
bab
il
it
y
of
the
ind
i
vidual
with
best
fitne
ss
is
1/N
an
d
the
ind
i
vidual
with
w
or
st
fitn
ess,
by
co
ntras
t,
is
1.
Af
te
r
c
om
pu
te
the
new
s
ol
ution
1
,
j
i
p
,
it
will
be
evaluated
a
nd
c
om
par
ed
wit
h
the
j
i
p
,
and
we
c
hoose
the
ind
ivi
du
al
with
a b
et
te
r
ob
j
ect
ive f
it
nes
s a
nd
denotes as
foll
ow
s:
O
t
h
e
r
w
i
s
e
;
1
,
t
h
a
n
b
e
t
t
e
r
is
,
if
;
1
,
,
j
i
j
i
i
p
j
i
f
j
i
f
p
P
(2)
In this
wor
k,
w
e u
se
Rast
rigin’s fu
nctio
n
f
or
fitness c
om
pu
ta
t
ion
a
nd d
e
fin
e as foll
ows:
n
i
i
p
j
i
f
1
i
2
10
p
2
c
os
10
,
(3)
wh
e
re
the
ra
nge
var
ia
bles
de
note
s
as
5
.
1
2
p
5
.
1
2
i
.
Ra
strigin'
s
functi
on
i
s
base
d
on
c
os
i
ne
m
od
ulati
on
to
produce
m
a
ny
local
m
ini
ma.
The
pa
ram
eter
s
for
cl
oud
s
el
ect
ion
are:
de
la
y
(D
),
j
it
te
r
(J)
,
BER
,
packet
loss
(P
L
),
c
omm
un
ic
at
ion
cost
(
C),
a
nd
netw
ork
l
oad
(
nL).
The
par
am
et
ers
are
ti
m
e
var
yi
ng
f
unct
ion
and
it
op
ti
m
iz
ed
by the
pro
po
se
d E
RM
O
2
al
gorith
m
. Th
e sele
ct
ion
c
rite
rion for
best
op
ti
m
al
cl
oud
a
s foll
ows,
nL
C
pL
B
E
R
J
D
,
,
,
,
,
m
i
n
c
l
o
u
d
B
e
s
t
(4)
Delay
is d
e
fine
d
as
the a
ver
a
ge
of the
su
m
o
f
the
distances
of all
the m
ob
il
e u
se
rs fr
om
their select
ed
cl
oud.
n
i
n
c
l
ou
d
B
e
s
t
m
di
s
D
1
,
(5)
Av
e
ra
ge
sin
k d
ist
ance is d
e
fin
ed
as
the
rati
o of dista
nce
b
et
ween I
DMD
to
the
DCG
as
fol
lows
,
D
C
G
c
l
o
u
d
B
e
s
t
d
i
s
D
A
v
g
d
,
(6)
All
m
ob
il
e
us
ers
co
nsum
e
so
m
e
ener
gy
due
to
their
proce
ssing
ti
m
e
and
to
con
s
um
e
l
ess
energy.
The
a
ve
rag
e
c
omm
un
ic
at
ion
dis
ta
nce
an
d
DCG
is
cl
os
e
to
al
l
the
m
obil
e
us
ers
.
T
o
m
ini
m
iz
e
the
aver
a
ge
delay
and
DC
G dist
ance
of a
ll
the Best cl
ou
d by opti
m
al
se
le
ct
ion
a
ppro
ac
h
i.e.
n
i
d
D
A
v
g
D
T
1
(7)
Si
m
ply,
j
it
te
r
i
s
def
ine
d
as
the
var
ia
ti
on
in
delay
of
recei
ved
pack
et
s.
D
u
e
to
netw
ork
congesti
on,
i
m
pr
oper
que
ui
ng
,
or
c
onfi
gurati
on
e
rro
rs,
this
ste
a
dy
stream
can
bec
om
e
delay
betw
een
eac
h
pac
ke
t
can
var
y
i
ns
te
ad
of
rem
ai
nin
g
c
onsta
nt.
Let
us
c
onside
r
tw
o
c
on
secuti
ve
pac
ke
ts
of
the
ta
gge
d
tra
ff
ic
,
0
P
and
1
P
.
Fo
r
pac
ket
1
,
0
,
j
P
j
an
d
node
k
= 1
,
...,
n,
le
t
τ
in
j
(
k)
a
nd
τ o
ut j
(
k)
b
e
the
arr
i
val
and
dep
a
rtu
re
tim
e
s
of
j
p
at
n
ode
k, le
t
k
w
j
be
the
wait
ing
ti
m
e o
f
j
p
at
nod
e
k,
a
nd f
i
nally
let
k
w
k
w
k
0
1
be
the
v
a
riat
ion
of
the inter
-
pa
cke
t delay
at n
ode
k.
The
end
-
to
-
e
nd j
it
te
r
is
de
fined as,
n
i
k
E
n
J
1
(8)
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
ERMO
2 alg
or
i
thm:
an en
er
gy
eff
ic
ie
nt m
obil
it
y m
anage
men
t i
n
m
obil
e clo
ud
co
m
pu
ti
ng
… (
L.
Pall
avi
)
1963
En
d
-
to
-
e
nd
j
it
te
r
is
the
exp
ect
ed
abs
ol
ute
value
of
th
e
su
m
of
inter
-
pack
et
dela
y
var
ia
ti
on
s
introd
uced
by
each
m
ob
il
e
us
er
al
ong
the
path
bet
ween
t
he
source
a
nd
the
destinat
io
n.
The
bit
error
rati
o
(BER)
is
the
nu
m
ber
of
bit
error
s
div
i
de
d
by
the
t
otal
nu
m
ber
of
tr
ansf
e
rr
e
d
bits
durin
g
a
stu
di
ed
tim
e
interval.
Bi
t
err
or
rati
o
is
a
unit
le
ss
per
f
orm
ance
m
easure
,
of
te
n
ex
pr
es
sed
as
a
per
ce
ntage.
Pac
ket
loss
(PL
)
ref
e
rs
to
the
num
ber
of
packet
s
lost
to
reach
the
desti
nation
w
he
n
bei
ng
transm
it
te
d
acro
s
s
the
net
wor
k.
It
is
one
of
th
e
i
m
po
rtant
Q
oS
pa
ram
et
er
for
real
ti
m
e
app
li
cat
ion
su
c
h
a
s
cl
ou
d
gam
ing
as
of
a
case
in
our
analy
sis.
The
com
m
un
ic
at
ion
cost
(C
)
is
su
m
of
the
co
st
of
al
l
the
ta
sk
s
im
ple
m
e
nting
t
hat
al
gorithm
.
We
s
hall
f
ocus
on
the
co
st
as
the
way
t
o
m
e
asur
e
the
e
ff
ic
i
ency
of
al
gorithm
and
the
am
ount
of
ti
m
e
it
ta
kes
each tas
k
to
exe
cute w
he
n
est
i
m
ating
t
he
r
un
ning ti
m
e o
f
an
algorit
hm
.
The net
w
ork
lo
ad
(
nL
)
is t
he wei
ghte
d w
ho
l
e o
f
whate
ve
r
i
s left o
f
the l
oa
d
of i
nd
i
vidual users o
f
th
e
baffli
ng
num
ber
of
us
e
rs.
Th
e
strag
glin
g
st
ay
s
of
the
l
oad
of
in
div
i
du
al
us
er
is
de
picte
d
as
t
he
m
ade
ho
l
ding
up
crit
ic
al
ness
out
of
t
he
use
r.
By
uprig
ht
ness
of
the
m
ob
il
e
us
er
s,
t
he
bundle
m
ight
be
retransm
i
t
te
d
to
ens
ur
e
the
ri
ght
trans
port.
T
he
crit
ic
al
ness
of
each
i
ns
id
e
and
thei
r
co
nd
it
io
n
in
li
ke
way
us
ed
t
o
pick
rem
ai
nin
g
load
of
the
e
ntir
e
us
er
.
We
co
ns
ide
r
the
stra
gg
li
ng
bits
of
the
load
of
the
entire
m
ob
il
e
us
er
regulat
e
as
th
e aggre
gate
of
t
he
weig
hted
ext
ra
li
feti
m
e
of
al
l
us
ers
in
t
he
ne
twork
.
T
hus,
t
he
nL
of
the
m
ob
il
e
us
ers
as
fo
ll
ows:
n
i
d
k
a
A
v
g
D
T
J
W
nL
1
m
a
x
(
9
)
wh
e
re
a
W
represe
nt
s the
weig
ht f
a
ct
or
of each
m
ob
il
e
us
er
and
represe
nted by,
d
a
A
v
g
c
W
1
(1
0
)
wh
e
re c
r
e
pr
es
ents the
consta
nt.
The w
orkin
g flow o
f
e
ntire
propose
d
E
2M
2M
C2 syste
m
i
l
lustrate
s in
Fig
ur
e
2.
Figure
2
.
Wo
r
ki
ng
flo
w of p
r
opose
d E2M
2M
C2 syste
m
clou
d sel
ect
ion us
ing
ERM
O
2
al
gorithm
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
3
,
June
201
9
:
1957
-
1967
1964
5.
PERFO
R
MANC
E
A
NA
L
Y
SIS
The
m
ob
il
e
clo
ud
c
om
pu
ti
ng
arc
hitec
tures
is
analy
zed
f
or
the
perf
or
m
ance
point
of
view.
M
ob
il
e
cl
oud
arc
hitec
ture
i
ncr
eases
r
eso
ur
ce
a
vaila
bili
ty
by
le
ver
agin
g
la
r
ge
nu
m
ber
of
nea
r
by
m
ob
il
e
dev
i
ces
in
public
places
l
ike
s
hoppin
g
m
al
l,
ci
ne
m
a,
and
ai
r
port
s
e
r
vice
a
vaila
bili
t
y
is
increa
sin
g
noti
ceably
.
It
al
so
enh
a
nces
sec
ur
it
y
becau
se
of
the
dynam
ic
par
ti
ti
on
in
g
of
the
com
m
un
ic
at
ion
cha
nnel
in
the
betwee
n
m
ob
il
e
dev
ic
e
a
nd
cl
oud
se
rv
e
r.
I
n
this
sect
ion,
we
prese
nt
the
evaluati
on
of
our
ene
rgy
eff
ic
ie
nt
m
ob
il
it
y
m
anag
em
ent
i
n
m
ob
il
e
cl
ou
d
com
pu
ti
ng
(
E2M2MC
2)
sy
stem
and
it
com
par
ed
with
the
existi
ng
dis
tribu
te
d
fo
ll
ow
m
e
cl
ou
d
c
ontr
oller
(
DF
MC
C).
The
perform
ance
of
pro
posed
E
2M2
MC
2
syst
e
m
is
analy
zed
by
the
diff
e
re
nt
te
sti
ng
sce
nar
i
os
:
i
m
pact
of
m
ob
il
e
us
er
den
sit
y
an
d
t
heir
s
pe
ed.
The
num
ber
of
m
ob
il
e
use
rs
is
var
ie
d
from
30
to
11
0
in
fi
r
st
te
st
and
the
m
ob
il
e
us
er
s
peed
is
var
ie
d
from
20
to
100
m
s
in
the
seco
nd
scenari
o.
F
or
t
his
te
sti
ng
,
we
us
e
f
our
DC
G
and
LM
A,
one
ID
MD
a
nd
one
FMC
C
with
high
de
ns
it
y
m
ob
il
e
us
ers
.
Bot
h
te
sts
are
im
ple
m
ented
in
Net
work
Sim
ulato
r
(NS2
)
to
ol
with
10
00
×
1000
m
2
network
siz
e
.
The
sim
ulati
on p
a
ram
et
ers
are su
m
m
arized in
Ta
ble
1.
Table
1
.
Sim
ul
at
ion
P
a
ram
et
e
rs
Para
m
eters
Valu
es
Nu
m
b
e
r
o
f
m
o
b
ile
u
sers
3
0
,
5
0
,
7
0
,
9
0
,
1
1
0
Mob
ile
u
ser sp
eed
(
m
s)
2
0
,
4
0
,
6
0
,
8
0
,
1
0
0
Nu
m
b
e
r
o
f
DCG
4
Nu
m
b
e
r
o
f
I
MA
4
Nu
m
b
e
r
o
f
I
D
MD
1
Nu
m
b
e
r
o
f
FM
CC
1
Netwo
rk size
1
0
0
0
×1
0
0
0
m
2
Tr
af
f
ic
m
o
d
el
Co
n
stan
t bit rate
Si
m
u
latio
n
ti
m
e
(
s
)
100
5.1.
Imp
ac
t of mobil
e
users
In
t
his
sce
nar
i
o,
we
var
yi
ng
the
num
ber
of
node
from
30
t
o
110
with
t
he
fixe
d
s
pee
d
as
60
m
s
and
the
pe
rfo
rm
ance
of
pr
opos
e
d
E2M2MC
2
is
com
par
ed
with
the
e
xisti
ng
D
FMC
C
syst
e
m
.
Fig
ur
e
3
sho
w
s
the
pack
et
l
os
s r
at
e o
f pro
posed E2M2MC
2
an
d
existi
ng
DFM
CC
syst
e
m
.
The plot cl
earl
y dep
ic
ts t
he p
acket l
os
s
rate o
f pro
po
se
d
E
2M2
MC
2
s
yst
e
m
is v
ery low
in
te
rm
s o
f 41% l
ess c
om
par
ed
to
e
xisti
ng
DF
MC
C sy
stem
.
Figure
3
.
Pac
ke
t l
os
s
rate co
m
par
ison
with
i
m
pact o
f
m
ob
i
le
u
ser
s
Figure
4
s
how
s
the
energy
con
s
um
ption
of
pr
op
os
e
d
E2
M2MC
2
an
d
existi
ng
DF
M
CC
syst
e
m
.
The
pl
ot
cl
early
dep
ic
ts
the
energy
co
ns
um
ption
of
propo
sed
E2M
2MC
2
syst
e
m
is
ver
y
low
in
te
rm
s
of
39
%
le
ss co
m
par
ed
to ex
ist
in
g DF
MC
C sy
stem
.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
ERMO
2 alg
or
i
thm:
an en
er
gy
eff
ic
ie
nt m
obil
it
y m
anage
men
t i
n
m
obil
e clo
ud
co
m
pu
ti
ng
… (
L.
Pall
avi
)
1965
Figure
4
.
Ene
r
gy cons
um
ption
c
om
par
isons
with im
pact o
f m
ob
il
e u
sers
Figure
5
s
how
s
the
thr
oughput
of
pro
po
se
d
E2M
2MC2
and
e
xisti
ng
DF
MC
C
syst
em
.
The
plot
cl
early
dep
ic
ts
the
th
rou
ghput
of
pr
opos
e
d
E
2M2
MC
2
syst
e
m
is
ver
y
high
in
te
rm
s
of
24%
high
c
om
par
ed
t
o
existi
ng DFM
CC
sys
tem
.
Figure
5
.
Th
r
ough
pu
t c
om
pari
so
ns wit
h i
m
p
act
o
f
m
ob
il
e use
rs
Figure
6
s
how
s
the
fair
ness
i
nd
e
x
of
propo
sed
E
2M
2MC2
a
nd
existi
ng
DF
MC
C
syst
e
m
.
The
plot
cl
early
dep
ic
ts
the
fair
ness
in
de
x
of
pr
opos
e
d
E2M2MC
2
syst
e
m
is
ver
y
hig
h
i
n
te
rm
s
of
44%
high
c
ompar
e
d
to ex
ist
in
g DF
MC
C sy
stem
.
Figure
6
.
Fair
ne
ss in
dex com
p
aris
ons w
it
h
i
m
pact o
f
m
ob
il
e u
se
rs
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
3
,
June
201
9
:
1957
-
1967
1966
Figure
7
sho
w
s
the
delay
of
pro
po
se
d
E
2M
2MC2
a
nd
e
xisti
ng
DF
MC
C
syst
e
m
.
The
plo
t
cl
early
dep
ic
ts
the
del
ay
of
pro
pose
d
E
2M2
MC
2
syst
e
m
is
ver
y
low
i
n
te
rm
s
of
20%
high
c
om
par
ed
to
ex
ist
ing
DF
MC
C sy
ste
m
.
Figure
7
.
Del
a
y com
par
isons
with im
pact o
f m
ob
il
e u
sers
6.
CONCL
US
I
O
N
Mob
il
e
cl
oud
com
pu
ti
ng
m
a
kes
us
e
of
the
resou
rces
from
the
cl
ou
d
to
i
ncr
ease
the
c
om
pu
ta
ti
on
al
capab
il
it
ie
s
of
m
ob
il
e
dev
ic
es.
Mob
il
e
use
rs
can
scal
e
the
a
vaila
ble
ser
vices
to
m
a
tc
h
their
nee
ds
,
c
us
tom
iz
e
apps
an
d
acce
ss
cl
oud
sp
ace
fr
om
any
locat
ion
with
a
n
act
ive
internet
connecti
on.
I
n
this
pa
per,
we
ha
ve
pro
po
se
d
e
nergy
ef
fici
ent
m
ob
il
it
y
m
anag
e
m
ent
in
m
ob
il
e
cl
oud
c
om
pu
ti
ng
(E
2M2M
C2)
syst
em
.
It
m
ai
n
obj
ect
ive
i
s
be
st
cl
oud
com
pu
ta
ti
on
,
w
hich
are
achie
ved
by
the
el
ect
ive
rep
eat
m
ulti
-
ob
j
ect
iv
e
opti
m
i
zat
ion
(ERMO
2)
al
go
rithm
resp
ect
ively
.
The
si
m
ulati
on
res
ults
prov
e
the
e
ff
ect
i
ven
e
ss
of
our
E2M2MC
2
syst
e
m
in
te
rm
s o
f
delay
,
p
ac
ket loss
r
at
e, th
rou
ghput,
fairn
e
ss
-
i
nd
e
x and ene
r
gy con
su
m
ption
.
REFERE
NCE
S
[1]
J.
Zha
ng,
et
a
l
.
,
“
Towa
rds
Secur
e
Data
Distr
ibut
i
on
S
y
stems
in
Mobile
Cloud
C
om
puti
ng,
”
IE
E
E
Tr
ansacti
ons
on
Mobil
e
Computi
ng
,
vol
/i
ss
ue:
16
(
11
)
,
pp
.
3222
-
3
235,
2017
.
[2]
T.
T
aleb,
et
a
l
.
,
“
On
Multi
-
Ac
ce
ss
Edge
Com
puti
ng:
A
Surv
e
y
of
th
e
Emer
ging
5G
Network
Edge
C
loud
Archi
tectur
e
an
d
Orche
stra
ti
on,
”
IEEE
Comm
unic
ati
ons
Surv
e
ys
&
T
utorial
s
,
vol
/i
ss
ue:
19
(
3
)
,
pp.
1657
-
168
1
,
2017.
[3]
S.
Nam
asudra
a
nd
P.
Ro
y
,
“
T
i
m
e
saving
proto
col
for
d
ata
a
ccess
ing
in
c
loud
computing,
”
IET
Comm
unic
ati
on
s
,
vol
/i
ss
ue:
11
(
10
)
,
pp
.
1558
-
1565
,
2017.
[4]
H.
S
.
Mansour
i,
et
al
.
,
“
Joint
Optimal
Prici
ng
and
Ta
sk
Schedul
ing
in
Mob
il
e
Cloud
Com
puti
ng
S
y
st
ems
,
”
IEE
E
Tr
ansacti
o
ns on
Wireless
C
omm
unic
ati
ons
,
vol
/i
ss
ue:
16
(
8
)
,
pp.
5218
-
5232
,
2017.
[5]
J.
Zha
ng,
et
al
.
,
“
Towa
rds
sec
u
re
dat
a
distr
ibution
s
y
stems
in
m
obil
e
cl
oud
co
m
pu
ti
ng
,”
IEEE
Tr
an
sacti
ons
on
Mobil
e
Computi
ng
,
vol
/i
ss
ue:
16
(
11
),
pp
.
3222
-
3
235
,
2017
.
[6]
K.
Kum
ar
and
Y
.
H
.
Lu,
“
Cloud
Com
puti
ng
for
Mobile
Us
ers:
Can
Offl
oading
Com
puta
ti
on
Save
Ene
rg
y?
”
Computer
,
vol
/i
s
sue:
43
(
4
)
,
pp
.
5
1
-
56,
2010
.
[7]
P.
Sim
oens,
et
al
.
,
“
Remot
e
Displa
y
Solut
io
ns
for
Mobile
Cloud
Com
puti
ng
,
”
Computer
,
vol
/i
ss
ue:
44
(
8
)
,
pp.
46
-
53
,
2011
.
[8]
V.
Haghighi
and
N
.
S.
Moa
y
e
dian
,
“
An
Offloa
di
ng
Strat
eg
y
in
Mobile
Cloud
C
om
puti
ng
Consideri
ng
En
erg
y
a
nd
Delay
Constr
ai
nt
s
,”
IE
EE A
cc
ess
,
vol.
6
,
pp.
1184
9
-
11861
,
2018
.
[9]
W
.
Ren,
et
al
.
,
“
Li
ghtwei
gh
t
and
comprom
ise
res
il
ie
n
t
storage
ou
tsourci
ng
with
distri
bute
d
sec
u
re
ac
ce
ss
ib
ilit
y
in
m
obil
e
c
loud co
m
puti
ng
,
”
Ts
ing
hua
Scienc
e
and
Technol
ogy
,
vo
l
/
issue:
16
(
5
)
,
pp.
520
-
528,
2011
.
[10]
R.
Barga,
e
t
al
.
,
“
The
Cl
ie
nt
a
nd
the
Cloud
:
Dem
ocr
at
izing
Resea
rch
Com
p
uti
ng,
”
IE
EE
In
te
rnet
Comput
in
g
,
vol
/i
ss
ue:
15
(
1
)
,
pp.
72
-
75
,
2011
.
[11]
S.
Tursunova
and
Y.
Kim
,
“
Rea
li
st
ic
IE
EE
802.
11e
EDCA
m
odel
for
QoS
-
awa
re
m
o
bil
e
cl
oud
serv
ice
provisioni
ng
,
”
I
EE
E
Tr
ansacti
o
ns on
Consum
er
El
e
ct
ronics
,
v
ol
/
issue:
58
(
1
)
,
pp.
60
-
68,
2012
.
[12]
J.
Zhe
ng
,
et
al
.
,
“
Com
ponent
I
m
porta
nce
An
aly
sis
of
Mobile
Cloud
Com
put
ing
S
y
st
em
in
the
Presen
ce
of
Com
m
on
-
Cause
Fail
ure
s
,”
I
EE
E
Ac
c
ess
,
2018
.
[13]
W
.
C
.
Peng
and
M
.
S
.
Chen,
“
Quer
y
proc
essing
in
a
m
obil
e
computing
envi
ro
nm
ent
:
expl
oi
ti
n
g
the
fea
tur
es
of
as
y
m
m
et
r
y
,
”
IE
EE
Tr
ansacti
ons
on
Knowl
edge a
nd
Data
Eng
inee
ring
,
vol
/i
ss
ue:
17
(
7
)
,
pp
.
982
-
99
6,
2005
.
Evaluation Warning : The document was created with Spire.PDF for Python.