T
E
L
K
O
M
N
I
K
A
T
elec
o
m
m
un
ica
t
io
n Co
m
pu
t
ing
E
lect
ro
nics
a
nd
Co
ntr
o
l
Vo
l.
2
4
,
No
.
5
,
Octo
b
er
2
0
2
6
,
p
p
.
1
5
1
3
~
1
5
2
5
I
SS
N:
1
6
9
3
-
6
9
3
0
,
DOI
: 1
0
.
1
2
9
2
8
/TE
L
KOM
NI
KA.
v
2
4
i
5
.
2
7
7
0
6
1513
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//telko
mn
ika
.
u
a
d
.
a
c.
i
d
Predic
tive a
nd
fault
-
tole
ra
nt
v
irt
u
a
l ma
chine mig
ra
tion for
energy
-
efficient
cl
o
ud da
ta cen
ters
Ra
k
s
ha
n G
.
K
.
,
P
.
Vino
t
hiy
a
la
k
s
hm
i
D
e
p
a
r
t
me
n
t
o
f
C
o
mp
u
t
e
r
S
c
i
e
n
c
e
,
S
r
i
V
e
n
k
a
t
e
sw
a
r
a
C
o
l
l
e
g
e
o
f
En
g
i
n
e
e
r
i
n
g
,
Ta
m
i
l
N
a
d
u
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Dec
10
,
2
0
2
5
R
ev
is
ed
Ap
r
7
,
2
0
2
6
Acc
ep
ted
May
25
,
2
0
2
6
Virtu
a
l
m
a
c
h
in
e
(VM)
m
ig
ra
ti
o
n
is
a
k
e
y
m
e
c
h
a
n
ism
fo
r
imp
ro
v
i
n
g
e
n
e
rg
y
e
fficie
n
c
y
,
se
rv
ice
c
o
n
ti
n
u
it
y
a
n
d
re
li
a
b
il
it
y
in
c
lo
u
d
d
a
ta
c
e
n
ters
.
Ho
we
v
e
r,
c
o
n
v
e
n
ti
o
n
a
l
m
ig
ra
ti
o
n
stra
teg
ies
a
re
larg
e
ly
re
a
c
ti
v
e
a
n
d
fa
il
to
a
c
c
o
u
n
t
f
o
r
wo
rk
l
o
a
d
flu
c
t
u
a
ti
o
n
s
a
n
d
p
o
ten
ti
a
l
fa
il
u
re
s,
o
ften
re
su
lt
i
n
g
i
n
i
n
e
fficie
n
t
re
so
u
rc
e
u
ti
li
z
a
ti
o
n
a
n
d
in
c
re
a
se
d
se
rv
ice
-
lev
e
l
a
g
re
e
m
e
n
t
(S
LA)
v
io
latio
n
s.
Th
is
p
a
p
e
r
p
r
o
p
o
se
s
th
e
p
re
d
ic
ti
v
e
VM
m
i
g
ra
ti
o
n
m
a
n
a
g
e
r
(
P
VMM
),
a
u
n
ifi
e
d
fra
m
e
wo
rk
th
a
t
in
teg
r
a
tes
wo
rk
lo
a
d
f
o
re
c
a
stin
g
a
n
d
fa
il
u
re
p
re
d
ictio
n
fo
r
p
r
o
a
c
ti
v
e
m
i
g
ra
t
io
n
c
o
n
tro
l
.
P
V
M
M
c
o
m
b
i
n
e
s
a
g
a
te
d
re
c
u
rre
n
t
u
n
it
(G
RU)
-
b
a
se
d
m
o
d
e
l
fo
r
sh
o
rt
-
term
re
so
u
rc
e
p
re
d
ict
io
n
wi
th
a
m
a
c
h
in
e
-
lea
rn
in
g
-
b
a
se
d
fa
il
u
re
p
re
d
icto
r
to
i
d
e
n
ti
f
y
h
ig
h
-
risk
VMs.
An
a
d
a
p
ti
v
e
d
y
n
a
m
ic
t
h
re
sh
o
l
d
b
a
se
d
o
n
e
n
e
r
g
y
c
o
n
su
m
p
ti
o
n
(AD
T
-
E
C)
d
e
tec
ts
h
o
st
o
v
e
rlo
a
d
c
o
n
d
i
ti
o
n
s,
wh
il
e
a
n
im
p
ro
v
e
d
e
n
e
rg
y
-
a
wa
re
b
e
st
-
fit
(IE
ABF)
h
e
u
risti
c
se
lec
ts
o
p
ti
m
a
l
m
ig
ra
ti
o
n
targ
e
ts.
E
x
p
e
rime
n
tal
re
su
lt
s
d
e
m
o
n
stra
te
th
a
t
P
VMM
re
d
u
c
e
s
u
n
n
e
c
e
ss
a
ry
m
ig
ra
ti
o
n
,
imp
r
o
v
e
s
S
LA
c
o
m
p
l
ian
c
e
a
n
d
e
n
h
a
n
c
e
s
o
v
e
ra
ll
e
n
e
rg
y
e
fficie
n
c
y
a
n
d
s
y
ste
m
re
li
a
b
il
it
y
.
T
h
e
se
fin
d
in
g
s
h
ig
h
li
g
h
t
t
h
e
e
ffe
c
ti
v
e
n
e
ss
o
f
p
re
d
ictiv
e
,
m
u
lt
i
-
o
b
jec
ti
v
e
m
ig
ra
ti
o
n
stra
teg
ies
fo
r
n
e
x
t
-
g
e
n
e
ra
ti
o
n
c
l
o
u
d
d
a
ta ce
n
ter m
a
n
a
g
e
m
e
n
t.
K
ey
w
o
r
d
s
:
Failu
r
e
p
r
ed
ictio
n
Gate
d
r
ec
u
r
r
e
n
t u
n
it
-
b
ased
f
o
r
ec
asti
n
g
Ma
ch
in
e
lear
n
in
g
R
eso
u
r
ce
u
tili
za
tio
n
p
r
ed
ictio
n
Vir
tu
al
m
ac
h
in
e
m
ig
r
atio
n
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
R
ak
s
h
an
G.
K.
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
,
Sri
Ven
k
ateswar
a
C
o
lleg
e
o
f
E
n
g
in
ee
r
in
g
Pen
n
alu
r
,
Srip
er
u
m
b
u
d
u
r
,
T
a
m
il Na
d
u
,
I
n
d
ia.
E
m
ail: Rak
s
h
an
GK@
o
u
tlo
o
k
.
co
m
1.
I
NT
RO
D
UCT
I
O
N
Vir
tu
al
m
ac
h
in
e
(
VM
)
m
ig
r
atio
n
p
lay
s
a
v
ital
r
o
le
i
n
a
d
d
r
ess
in
g
th
ese
is
s
u
es
b
y
d
y
n
am
ically
r
ed
is
tr
ib
u
tin
g
wo
r
k
lo
a
d
s
to
m
ain
tain
lo
ad
b
alan
ce
a
n
d
s
er
v
i
ce
co
n
tin
u
ity
[
1
]
–
[
4
]
.
H
o
wev
e
r
,
b
alan
ci
n
g
e
n
er
g
y
ef
f
icien
cy
,
m
i
g
r
atio
n
o
v
er
h
ea
d
,
an
d
r
eliab
ilit
y
r
em
ain
s
c
h
allen
g
in
g
i
n
lar
g
e
-
s
ca
le
clo
u
d
e
n
v
ir
o
n
m
e
n
ts
[
5
]
–
[
7
]
.
C
o
n
v
e
n
t
i
o
n
a
l
m
i
g
r
at
i
o
n
t
e
c
h
n
iq
u
e
s
,
i
n
c
l
u
d
i
n
g
s
t
a
ti
c
a
n
d
d
y
n
am
i
c
t
h
r
e
s
h
o
l
d
-
b
as
e
d
p
o
l
i
ci
e
s
,
r
e
s
p
o
n
d
r
e
a
c
t
i
v
el
y
t
o
o
v
e
r
l
o
a
d
s
w
i
t
h
o
u
t
a
n
t
i
ci
p
a
t
i
n
g
f
u
t
u
r
e
w
o
r
k
l
o
a
d
c
h
a
n
g
e
s
o
r
p
o
t
e
n
t
i
a
l
f
a
il
u
r
e
s
[
8
]
–
[
1
0
]
.
S
u
c
h
m
e
t
h
o
d
s
c
a
n
t
r
i
g
g
e
r
r
e
d
u
n
d
a
n
t
m
i
g
r
a
t
i
o
n
s
o
r
d
e
l
a
y
c
r
i
t
i
c
al
a
c
ti
o
n
s
,
i
n
c
r
e
as
i
n
g
s
e
r
v
i
c
e
-
l
e
v
e
l
a
g
r
ee
m
e
n
t
(
S
L
A
)
v
i
o
la
t
i
o
n
s
[
7
]
.
Ad
v
an
ce
d
ap
p
r
o
ac
h
es
lev
e
r
a
g
in
g
m
ac
h
in
e
lear
n
in
g
(
ML
)
h
av
e
im
p
r
o
v
ed
d
ec
is
io
n
-
m
ak
in
g
i
n
m
ig
r
atio
n
co
n
t
r
o
l
[
6
]
–
[
1
2
]
.
Fo
r
ex
am
p
le,
Pau
lr
aj
et
a
l.
[
1
1
]
d
e
v
elo
p
e
d
a
co
m
b
in
e
d
f
o
r
ec
ast
-
lo
ad
-
awa
r
e
s
tr
ateg
y
,
wh
ile
Mish
r
a
et
a
l.
[
8
]
in
tr
o
d
u
ce
d
an
in
te
r
q
u
ar
tile r
an
g
e
(
I
QR
)
-
b
ased
m
o
d
el.
Yu
an
et
a
l.
[
1
2
]
f
u
r
t
h
er
im
p
r
o
v
e
d
wo
r
k
lo
a
d
p
r
ed
ictio
n
ac
cu
r
ac
y
u
s
in
g
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
L
STM
)
-
b
ased
m
o
d
els.
Similar
ly
,
B
o
m
m
ala
et
a
l.
[
1
3
]
ap
p
lied
s
u
p
er
v
is
ed
lea
r
n
in
g
to
p
r
ed
i
ct
VM
f
ailu
r
es.
Ho
wev
e
r
,
th
ese
m
eth
o
d
s
o
f
ten
o
p
tim
ize
a
s
in
g
le
o
b
jectiv
e
ei
th
er
en
er
g
y
o
r
r
eliab
ilit
y
with
o
u
t
jo
in
tly
a
d
d
r
ess
in
g
b
o
th
u
n
d
er
v
o
latile
clo
u
d
co
n
d
itio
n
s
.
T
h
is
s
tu
d
y
p
r
esen
ts
th
e
p
r
ed
ictiv
e
VM
m
ig
r
atio
n
m
an
ag
er
(
PVMM
)
,
an
in
tellig
en
t,
f
au
lt
-
to
ler
an
t
f
r
am
ewo
r
k
t
h
at
in
teg
r
ates
f
o
r
ec
asti
n
g
an
d
f
ailu
r
e
p
r
ed
ictio
n
to
en
ab
le
p
r
o
ac
tiv
e
m
i
g
r
atio
n
d
ec
is
io
n
s
.
PVMM
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N:
1
6
9
3
-
6
9
3
0
TEL
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
,
Vo
l.
2
4
,
No
.
5
,
Octo
b
e
r
2
0
2
6
:
1
5
1
3
-
1
5
2
5
1514
em
p
lo
y
s
a
g
ated
r
ec
u
r
r
en
t
u
n
i
t
(
GR
U)
m
o
d
el
f
o
r
s
h
o
r
t
-
ter
m
r
eso
u
r
ce
u
tili
za
tio
n
f
o
r
ec
asti
n
g
an
d
a
ML
-
b
ased
f
ailu
r
e
p
r
ed
icto
r
to
id
e
n
tify
h
ig
h
-
r
is
k
VM
s
.
An
ad
ap
tiv
e
d
y
n
am
ic
th
r
esh
o
ld
b
ased
o
n
en
er
g
y
co
n
s
u
m
p
tio
n
(
ADT
-
E
C
)
co
n
tin
u
o
u
s
ly
class
i
f
ies h
o
s
ts
as u
n
d
er
lo
ad
ed
o
r
o
v
er
lo
ad
ed
,
wh
ile
an
im
p
r
o
v
ed
en
er
g
y
-
awa
r
e
b
est
-
f
it (
I
E
AB
F)
alg
o
r
ith
m
s
elec
ts
o
p
tim
al
m
ig
r
atio
n
tar
g
ets,
en
s
u
r
in
g
e
f
f
icien
t r
eso
u
r
ce
u
s
e
an
d
SLA
co
m
p
lian
ce
.
T
h
e
p
r
o
p
o
s
ed
PVMM
f
r
am
e
wo
r
k
in
tr
o
d
u
ce
s
a
m
u
lti
-
o
b
jectiv
e,
p
r
ed
ictiv
e
m
ig
r
atio
n
s
tr
ateg
y
th
a
t
u
n
if
ies
r
eliab
ilit
y
,
en
er
g
y
e
f
f
i
cien
cy
an
d
o
p
er
atio
n
al
s
tab
ilit
y
.
Un
lik
e
co
n
v
e
n
tio
n
al
r
ea
cti
v
e
m
o
d
els,
PVMM
an
ticip
ates
o
v
er
lo
ad
an
d
f
ailu
r
e
co
n
d
itio
n
s
b
ef
o
r
e
th
ey
o
cc
u
r
,
th
er
eb
y
r
ed
u
cin
g
u
n
n
e
ce
s
s
ar
y
m
ig
r
atio
n
s
,
lo
wer
in
g
en
er
g
y
co
n
s
u
m
p
tio
n
(
E
C
)
an
d
im
p
r
o
v
in
g
o
v
er
all
d
ata
ce
n
ter
r
esil
ien
ce
.
T
h
ese
r
esu
lts
co
n
tr
ib
u
te
t
o
ad
v
an
cin
g
in
tellig
en
t
an
d
s
u
s
tain
ab
le
clo
u
d
r
eso
u
r
ce
m
an
ag
em
en
t
f
o
r
n
e
x
t
-
g
en
er
atio
n
co
m
p
u
tin
g
en
v
ir
o
n
m
en
ts
.
2.
L
I
T
E
R
AT
U
RE
SU
RVE
Y
VM
m
ig
r
atio
n
is
a
f
u
n
d
am
en
t
al
m
ec
h
an
is
m
i
n
clo
u
d
d
ata
ce
n
ter
s
f
o
r
im
p
r
o
v
in
g
r
eso
u
r
ce
u
tili
za
tio
n
,
en
s
u
r
in
g
s
er
v
ice
co
n
tin
u
ity
,
a
n
d
r
e
d
u
cin
g
EC
.
E
ar
ly
s
tu
d
ies
p
r
im
ar
ily
f
o
c
u
s
ed
o
n
p
er
f
o
r
m
an
ce
m
o
d
elin
g
a
n
d
en
er
g
y
-
ef
f
icien
t
co
n
s
o
lid
atio
n
.
Kh
az
ae
i
et
a
l.
[
1
]
in
tr
o
d
u
c
ed
q
u
eu
i
n
g
-
b
ased
an
aly
tical
m
o
d
els
to
ev
alu
ate
p
er
f
o
r
m
an
ce
in
clo
u
d
en
v
ir
o
n
m
en
ts
,
wh
ile
B
elo
g
lazo
v
a
n
d
B
u
y
y
a
[
2
]
p
r
o
p
o
s
ed
ad
a
p
tiv
e
h
eu
r
is
tics
f
o
r
d
y
n
am
ic
VM
co
n
s
o
lid
atio
n
,
f
o
r
m
in
g
th
e
b
asis
f
o
r
m
an
y
en
er
g
y
-
awa
r
e
m
ig
r
atio
n
s
tr
ateg
ies.
T
h
ese
f
o
u
n
d
atio
n
al
wo
r
k
s
estab
lis
h
ed
th
e
im
p
o
r
tan
ce
o
f
b
alan
ci
n
g
r
eso
u
r
ce
u
tili
za
tio
n
a
n
d
SLA
c
o
m
p
lian
ce
.
Su
b
s
eq
u
en
t
r
esear
ch
s
h
if
ted
to
war
d
en
er
g
y
-
awa
r
e
an
d
o
p
tim
izatio
n
-
d
r
iv
en
m
i
g
r
atio
n
t
ec
h
n
iq
u
es.
Ma
et
a
l.
[
3
]
p
r
o
v
i
d
ed
a
co
m
p
r
eh
en
s
iv
e
r
ev
iew
o
f
VM
m
i
g
r
atio
n
s
tr
ateg
ies
f
o
r
en
er
g
y
m
in
im
izatio
n
,
wh
ile
W
an
g
et
a
l.
[
4
]
,
Mu
s
taf
a
et
a
l.
[
7
]
p
r
o
p
o
s
ed
en
er
g
y
-
a
n
d
SLA
-
awa
r
e
co
n
s
o
lid
atio
n
ap
p
r
o
ac
h
es
to
im
p
r
o
v
e
o
p
er
atio
n
al
ef
f
icien
c
y
.
R
ein
f
o
r
ce
m
en
t
lear
n
in
g
a
n
d
o
p
tim
iz
atio
n
-
b
ased
m
eth
o
d
s
f
u
r
th
er
a
d
v
an
ce
d
th
is
ar
ea
.
Fo
r
in
s
tan
ce
,
Z
h
an
g
et
a
l.
[
5
]
in
tr
o
d
u
ce
d
a
d
ee
p
Q
-
lear
n
in
g
m
o
d
el
f
o
r
en
er
g
y
-
ef
f
icien
t
s
ch
ed
u
lin
g
,
a
n
d
Go
m
ath
i
et
a
l.
[
6
]
ex
p
lo
r
ed
m
u
lti
-
o
b
jectiv
e
VM
p
lace
m
en
t
s
tr
ateg
ies
co
n
s
id
er
in
g
en
er
g
y
an
d
r
eso
u
r
ce
u
tili
za
tio
n
.
A
d
a
p
t
i
v
e
a
n
d
t
h
r
es
h
o
l
d
-
b
a
s
e
d
m
i
g
r
a
t
i
o
n
m
e
c
h
a
n
is
m
s
h
a
v
e
a
ls
o
b
e
e
n
w
i
d
el
y
s
t
u
d
i
e
d
.
M
is
h
r
a
e
t
a
l
.
[
8
]
p
r
o
p
o
s
e
d
a
d
y
n
a
m
i
c
t
h
r
e
s
h
o
l
d
-
b
a
s
e
d
l
o
a
d
b
a
l
a
n
c
i
n
g
t
e
c
h
n
i
q
u
e
,
w
h
i
l
e
J
i
a
n
g
a
n
d
C
h
e
n
[
9
]
d
e
v
e
l
o
p
e
d
a
s
e
l
f
-
a
d
a
p
t
i
v
e
V
M
p
la
c
e
m
e
n
t
s
t
r
a
te
g
y
t
o
i
m
p
r
o
v
e
e
n
e
r
g
y
e
f
f
i
c
i
e
n
c
y
u
n
d
e
r
d
y
n
a
m
i
c
w
o
r
k
l
o
a
d
s
.
S
im
i
l
a
r
l
y
,
W
a
n
g
e
t
a
l.
[
1
0
]
d
e
m
o
n
s
t
r
a
t
e
d
t
h
e
e
f
f
e
c
t
i
v
e
n
e
s
s
o
f
d
e
e
p
r
e
i
n
f
o
r
c
e
m
e
n
t
l
e
a
r
n
i
n
g
f
o
r
e
n
e
r
g
y
-
e
f
f
i
c
i
e
n
t
V
M
s
c
h
e
d
u
l
i
n
g
.
A
l
t
h
o
u
g
h
t
h
e
s
e
a
p
p
r
o
a
c
h
e
s
i
m
p
r
o
v
e
a
d
a
p
t
a
b
i
l
i
t
y
,
t
h
e
y
l
a
r
g
e
l
y
r
e
l
y
o
n
r
e
a
c
t
i
v
e
d
e
ci
s
i
o
n
-
m
a
k
i
n
g
b
a
s
e
d
o
n
c
u
r
r
e
n
t
s
y
s
te
m
s
ta
t
es
.
T
o
o
v
e
r
c
o
m
e
r
e
a
c
t
i
v
e
l
i
m
i
t
a
ti
o
n
s
,
p
r
e
d
i
c
ti
v
e
a
n
d
m
a
c
h
i
n
e
-
l
e
a
r
n
i
n
g
-
b
a
s
e
d
a
p
p
r
o
a
c
h
es
h
a
v
e
b
e
e
n
i
n
t
r
o
d
u
c
e
d
.
P
a
u
l
r
a
j
e
t
a
l
.
[
1
1
]
p
r
o
p
o
s
e
d
a
f
o
r
e
c
a
s
t
-
b
a
s
e
d
m
i
g
r
a
t
i
o
n
s
t
r
a
te
g
y
t
h
a
t
i
n
t
e
g
r
a
t
es
w
o
r
k
l
o
a
d
p
r
e
d
i
c
t
i
o
n
w
i
t
h
m
i
g
r
a
ti
o
n
c
o
n
t
r
o
l
.
Y
u
a
n
e
t
a
l
.
[
1
2
]
f
u
r
t
h
e
r
i
m
p
r
o
v
e
d
w
o
r
k
l
o
a
d
p
r
e
d
i
c
t
i
o
n
a
c
c
u
r
a
c
y
u
s
i
n
g
L
S
T
M
-
b
as
e
d
m
o
d
e
l
s
,
c
a
p
t
u
r
i
n
g
t
e
m
p
o
r
a
l
d
ep
e
n
d
e
n
c
i
e
s
i
n
l
a
r
g
e
-
s
c
a
le
d
a
t
a
c
e
n
t
e
r
s
.
I
n
a
d
d
i
t
i
o
n
,
B
o
m
m
al
a
e
t
a
l
.
[
1
3
]
d
e
v
el
o
p
e
d
a
m
a
c
h
i
n
e
-
l
e
a
r
n
i
n
g
-
b
a
s
e
d
f
a
i
lu
r
e
p
r
e
d
i
c
t
i
o
n
m
o
d
e
l
t
o
i
d
e
n
t
i
f
y
p
o
t
e
n
t
i
a
l
s
y
s
t
e
m
a
n
o
m
a
l
ie
s
.
T
h
e
s
e
s
t
u
d
i
es
h
i
g
h
l
i
g
h
t
t
h
e
g
r
o
w
i
n
g
r
o
l
e
o
f
p
r
e
d
i
c
t
i
v
e
a
n
a
l
y
t
i
cs
i
n
e
n
a
b
l
i
n
g
p
r
o
a
c
t
i
v
e
m
i
g
r
a
t
i
o
n
d
e
ci
s
i
o
n
s
.
R
ec
en
t
wo
r
k
s
h
av
e
also
ex
p
lo
r
ed
in
tellig
en
t
an
d
h
eu
r
is
tic
-
b
ased
s
ch
ed
u
lin
g
f
r
am
ewo
r
k
s
.
Ab
d
u
llah
et
a
l.
[
1
4
]
p
r
o
p
o
s
ed
a
h
e
u
r
is
tic
-
b
ased
VM
co
n
s
o
lid
atio
n
ap
p
r
o
ac
h
,
wh
ile
B
u
y
y
a
et
a
l.
[
1
5
]
o
u
tlin
ed
th
e
v
is
io
n
o
f
au
t
o
n
o
m
ic
clo
u
d
s
y
s
tem
s
ca
p
ab
le
o
f
s
elf
-
m
an
ag
in
g
r
es
o
u
r
ce
s
.
Su
r
v
ey
s
tu
d
ies
[
1
6
]
,
[
1
7
]
em
p
h
asize
t
h
e
in
cr
ea
s
in
g
in
teg
r
atio
n
o
f
a
r
tific
ial
in
tellig
en
ce
an
d
f
a
u
lt
-
to
ler
an
ce
m
ec
h
an
is
m
s
in
clo
u
d
r
eso
u
r
ce
m
an
ag
em
en
t,
wh
ile
also
n
o
tin
g
th
e
lim
itatio
n
s
o
f
e
x
is
tin
g
r
e
ac
tiv
e
ap
p
r
o
ac
h
es.
Mu
lti
-
o
b
jectiv
e
an
d
s
u
s
tain
ab
ilit
y
-
awa
r
e
m
ig
r
atio
n
s
tr
ateg
ies
h
av
e
g
ain
e
d
atten
tio
n
in
r
ec
en
t
y
ea
r
s
.
Sin
g
h
et
a
l.
[
1
8
]
p
r
o
p
o
s
ed
a
Qo
S
-
awa
r
e
task
co
n
s
o
lid
atio
n
m
eth
o
d
to
r
ed
u
ce
SLA
v
io
latio
n
s
,
wh
ile
Kh
o
s
r
av
i
et
a
l.
[
1
9
]
in
tr
o
d
u
ce
d
a
ca
r
b
o
n
-
an
d
en
e
r
g
y
-
awa
r
e
VM
p
lace
m
en
t
s
tr
ateg
y
f
o
r
g
eo
g
r
a
p
h
ica
lly
d
is
tr
ib
u
ted
d
ata
ce
n
ter
s
.
Similar
ly
,
Far
ah
n
ak
ian
et
a
l.
[
2
0
]
a
p
p
lied
a
n
t
co
lo
n
y
o
p
tim
izatio
n
f
o
r
e
n
er
g
y
-
ef
f
icien
t
VM
co
n
s
o
lid
atio
n
.
C
o
m
p
ar
ativ
e
a
n
aly
s
es,
s
u
ch
as
Nag
m
a
et
a
l
.
[
2
1
]
,
f
u
r
th
er
h
ig
h
li
g
h
t
th
e
tr
ad
e
-
o
f
f
s
b
etwe
en
d
if
f
er
en
t m
i
g
r
atio
n
a
n
d
c
o
n
s
o
l
id
atio
n
s
tr
ateg
ies u
n
d
er
v
ar
y
i
n
g
wo
r
k
lo
a
d
s
.
Mo
r
e
r
ec
e
n
t
s
tu
d
ies
h
a
v
e
em
p
h
asized
p
o
wer
-
awa
r
e
an
d
co
m
m
u
n
icatio
n
-
awa
r
e
o
p
tim
izatio
n
in
VM
p
lace
m
en
t.
Ar
s
h
ad
et
a
l.
[
2
2
]
an
d
Su
n
il
an
d
Patel
[
2
3
]
d
em
o
n
s
tr
ated
im
p
r
o
v
em
e
n
ts
in
en
er
g
y
ef
f
icien
c
y
th
r
o
u
g
h
d
y
n
a
m
ic
p
o
wer
u
tili
za
tio
n
m
o
d
elin
g
.
I
n
ad
d
itio
n
,
F
ar
za
i
et
a
l.
[
2
4
]
p
r
o
p
o
s
ed
a
co
m
m
u
n
icatio
n
-
awa
r
e
m
u
lti
-
o
b
jectiv
e
o
p
tim
izatio
n
m
o
d
el,
h
ig
h
lig
h
tin
g
t
h
e
im
p
o
r
tan
ce
o
f
n
etwo
r
k
co
n
s
id
er
atio
n
s
in
VM
p
lace
m
en
t
d
ec
is
io
n
s
.
Acc
u
r
ate
p
o
wer
m
o
d
elin
g
h
as
also
b
ee
n
r
ec
o
g
n
i
ze
d
as
a
k
ey
f
ac
to
r
in
e
v
alu
at
in
g
en
er
g
y
-
e
f
f
icien
t
clo
u
d
s
y
s
tem
s
.
L
in
et
a
l.
[
2
5
]
p
r
esen
ted
a
q
u
a
n
titativ
e
s
er
v
er
p
o
wer
m
o
d
el
wid
ely
u
s
ed
f
o
r
esti
m
atin
g
EC
in
clo
u
d
d
ata
ce
n
ter
s
,
en
ab
lin
g
m
o
r
e
r
ea
lis
tic
s
im
u
latio
n
an
d
e
v
alu
atio
n
o
f
m
ig
r
atio
n
s
tr
ateg
ies.
Desp
ite
th
ese
ad
v
an
ce
m
en
ts
,
m
o
s
t
ex
is
tin
g
ap
p
r
o
ac
h
es
ad
d
r
ess
eith
er
en
er
g
y
ef
f
icien
c
y
o
r
r
eliab
ilit
y
in
d
ep
en
d
en
tly
a
n
d
o
f
ten
r
ely
o
n
r
ea
cti
v
e
o
r
s
in
g
le
-
o
b
jectiv
e
o
p
tim
izatio
n
s
tr
ateg
ies.
Ho
wev
er
,
lim
ited
wo
r
k
h
as
f
o
cu
s
ed
o
n
in
teg
r
atin
g
wo
r
k
lo
ad
f
o
r
ec
asti
n
g
,
f
ailu
r
e
p
r
e
d
ictio
n
,
a
n
d
e
n
er
g
y
-
awa
r
e
d
ec
is
io
n
-
m
ak
in
g
in
to
a
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
P
r
ed
ictive
a
n
d
fa
u
lt
-
to
lera
n
t
virt
u
a
l m
a
ch
in
e
mig
r
a
tio
n
fo
r
en
erg
y
-
efficien
t c
lo
u
d
…
(
R
a
k
s
h
a
n
G.
K
.
)
1515
u
n
if
ied
m
ig
r
atio
n
f
r
a
m
ewo
r
k
.
T
o
ad
d
r
ess
th
is
g
ap
,
th
is
s
tu
d
y
p
r
o
p
o
s
es
th
e
PVMM
,
wh
ich
in
teg
r
ates
GR
U
-
b
ased
wo
r
k
lo
ad
f
o
r
ec
asti
n
g
,
m
ac
h
in
e
-
lear
n
i
n
g
-
b
ased
f
ailu
r
e
p
r
ed
ictio
n
,
a
d
ap
tiv
e
en
er
g
y
-
awa
r
e
th
r
esh
o
ld
in
g
,
an
d
o
p
tim
ize
d
h
o
s
t
s
elec
tio
n
.
T
h
is
u
n
if
ied
a
p
p
r
o
ac
h
en
a
b
les
p
r
o
ac
tiv
e,
m
u
lti
-
o
b
jecti
v
e
VM
m
ig
r
atio
n
,
im
p
r
o
v
in
g
en
e
r
g
y
e
f
f
icien
cy
,
SLA
co
m
p
lian
ce
,
an
d
s
y
s
tem
r
eliab
ilit
y
in
d
y
n
am
ic
clo
u
d
en
v
ir
o
n
m
e
n
ts
.
3.
M
E
T
H
O
DO
L
O
G
Y
T
h
e
p
r
o
p
o
s
ed
PVMM
in
tr
o
d
u
ce
s
a
p
r
o
ac
tiv
e
an
d
f
au
lt
-
t
o
ler
an
t
VM
m
ig
r
atio
n
f
r
am
e
wo
r
k
th
at
ad
d
r
ess
es
th
e
lim
itatio
n
s
o
f
r
e
ac
tiv
e
ap
p
r
o
ac
h
es.
Un
lik
e
c
o
n
v
en
tio
n
al
m
eth
o
d
s
,
PVMM
i
n
teg
r
ates
wo
r
k
lo
ad
f
o
r
ec
asti
n
g
,
f
ailu
r
e
p
r
ed
ictio
n
an
d
en
er
g
y
-
awa
r
e
d
ec
is
io
n
-
m
ak
in
g
to
en
ab
le
in
tellig
e
n
t
m
ig
r
atio
n
u
n
d
er
d
y
n
am
ic
clo
u
d
co
n
d
itio
n
s
.
T
h
e
PVMM
ar
ch
itectu
r
e
co
n
s
is
ts
o
f
s
ix
m
o
d
u
les
(
Fig
u
r
e
1
)
,
ea
ch
r
esp
o
n
s
ib
le
f
o
r
a
s
p
ec
if
ic
s
tag
e
o
f
p
r
ed
ictiv
e
m
ig
r
atio
n
:
d
ata
p
r
e
p
r
o
ce
s
s
in
g
,
ad
ap
tiv
e
th
r
esh
o
ld
in
g
,
f
o
r
ec
asti
n
g
,
f
ailu
r
e
p
r
ed
ictio
n
,
r
is
k
-
b
ased
VM
s
elec
tio
n
,
h
o
s
t
s
elec
tio
n
an
d
m
ig
r
atio
n
e
x
ec
u
tio
n
.
T
h
ese
m
o
d
u
les
o
p
er
ate
p
er
io
d
ically
to
e
n
s
u
r
e
co
n
tin
u
o
u
s
m
o
n
ito
r
i
n
g
a
n
d
d
ec
is
io
n
-
m
ak
in
g
.
Fig
u
r
e
1
.
Ar
c
h
itectu
r
e
o
f
th
e
p
r
o
p
o
s
ed
wo
r
k
3
.
1
.
Da
t
a
c
o
llect
io
n a
nd
prepro
ce
s
s
ing
m
o
du
le
T
h
is
m
o
d
u
le
co
llects
tim
e
-
s
er
ies
d
ata
f
o
r
ea
c
h
VM
,
in
clu
d
in
g
ce
n
tr
al
p
r
o
ce
s
s
in
g
u
n
it
(
C
PU)
,
m
em
o
r
y
,
d
is
k
u
tili
za
tio
n
,
SL
A
v
io
latio
n
s
an
d
h
o
s
t
-
lev
el
EC
.
Data
i
s
n
o
r
m
alize
d
an
d
s
eg
m
en
ted
in
to
s
lid
in
g
win
d
o
ws
f
o
r
m
o
d
el
in
p
u
t.
A
d
d
itio
n
al
f
ea
tu
r
es
s
u
ch
as
task
p
r
io
r
ity
,
r
etr
y
c
o
u
n
t
an
d
s
ch
ed
u
lin
g
d
elay
ar
e
ex
tr
ac
ted
f
o
r
f
ailu
r
e
p
r
ed
ictio
n
.
3
.
1
.
1
.
ADT
-
EC
T
h
e
ADT
-
E
C
m
ec
h
an
is
m
d
y
n
am
ically
class
if
ies
h
o
s
t
en
er
g
y
s
tates
to
d
etec
t
o
v
er
lo
ad
c
o
n
d
itio
n
s
.
Un
lik
e
s
tatic
th
r
esh
o
ld
s
,
it a
d
a
p
ts
b
ased
o
n
r
ea
l
-
tim
e
EC
.
A
h
o
s
t is cla
s
s
if
ied
as:
(
ℎ
)
<
×
max
⇒
B
L
UE
×
max
≤
(
ℎ
)
≤
×
max
⇒
GR
E
E
N
(
ℎ
)
>
×
max
⇒
R
E
D
wh
er
e
(
ℎ
)
:
cu
r
r
en
t
h
o
s
t
en
er
g
y
,
max
:
m
ax
im
u
m
allo
wab
le
en
er
g
y
,
,
:
ad
ap
tiv
e
th
r
esh
o
ld
f
ac
to
r
s
(
0
.
6
≤
a
<
b
≤
0
.
9
)
.
T
h
ese
v
alu
es
wer
e
s
elec
ted
b
ased
o
n
p
r
io
r
th
r
esh
o
ld
-
b
ased
s
tu
d
ies
an
d
em
p
ir
ic
al
tu
n
in
g
to
b
alan
ce
m
ig
r
atio
n
s
en
s
itiv
ity
an
d
s
tab
i
lity
.
T
h
e
co
m
p
lete
ADT
-
E
C
alg
o
r
ith
m
is
p
r
esen
ted
in
Alg
o
r
i
th
m
1
,
w
h
en
a
h
o
s
t
en
ter
s
th
e
R
E
D
s
tate,
m
ig
r
atio
n
is
tr
ig
g
er
ed
.
B
L
UE
h
o
s
ts
ar
e
p
r
eser
v
ed
t
o
av
o
i
d
u
n
n
ec
ess
ar
y
m
ig
r
atio
n
s
.
Alg
o
r
ith
m
1.
ADT
-
EC
Input:
: Set of all hosts in the data center
(
ℎ
)
: Current energy consumption of host
ℎ
max
: Maximum allowable energy for host
ℎ
ℎ
: Load history of host
ℎ
Threshold factors
,
(0.6 ≤ a < b ≤ 0.9)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N:
1
6
9
3
-
6
9
3
0
TEL
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
,
Vo
l.
2
4
,
No
.
5
,
Octo
b
e
r
2
0
2
6
:
1
5
1
3
-
1
5
2
5
1516
Output
:
Host energy state classification
Level
(
ℎ
)
∈
{
BLUE, GREEN, RED
}
Updated migration threshold values for each host
Steps:
1.
Initialize:
Energy Status List (ESL) = Null and Migration Threshold List (MTL) =
Null
2.
For each host
ℎ
∈
:
a. Retrieve
(
ℎ
)
and
ℎ
.
b. Compute the average energy utilization
‾
(
ℎ
)
over the monitoring window.
3.
Classify energy consumption level:
i.
If
(
ℎ
)
<
×
max
:
-
Set
Level
(
ℎ
)
=
BLUE
// Host is idle or underutilized.
ii.
Else if
×
max
≤
(
ℎ
)
≤
×
max
:
-
Set
Level
(
ℎ
)
=
GREEN
// Host is in normal range.
iii.
Else if
(
ℎ
)
>
×
max
:
-
Set
Level
(
ℎ
)
=
RED
// Host is overloaded.
4.
Adjust migration threshold (MT)
based on the classified level:
-
For
BLUE
, increase MT to avoid unnecessary migrations.
-
For
GREEN
, maintain MT at nominal value.
-
For
RED
, lower MT to trigger immediate VM migration.
5.
Update
ESL
and
MTL
with
Level
(
ℎ
)
and corresponding threshold.
6.
Identify hosts as:
Overloaded
if
Level
(
ℎ
)
=
RED
Underutilized
if
Level
(
ℎ
)
=
BLUE
7.
End For
8.
Return
[
ESL
,
MTL
]
to the VM Selection Module.
9.
End
3
.
2
.
G
RU
-
ba
s
ed
re
s
o
urce
f
o
re
ca
s
t
ing
m
o
du
le
A
GR
U
m
o
d
el
is
u
s
ed
to
p
r
ed
ict
s
h
o
r
t
-
ter
m
VM
r
eso
u
r
ce
u
t
ilizatio
n
(
C
PU,
m
em
o
r
y
,
d
is
k
)
.
GR
U
is
s
elec
ted
d
u
e
t
o
its
ef
f
icien
cy
in
ca
p
tu
r
in
g
tem
p
o
r
al
d
e
p
en
d
en
cies
with
lo
wer
co
m
p
u
tatio
n
al
o
v
e
r
h
ea
d
th
an
L
STM
.
T
h
e
m
o
d
el
is
tr
ain
ed
o
n
h
is
to
r
ical
wo
r
k
lo
ad
tr
ac
es
an
d
p
r
ed
icts
f
u
tu
r
e
u
tili
za
tio
n
to
en
ab
le
p
r
o
ac
tiv
e
o
v
er
lo
ad
d
etec
tio
n
.
3
.
3
.
F
a
ilu
re
predict
io
n m
o
du
le
A
s
u
p
er
v
is
ed
lear
n
i
n
g
m
o
d
el
(
r
an
d
o
m
f
o
r
est
(
R
F)
/
ex
tr
em
e
g
r
ad
ien
t
b
o
o
s
tin
g
(
XGBo
o
s
t
)
)
p
r
e
d
icts
VM
f
ailu
r
e
p
r
o
b
ab
ilit
y
:
FR
(
)
∈
[
0
,
1
]
Featu
r
e
v
ec
to
r
s
in
clu
d
e
C
PU
p
r
ess
u
r
e
lev
els,
task
r
etr
y
f
r
e
q
u
en
cy
,
r
eso
u
r
ce
q
u
eu
e
len
g
t
h
,
r
ec
en
t
h
ar
d
war
e
er
r
o
r
lo
g
s
,
a
n
d
s
ch
e
d
u
lin
g
an
o
m
aly
in
d
icato
r
s
.
I
t
is
tr
ain
ed
o
n
lab
elled
clo
u
d
tr
ac
e
d
atasets
to
class
if
y
VM
s
in
to
lo
w
-
r
is
k
an
d
h
ig
h
-
r
is
k
ca
t
eg
o
r
ies b
ased
o
n
lik
elih
o
o
d
o
f
f
ailu
r
e.
3
.
4
.
O
v
er
lo
a
d
risk
s
co
ring
a
nd
VM
prio
rit
iza
t
io
n m
o
du
le
Fo
r
ev
er
y
VM
,
a
co
m
p
o
s
ite
r
is
k
s
co
r
e
is
co
m
p
u
ted
:
=
∗
(
)
+
∗
Û
+
z
∗
(
)
(
1
)
wh
er
e
,
(
)
is
th
e
p
r
ed
icted
f
ailu
r
e
r
is
k
,
Û
is
th
e
f
o
r
ec
asted
r
eso
u
r
ce
u
tili
za
tio
n
(
av
er
ag
e
o
f
C
PU,
m
em
o
r
y
(
ME
M)
,
an
d
d
is
k
(
DI
SK)
)
,
(
)
is
th
e
r
ec
en
t
SLA
v
io
latio
n
f
r
e
q
u
en
cy
,
[
,
,
]
ar
e
th
e
e
m
p
ir
ically
tu
n
ed
weig
h
tin
g
p
ar
am
eter
s
.
T
h
e
weig
h
ts
ar
e
d
eter
m
in
ed
th
r
o
u
g
h
em
p
ir
ical
tu
n
in
g
an
d
s
en
s
itiv
ity
an
aly
s
is
to
b
alan
ce
r
eliab
ilit
y
a
n
d
en
er
g
y
ef
f
icien
cy
.
T
h
e
VM
s
ar
e
s
o
r
t
ed
b
y
d
escen
d
i
n
g
s
co
r
e
an
d
th
e
ca
n
d
i
d
ates
ar
e
s
elec
ted
f
o
r
m
ig
r
atio
n
.
3
.
4
.
1
.
Co
rr
ela
t
io
n
-
ba
s
ed
VM
ca
nd
ida
t
e
s
elec
t
io
n
(
SS
-
CA
U)
T
h
e
SS
-
C
AU
s
tr
ateg
y
s
elec
ts
VM
s
co
n
tr
ib
u
tin
g
m
o
s
t to
o
v
e
r
lo
ad
u
s
in
g
co
r
r
elatio
n
:
=
co
r
r
(
C
P
U
,
ℎ
C
P
U
)
VM
s
with
>
0
.
6
ar
e
s
elec
ted
an
d
s
o
r
ted
b
y
ascen
d
in
g
u
tili
za
tio
n
to
m
in
im
ize
m
ig
r
atio
n
o
v
er
h
ea
d
.
T
h
e
co
m
p
lete
SS
-
C
AU
p
r
o
ce
d
u
r
e
is
p
r
esen
ted
i
n
Alg
o
r
ith
m
2
.
T
h
is
en
s
u
r
es
tar
g
eted
l
o
ad
r
ed
u
ctio
n
with
m
in
im
al
d
ata
tr
an
s
f
er
co
s
t.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
P
r
ed
ictive
a
n
d
fa
u
lt
-
to
lera
n
t
virt
u
a
l m
a
ch
in
e
mig
r
a
tio
n
fo
r
en
erg
y
-
efficien
t c
lo
u
d
…
(
R
a
k
s
h
a
n
G.
K
.
)
1517
Alg
o
r
ith
m
2.
SS
-
C
AU
Input:
o
v
e
r
: Set of overloaded hosts identified by ADT
-
EC
(
ℎ
)
: Set of VMs residing on host
ℎ
CPU
: CPU utilization history of VM
ℎ
CPU
: CPU utilization history of host
ℎ
Correlation threshold
min
=
0
.
6
Output
:
Candidate VM list for migration
VM
Steps:
1.
Initialize:
VM
=
∅
and
is the Correlation coefficient for VM
2.
For each overloaded host
ℎ
∈
o
v
e
r
:
a. Retrieve all VMs
(
ℎ
)
associated with host
ℎ
.
b. For each VM
∈
(
ℎ
)
:
i. Compute the correlation coefficient
between VM and host CPU utilization using
regression:
=
corr
(
CPU
,
ℎ
CPU
)
ii.
If
>
min
:
-
Add
to candidate list
VM
.
c.
End For
3.
Sort
all candidate VMs
∈
VM
in
ascending order of current CPU utilization
CPU
(
)
.
4.
Select
VMs sequentially for migration starting from lowest CPU utilization, as
smaller workloads migrate faster and with less data transfer overhead.
5.
Continue migration
until the host’s utilization
ℎ
CPU
drops below the adaptive
threshold
determined by the ADT
-
EC algorithm.
6.
End For
7.
Return
the final candidate list
VM
to the
Host Selection (IEABF)
module.
8.
End
3
.
5
.
I
E
AB
F
ho
s
t
s
elec
t
io
n m
o
du
le
T
h
e
I
E
AB
F a
lg
o
r
ith
m
s
elec
ts
o
p
tim
al
d
esti
n
atio
n
h
o
s
ts
.
Fo
r
ea
ch
VM
:
Δ
(
ℎ
)
=
pos
t
(
ℎ
)
−
c
ur
r
e
nt
(
ℎ
)
(
2
)
T
h
e
h
o
s
t
with
th
e
m
in
im
u
m
Δ
(
ℎ
)
v
alu
e
an
d
s
u
f
f
icien
t
av
ailab
le
ca
p
ac
ity
is
s
elec
ted
as
th
e
d
esti
n
atio
n
.
Alg
o
r
ith
m
3
s
u
m
m
ar
izes
th
e
I
E
AB
F
h
o
s
t
s
elec
tio
n
p
r
o
ce
d
u
r
e.
T
h
is
b
est
-
f
it
h
eu
r
is
tic
en
s
u
r
es
o
p
tim
al
lo
ad
d
is
tr
ib
u
tio
n
an
d
r
e
d
u
ce
d
p
o
wer
o
v
er
h
ea
d
wh
ile
m
ain
tai
n
in
g
SLA
co
m
p
lian
ce
.
B
y
p
r
io
r
itizin
g
e
n
er
g
y
ef
f
icien
cy
an
d
o
p
er
atio
n
al
s
t
ab
ilit
y
,
I
E
AB
F
en
h
an
ce
s
th
e
o
v
er
all
ef
f
ec
tiv
e
n
ess
o
f
th
e
P
VM
M
f
r
am
ewo
r
k
,
lead
in
g
to
im
p
r
o
v
ed
r
eso
u
r
ce
u
tili
za
tio
n
,
lo
wer
m
ig
r
atio
n
f
r
eq
u
en
c
y
an
d
ex
ten
d
ed
d
ata
c
en
ter
life
s
p
an
.
T
h
e
alg
o
r
ith
m
o
p
er
ates
with
p
o
ly
n
o
m
ial
tim
e
co
m
p
le
x
ity
r
elat
iv
e
to
th
e
n
u
m
b
er
o
f
h
o
s
ts
an
d
VM
s
,
m
ak
in
g
it
s
u
itab
le
f
o
r
s
ca
lab
le
clo
u
d
en
v
ir
o
n
m
en
ts
.
Alg
o
r
ith
m
3.
I
E
AB
F
h
o
s
t s
elec
tio
n
Input:
VM
: List of VMs selected for migration (from SS
-
CAU)
: Set of all available hosts in the data center
(
ℎ
)
: Current energy consumption of host
ℎ
max
: Maximum allowable energy for host
ℎ
ℎ
: Available computing capacity (CPU, memory, disk) of host
ℎ
SLA
(
ℎ
)
: Current SLA violation rate of host
ℎ
Thresholds for utilization and SLA compliance
Output
:
Optimized VM
-
to
-
host mapping list
=
{
(
,
ℎ
t
a
r
g
e
t
)
}
Steps:
1.
Initialize:
=
∅
and
v
a
l
i
d
=
∅
2.
Preprocessing:
a. Exclude hosts that are
overloaded or underloaded
according to the ADT
-
EC classification.
b. Retain only hosts satisfying
(
ℎ
)
<
max
and
SLA
(
ℎ
)
<
threshold
.
c. Add all such hosts to
v
a
l
i
d
.
3.
Sort
VMs in
VM
by
descending CPU utilization
, prioritizing high
-
demand VMs for
placement first.
4.
For each VM
∈
VM
:
a. Initialize
Δ
min
=
∞
;
ℎ
b
e
s
t
=
null
.
b.
For each candidate host
ℎ
∈
v
a
l
i
d
:
i. Check resource capacity constraint:
ℎ
≥
ResourceDemand
(
)
.
ii. Estimate the expected post
-
placement energy:
post
(
ℎ
)
=
c
u
r
r
e
n
t
(
ℎ
)
+
EnergyIncrement
(
)
iii. Compute energy difference:
Δ
(
ℎ
)
=
p
o
s
t
(
ℎ
)
−
c
u
r
r
e
n
t
(
ℎ
)
iv.
If
Δ
(
ℎ
)
<
Δ
min
and capacity constraints are satisfied:
-
Update
Δ
min
=
Δ
(
ℎ
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N:
1
6
9
3
-
6
9
3
0
TEL
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
,
Vo
l.
2
4
,
No
.
5
,
Octo
b
e
r
2
0
2
6
:
1
5
1
3
-
1
5
2
5
1518
-
Set
ℎ
best
=
ℎ
c. Assign VM
to
ℎ
best
and update host energy state.
d. Add mapping
(
,
ℎ
b
e
s
t
)
to
.
e. Remove
ℎ
best
from
v
a
l
i
d
if capacity becomes saturated.
5.
End For
6.
Return
final mapping list
for migration execution.
7.
End
3
.
6
.
M
ig
ra
t
i
o
n
s
cheduli
ng
a
nd
ex
ec
utio
n m
o
du
le
T
h
is
m
o
d
u
le
c
o
o
r
d
i
n
ates m
ig
r
atio
n
d
ec
is
io
n
s
an
d
e
x
ec
u
tio
n
.
3
.
6
.
1
.
M
ig
ra
t
io
n
t
rig
g
er
ing
A
m
ig
r
atio
n
r
ea
d
in
ess
in
d
ex
(
MRI)
is
d
ef
in
ed
:
MRI
(
)
=
⋅
f
or
e
c
a
s
t
(
)
+
⋅
r
is
k
(
)
+
⋅
s
tate
(
ℎ
)
wh
er
e
f
or
e
c
a
s
t
(
)
r
ep
r
esen
ts
th
e
f
o
r
ec
a
s
ted
r
eso
u
r
ce
d
e
m
an
d
g
r
o
wt
h
r
ate,
r
is
k
(
)
th
e
p
r
e
d
icted
f
ailu
r
e
p
r
o
b
a
b
ilit
y
an
d
s
tate
(
ℎ
)
th
e
cu
r
r
e
n
t
en
er
g
y
lev
el
o
f
th
e
h
o
s
t
ℎ
.
T
h
e
weig
h
tin
g
p
ar
a
m
eter
s
(
,
,
)
ar
e
em
p
ir
ically
tu
n
ed
to
p
r
io
r
itize
eith
er
f
ailu
r
e
av
o
id
a
n
ce
o
r
en
er
g
y
ef
f
icien
cy
.
Mig
r
atio
n
is
t
r
ig
g
er
ed
o
n
ly
w
h
en
th
e
co
m
p
u
ted
MRI
ex
ce
ed
s
a
d
ef
in
ed
ac
tiv
atio
n
th
r
esh
o
l
d
,
p
r
ev
en
tin
g
u
n
n
ec
ess
ar
y
VM
r
el
o
ca
tio
n
s
.
3
.
6
.
2
.
M
ig
ra
t
io
n
ex
ec
utio
n
On
ce
a
m
ig
r
atio
n
d
ec
is
io
n
is
ap
p
r
o
v
ed
,
liv
e
VM
m
ig
r
atio
n
is
p
er
f
o
r
m
ed
u
s
in
g
a
p
r
e
-
co
p
y
m
ec
h
an
is
m
,
wh
er
e
m
em
o
r
y
p
ag
es
ar
e
iter
ativ
ely
tr
an
s
f
er
r
e
d
to
th
e
d
esti
n
atio
n
h
o
s
t
wh
ile
th
e
VM
r
em
ain
s
ac
tiv
e,
m
in
im
izin
g
s
er
v
ice
d
o
wn
tim
e.
Prio
r
to
m
ig
r
atio
n
,
th
e
s
ch
ed
u
ler
v
er
if
ies
th
e
s
u
itab
ilit
y
o
f
th
e
d
esti
n
atio
n
h
o
s
t
b
ased
o
n
ca
p
ac
ity
,
en
er
g
y
co
n
s
tr
ain
ts
an
d
SLA
r
eq
u
ir
em
en
ts
d
ef
in
ed
b
y
th
e
I
E
AB
F
m
o
d
u
le.
T
h
e
m
ig
r
atio
n
p
r
o
ce
s
s
co
n
s
is
ts
o
f
th
r
ee
p
h
ases
:
(
i)
p
r
e
-
co
p
y
,
wh
e
r
e
m
em
o
r
y
p
ag
es
ar
e
p
r
o
g
r
ess
iv
ely
tr
an
s
f
er
r
ed
;
(
ii)
s
to
p
-
an
d
-
co
p
y
,
wh
er
e
th
e
VM
is
b
r
ief
ly
p
au
s
ed
to
tr
an
s
f
er
r
em
ain
in
g
s
tate
;
an
d
(
iii)
s
tat
e
u
p
d
ate,
wh
e
r
e
VM
m
ap
p
i
n
g
s
ar
e
u
p
d
ate
d
an
d
p
o
s
t
-
m
ig
r
atio
n
en
er
g
y
is
r
ec
alcu
lated
.
T
o
en
s
u
r
e
n
etwo
r
k
s
tab
ilit
y
d
u
r
in
g
co
n
c
u
r
r
e
n
t
m
ig
r
atio
n
s
,
th
e
s
ch
ed
u
ler
ap
p
lies
ad
ap
tiv
e
b
an
d
wid
t
h
co
n
tr
o
l,
t
h
r
o
ttli
n
g
m
ig
r
atio
n
tr
af
f
ic
wh
en
n
etwo
r
k
u
tili
za
tio
n
ex
ce
ed
s
p
r
e
d
ef
in
e
d
th
r
esh
o
ld
s
.
Af
ter
m
ig
r
atio
n
,
v
alid
atio
n
ch
ec
k
s
ar
e
p
er
f
o
r
m
ed
to
en
s
u
r
e
s
y
s
tem
co
n
s
is
ten
cy
an
d
SLA
co
m
p
lian
c
e.
T
h
ese
in
clu
d
e
v
er
if
icatio
n
o
f
VM
s
tate
in
teg
r
ity
an
d
m
o
n
ito
r
i
n
g
o
f
r
esp
o
n
s
e
tim
e
an
d
t
h
r
o
u
g
h
p
u
t.
I
f
p
e
r
f
o
r
m
an
ce
d
eg
r
ad
atio
n
ex
ce
ed
s
ac
ce
p
tab
l
e
lim
its
,
a
r
o
llb
ac
k
m
ec
h
a
n
is
m
is
tr
ig
g
er
ed
.
3
.
6
.
3
.
M
ig
ra
t
io
n
co
s
t
a
nd
o
v
er
hea
d m
ini
m
iza
t
io
n
Mig
r
atio
n
o
v
er
h
ea
d
is
ev
alu
at
ed
u
s
in
g
th
e
c
o
s
t f
u
n
ctio
n
:
mig
(
,
ℎ
s
r
c
,
ℎ
ds
t
)
=
1
mig
+
2
mi
g
+
3
Δ
(
ℎ
ds
t
)
wh
er
e
mi
g
is
m
ig
r
atio
n
tim
e,
m
ig
is
b
an
d
wid
t
h
u
s
ag
e
an
d
Δ
(
ℎ
ds
t
)
is
th
e
in
cr
em
en
tal
e
n
er
g
y
co
s
t
o
f
t
h
e
d
esti
n
atio
n
h
o
s
t.
T
h
e
weig
h
tin
g
f
ac
to
r
s
(
1
,
2
,
3
)
tu
n
ed
to
b
ala
n
ce
laten
cy
an
d
en
er
g
y
ef
f
i
cien
cy
.
Mig
r
atio
n
is
ex
ec
u
ted
o
n
ly
wh
en
th
e
b
en
ef
it
-
to
-
co
s
t
r
atio
ex
ce
ed
s
a
d
ef
i
n
ed
th
r
esh
o
ld
,
p
r
e
v
en
tin
g
u
n
n
ec
ess
ar
y
m
ig
r
atio
n
s
.
3
.
6
.
4
.
Su
m
m
a
r
y
o
f
o
pera
t
io
n
T
h
e
s
ch
ed
u
ler
in
teg
r
ates
d
ec
is
io
n
s
f
r
o
m
ADT
-
E
C
,
SS
-
C
AU
an
d
I
E
AB
F
to
p
e
r
f
o
r
m
co
o
r
d
i
n
ated
VM
m
ig
r
atio
n
.
I
t
h
an
d
les
m
ig
r
ati
o
n
tr
ig
g
er
in
g
,
ex
ec
u
tio
n
an
d
v
alid
atio
n
wh
ile
m
i
n
im
izin
g
o
v
er
h
e
ad
th
r
o
u
g
h
p
r
ed
ictiv
e
s
ch
ed
u
lin
g
an
d
n
etwo
r
k
-
awa
r
e
c
o
n
tr
o
l.
T
h
is
i
n
teg
r
ated
a
p
p
r
o
ac
h
e
n
s
u
r
es
r
ed
u
ce
d
d
o
wn
tim
e
,
im
p
r
o
v
e
d
en
er
g
y
ef
f
icien
cy
a
n
d
s
u
s
tain
ed
SLA
co
m
p
lian
ce
.
4.
E
XP
E
R
I
M
E
N
T
A
L
SE
T
UP
T
h
e
p
r
o
p
o
s
ed
PVMM
f
r
am
e
wo
r
k
was
ev
alu
ated
u
s
in
g
b
o
th
r
ea
l
-
wo
r
ld
clo
u
d
wo
r
k
lo
ad
tr
ac
es
an
d
s
im
u
latio
n
-
b
ased
en
v
ir
o
n
m
en
t
s
to
as
s
ess
its
p
er
f
o
r
m
an
ce
in
en
er
g
y
ef
f
icien
cy
,
m
ig
r
atio
n
f
r
eq
u
en
c
y
an
d
f
au
lt
-
to
ler
an
t
p
r
ed
ictio
n
ac
cu
r
ac
y
.
T
h
e
f
r
am
ewo
r
k
in
te
g
r
ates
en
e
r
g
y
-
awa
r
e
m
ig
r
atio
n
co
n
tr
o
l,
GR
U
-
b
ased
r
eso
u
r
ce
f
o
r
ec
asti
n
g
an
d
m
ac
h
i
n
e
-
lear
n
in
g
-
d
r
iv
en
f
ailu
r
e
p
r
ed
ictio
n
to
en
a
b
le
p
r
o
ac
tiv
e
an
d
i
n
tellig
en
t
m
ig
r
atio
n
d
ec
is
io
n
s
with
in
d
y
n
am
ic
cl
o
u
d
in
f
r
astru
ctu
r
es.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
P
r
ed
ictive
a
n
d
fa
u
lt
-
to
lera
n
t
virt
u
a
l m
a
ch
in
e
mig
r
a
tio
n
fo
r
en
erg
y
-
efficien
t c
lo
u
d
…
(
R
a
k
s
h
a
n
G.
K
.
)
1519
4
.
1
.
Sim
ula
t
i
o
n
env
iro
nm
en
t
E
x
p
er
im
en
ts
wer
e
co
n
d
u
cted
u
s
in
g
th
e
C
lo
u
d
Sim
3
.
0
.
3
to
o
lk
it,
a
well
-
estab
lis
h
ed
s
im
u
latio
n
f
r
am
ewo
r
k
f
o
r
ev
alu
atin
g
en
er
g
y
-
awa
r
e
r
eso
u
r
ce
m
an
a
g
e
m
en
t
an
d
VM
m
ig
r
atio
n
s
tr
ateg
ies
in
clo
u
d
d
ata
ce
n
ter
s
.
T
h
e
s
im
u
lated
en
v
ir
o
n
m
en
t
c
o
m
p
r
is
ed
1
0
0
h
eter
o
g
en
eo
u
s
p
h
y
s
ical
h
o
s
ts
m
o
d
ele
d
af
ter
SP
E
C
p
o
wer
b
en
ch
m
ar
k
s
er
v
er
s
with
HP
P
r
o
L
ian
t
ML
1
1
0
G4
(
I
n
tel
Xeo
n
3
0
4
0
,
2
co
r
es,
1
.
8
6
GHz
,
4
GB
R
AM
)
an
d
HP
Pro
L
ian
t
ML
1
1
0
G5
(
I
n
tel
Xeo
n
3
0
7
5
,
2
c
o
r
es,
2
.
6
6
GHz
,
4
GB
R
AM
)
s
p
ec
if
icatio
n
s
.
T
h
e
p
o
wer
co
n
s
u
m
p
tio
n
p
r
o
f
iles
o
f
th
ese
h
o
s
ts
wer
e
d
er
iv
ed
f
r
o
m
e
s
tab
lis
h
ed
s
er
v
er
p
o
wer
m
o
d
elin
g
ap
p
r
o
ac
h
es
in
clo
u
d
d
ata
ce
n
ter
s
[
2
5
]
,
wh
ic
h
ar
e
wid
ely
u
s
ed
to
ap
p
r
o
x
im
a
te
en
er
g
y
u
tili
za
tio
n
b
ased
o
n
wo
r
k
lo
ad
in
te
n
s
ity
.
E
ac
h
h
o
s
t
s
u
p
p
o
r
ted
1
0
to
1
5
VM
s
wi
th
v
ar
iab
le
r
eso
u
r
ce
d
em
an
d
s
an
d
lo
ad
p
atter
n
s
d
er
i
v
ed
f
r
o
m
Plan
etL
ab
wo
r
k
lo
ad
tr
ac
es,
en
s
u
r
in
g
r
e
alis
tic
tem
p
o
r
al
u
tili
za
tio
n
f
l
u
ctu
atio
n
s
th
at
em
u
late
r
ea
l
-
wo
r
ld
o
p
er
atio
n
al
co
n
d
itio
n
s
.
E
ac
h
s
im
u
latio
n
r
e
p
r
esen
ted
2
4
h
o
u
r
s
o
f
co
n
tin
u
o
u
s
o
p
er
at
io
n
,
d
u
r
in
g
wh
ich
t
h
e
C
lo
u
d
S
im
en
er
g
y
m
o
d
el
was
ex
ten
d
ed
to
co
m
p
u
te
d
y
n
am
ic
EC
s
tates
f
o
r
ea
ch
h
o
s
t,
r
ep
r
esen
ted
as
an
d
,
to
ca
p
tu
r
e
p
r
e
-
a
n
d
p
o
s
t
-
m
ig
r
ati
o
n
en
e
r
g
y
v
ar
iatio
n
s
.
T
h
e
AD
T
-
E
C
m
ec
h
an
is
m
was
in
teg
r
a
ted
to
class
if
y
h
o
s
t
en
er
g
y
s
tates a
n
d
d
y
n
am
ically
ad
ju
s
t m
ig
r
atio
n
t
h
r
esh
o
ld
s
b
ased
o
n
lo
ad
h
is
to
r
y
a
n
d
p
o
we
r
u
tili
za
tio
n
tr
en
d
s
.
T
h
e
GR
U
-
b
ased
r
eso
u
r
ce
f
o
r
ec
asti
n
g
an
d
RF
-
b
ased
f
ailu
r
e
p
r
ed
ictio
n
m
o
d
els
wer
e
p
r
e
-
tr
ain
ed
ex
ter
n
ally
in
Py
th
o
n
u
s
in
g
t
h
e
Go
o
g
le
C
lu
s
ter
,
L
ANL
Mu
s
tan
g
an
d
T
r
in
ity
tr
ac
es
an
d
th
e
n
em
b
ed
d
ed
with
in
C
lo
u
d
Sim
f
o
r
in
f
er
en
ce
d
u
r
i
n
g
r
u
n
tim
e.
Pre
d
ictio
n
s
wer
e
ex
ch
an
g
e
d
b
etwe
en
th
e
ML
m
o
d
els
an
d
th
e
s
im
u
lato
r
th
r
o
u
g
h
a
lig
h
tweig
h
t
J
av
aScr
ip
t
O
b
ject
N
o
tatio
n
(
J
SON
)
-
b
ased
in
ter
f
ac
e,
en
a
b
lin
g
C
lo
u
d
Sim
to
r
eq
u
est
f
o
r
ec
ast
a
n
d
f
ailu
r
e
p
r
o
b
ab
ilit
y
v
al
u
es
at
ea
ch
s
ch
ed
u
lin
g
in
ter
v
al.
T
h
e
in
f
er
e
n
ce
l
aten
cy
o
f
th
e
ML
m
o
d
els
is
ass
u
m
ed
to
b
e
n
e
g
lig
ib
le
in
th
e
s
im
u
latio
n
e
n
v
ir
o
n
m
e
n
t,
as
p
r
ed
ictio
n
s
ar
e
p
r
e
-
co
m
p
u
ted
a
n
d
in
v
o
k
e
d
at
d
is
cr
ete
s
ch
ed
u
lin
g
in
ter
v
als
r
ath
er
th
an
ex
ec
u
ted
as
r
ea
l
-
tim
e
s
tr
ea
m
i
n
g
in
f
er
en
ce
.
T
h
is
ab
s
tr
ac
tio
n
allo
ws th
e
s
tu
d
y
to
f
o
cu
s
o
n
m
ig
r
atio
n
d
ec
is
io
n
e
f
f
ec
tiv
en
ess
r
ath
er
th
a
n
r
u
n
tim
e
ML
o
v
er
h
ea
d
.
T
o
en
s
u
r
e
r
ep
r
o
d
u
cib
ilit
y
an
d
r
ed
u
ce
s
to
ch
asti
c
b
ias,
ea
ch
e
x
p
er
im
en
t
was
r
ep
ea
ted
f
iv
e
t
im
es
with
d
if
f
er
en
t
r
a
n
d
o
m
s
ee
d
s
an
d
av
er
ag
e
r
esu
lts
wer
e
r
ep
o
r
te
d
f
o
r
all
ev
al
u
atio
n
m
etr
ics.
T
h
e
f
r
am
ewo
r
k
is
d
esig
n
ed
to
s
ca
le
with
in
cr
ea
s
in
g
n
u
m
b
er
s
o
f
h
o
s
ts
an
d
VM
s
,
as
th
e
m
o
d
u
lar
ar
ch
itectu
r
e
d
is
tr
ib
u
tes
f
o
r
ec
asti
n
g
,
p
r
ed
ictio
n
a
n
d
s
ch
ed
u
lin
g
task
s
in
d
e
p
en
d
e
n
tly
.
Ho
wev
er
,
c
o
m
p
u
tatio
n
al
o
v
er
h
ea
d
ass
o
ciate
d
with
lar
g
e
-
s
ca
le
in
f
er
en
ce
a
n
d
m
i
g
r
atio
n
co
o
r
d
in
atio
n
is
n
o
t
ex
p
licitly
m
o
d
ele
d
in
C
lo
u
d
Sim
an
d
th
u
s
s
ca
lab
ilit
y
is
ev
alu
ated
p
r
im
ar
ily
in
ter
m
s
o
f
s
y
s
tem
b
e
h
av
io
r
r
ath
er
th
a
n
ex
ec
u
tio
n
tim
e.
4
.
2
.
Da
t
a
s
et
a
nd
wo
r
k
lo
a
d t
ra
ce
s
T
h
r
ee
p
u
b
licly
a
v
ailab
le
clo
u
d
wo
r
k
lo
a
d
d
atasets
wer
e
u
tili
ze
d
f
o
r
tr
ain
in
g
an
d
v
ali
d
atin
g
th
e
p
r
ed
ictiv
e
m
o
d
els
in
teg
r
ate
d
in
to
th
e
PVMM
f
r
am
ewo
r
k
.
T
h
e
G
o
o
g
le
C
lu
s
ter
T
r
ac
e
was
em
p
lo
y
e
d
to
r
ep
r
esen
t
p
r
o
d
u
ctio
n
-
s
ca
le,
h
eter
o
g
en
eo
u
s
wo
r
k
lo
a
d
s
ty
p
i
ca
l
o
f
lar
g
e
clo
u
d
in
f
r
astru
c
tu
r
es.
T
h
e
L
ANL
Mu
s
tan
g
an
d
T
r
i
n
ity
tr
ac
es
wer
e
u
s
ed
t
o
ca
p
tu
r
e
h
ig
h
-
p
er
f
o
r
m
a
n
ce
c
o
m
p
u
tin
g
wo
r
k
l
o
ad
c
h
ar
ac
ter
is
tics
,
in
clu
d
in
g
v
ar
y
in
g
task
co
m
p
l
ex
ities
,
ex
ec
u
tio
n
d
u
r
atio
n
s
a
n
d
r
eso
u
r
ce
c
o
n
ten
tio
n
b
e
h
av
io
r
s
.
E
ac
h
d
ataset
co
n
tain
ed
d
etailed
in
f
o
r
m
ati
o
n
o
n
C
PU
u
tili
za
tio
n
,
m
e
m
o
r
y
co
n
s
u
m
p
tio
n
,
d
is
k
in
p
u
t/o
u
tp
u
t
(
I
/O
)
,
task
p
r
io
r
ity
,
r
etr
y
c
o
u
n
t a
n
d
ex
it c
o
d
es.
All d
atasets
wer
e
clea
n
ed
an
d
p
r
ep
r
o
ce
s
s
ed
to
r
em
o
v
e
d
u
p
lic
ates,
in
co
m
p
lete
r
ec
o
r
d
s
an
d
a
n
o
m
alies,
f
o
llo
win
g
s
tan
d
a
r
d
ized
p
r
o
ce
d
u
r
es
d
escr
ib
ed
in
r
ec
e
n
t
m
ac
h
i
n
e
-
lear
n
in
g
-
b
ased
f
ailu
r
e
p
r
ed
ictio
n
s
tu
d
ies
[
1
3
]
.
T
h
e
p
r
ep
r
o
ce
s
s
ed
d
ata
wer
e
p
ar
titi
o
n
ed
in
to
tr
ain
in
g
an
d
test
in
g
s
ets
in
th
e
r
atio
o
f
7
0
:3
0
f
o
r
f
o
r
ec
asti
n
g
an
d
8
0
:2
0
f
o
r
f
ailu
r
e
p
r
ed
ictio
n
.
T
h
ey
wer
e
th
en
u
s
ed
t
o
d
e
v
el
o
p
th
e
GR
U
-
b
ased
r
eso
u
r
ce
f
o
r
ec
asti
n
g
an
d
RF
-
b
ased
f
ailu
r
e
p
r
ed
ictio
n
m
o
d
e
ls
in
T
en
s
o
r
Flo
w
v
2
.
1
2
an
d
S
cik
it
-
lear
n
v
1
.
4
,
r
esp
ec
tiv
ely
.
T
h
e
tr
ain
ed
m
o
d
els
wer
e
th
en
ex
p
o
r
ted
an
d
i
n
ter
f
a
ce
d
with
C
lo
u
d
Sim
3
.
0
.
3
to
p
r
o
v
id
e
r
ea
l
-
tim
e
p
r
ed
ictio
n
s
d
u
r
in
g
s
im
u
latio
n
.
Fo
r
s
im
u
latio
n
wo
r
k
l
o
ad
s
,
Plan
etL
ab
tr
ac
es
wer
e
em
p
l
o
y
ed
to
g
e
n
er
ate
tim
e
-
v
a
r
y
in
g
u
tili
za
tio
n
in
p
u
ts
f
o
r
th
e
VM
s
h
o
s
ted
o
n
ea
ch
s
im
u
lated
s
er
v
er
.
T
h
ese
tr
ac
es
ef
f
ec
tiv
ely
r
ep
r
o
d
u
ce
d
y
n
am
ic
clo
u
d
co
n
d
itio
n
s
b
y
in
t
r
o
d
u
cin
g
r
ea
lis
tic
f
lu
ctu
atio
n
s
in
C
PU
an
d
m
em
o
r
y
d
em
an
d
.
T
h
is
co
m
b
in
ed
u
s
e
o
f
r
ea
l
-
wo
r
ld
tr
ac
e
d
ata
f
o
r
m
o
d
el
tr
ain
in
g
an
d
Plan
etL
ab
-
b
ased
wo
r
k
lo
ad
e
m
u
latio
n
e
n
s
u
r
ed
th
at
th
e
s
im
u
latio
n
en
v
ir
o
n
m
en
t
ac
c
u
r
ately
r
ef
lec
ted
b
o
t
h
o
p
er
atio
n
al
u
n
p
r
e
d
ictab
ilit
y
an
d
p
r
e
d
ictiv
e
d
ec
is
io
n
-
m
ak
in
g
b
eh
av
io
r
with
in
lar
g
e
-
s
ca
le
clo
u
d
d
ata
c
en
ter
s
.
4
.
3
.
M
o
del
t
ra
ini
ng
a
nd
pa
r
a
m
et
er
c
o
nfig
ura
t
io
n
T
h
e
r
eso
u
r
ce
f
o
r
ec
asti
n
g
m
o
d
el
was
im
p
lem
en
ted
u
s
in
g
a
GR
U
n
eu
r
al
n
etwo
r
k
d
e
v
elo
p
ed
in
T
en
s
o
r
Flo
w
v
2
.
1
2
.
T
h
e
GR
U
was
tr
ain
ed
o
n
7
0
%
o
f
th
e
h
is
to
r
ical
tim
e
-
s
er
ies
d
ata,
w
h
ich
in
clu
d
ed
C
PU,
m
em
o
r
y
an
d
d
is
k
u
tili
za
tio
n
m
etr
ics,
wh
ile
th
e
r
em
ain
in
g
3
0
%
was
r
eser
v
ed
f
o
r
test
in
g
an
d
v
alid
atio
n
.
E
ac
h
in
p
u
t
s
eq
u
en
ce
was
co
n
s
tr
u
cted
u
s
in
g
a
1
0
-
s
tep
s
lid
in
g
win
d
o
w,
co
r
r
esp
o
n
d
in
g
to
5
-
m
in
u
te
s
am
p
lin
g
in
ter
v
als,
en
ab
lin
g
th
e
m
o
d
el
to
ca
p
tu
r
e
tem
p
o
r
al
d
ep
e
n
d
en
cies
ac
r
o
s
s
s
h
o
r
t
-
ter
m
wo
r
k
lo
ad
f
lu
ctu
atio
n
s
.
T
h
e
n
etwo
r
k
ar
c
h
itectu
r
e
co
n
s
is
ted
o
f
two
h
id
d
e
n
lay
er
s
with
6
4
n
eu
r
o
n
s
ea
ch
an
d
was
o
p
tim
iz
ed
u
s
in
g
th
e
Ad
am
o
p
tim
izer
with
a
lear
n
in
g
r
at
e
o
f
0
.
0
0
1
.
T
h
e
m
ea
n
s
q
u
ar
ed
er
r
o
r
(
MSE
)
lo
s
s
f
u
n
ctio
n
was
em
p
lo
y
ed
to
m
in
im
ize
p
r
ed
ictio
n
er
r
o
r
d
u
r
in
g
tr
ain
in
g
.
T
h
is
co
n
f
i
g
u
r
ati
o
n
allo
wed
th
e
GR
U
m
o
d
el
t
o
ef
f
ec
tiv
ely
lear
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N:
1
6
9
3
-
6
9
3
0
TEL
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
,
Vo
l.
2
4
,
No
.
5
,
Octo
b
e
r
2
0
2
6
:
1
5
1
3
-
1
5
2
5
1520
d
y
n
am
ic
u
tili
za
tio
n
p
atter
n
s
an
d
f
o
r
ec
ast
n
ea
r
-
f
u
tu
r
e
w
o
r
k
lo
ad
tr
en
d
s
,
f
ac
ilit
atin
g
p
r
o
ac
tiv
e
m
ig
r
atio
n
d
ec
is
io
n
s
b
ef
o
r
e
o
v
er
lo
a
d
co
n
d
itio
n
s
o
cc
u
r
r
e
d
.
T
h
e
f
ailu
r
e
p
r
e
d
ictio
n
m
o
d
el
was
im
p
lem
en
ted
u
s
in
g
s
u
p
er
v
is
ed
ML
tech
n
iq
u
es
in
Scik
it
-
lear
n
v
1
.
4
,
em
p
lo
y
in
g
an
8
0
:2
0
s
p
lit
f
o
r
tr
ain
in
g
an
d
test
in
g
.
T
h
e
f
ea
t
u
r
e
s
et
in
clu
d
ed
k
e
y
b
eh
a
v
io
r
al
an
d
s
y
s
tem
-
lev
el
in
d
icato
r
s
s
u
ch
as
task
r
etr
y
co
u
n
t,
r
eso
u
r
ce
co
n
ten
tio
n
in
d
ex
,
SLA
v
io
latio
n
f
r
e
q
u
en
c
y
an
d
C
PU
q
u
eu
e
len
g
th
.
Mu
ltip
le
class
if
ier
s
,
in
clu
d
in
g
RF
an
d
XGBo
o
s
t,
wer
e
ev
alu
ated
to
d
eter
m
in
e
o
p
t
im
al
p
er
f
o
r
m
an
ce
.
Am
o
n
g
th
ese,
th
e
RF
m
o
d
el,
co
n
f
ig
u
r
ed
with
2
0
0
esti
m
ato
r
s
,
d
em
o
n
s
tr
ated
s
u
p
e
r
io
r
p
r
ed
ictiv
e
ca
p
ab
ilit
y
,
ac
h
iev
in
g
an
F1
-
s
co
r
e
o
f
0
.
9
4
,
co
n
s
is
ten
t
with
p
r
io
r
f
in
d
in
g
s
in
h
ig
h
-
co
n
f
id
e
n
ce
co
m
p
u
tin
g
B
o
m
m
ala
et
a
l.
[
1
3
]
.
T
h
e
f
ailu
r
e
p
r
o
b
ab
ilit
y
s
co
r
es
g
en
er
ated
b
y
th
is
m
o
d
e
l
wer
e
in
teg
r
ated
in
to
t
h
e
PVMM
f
r
am
ewo
r
k
t
o
p
r
io
r
itize
m
ig
r
atio
n
o
f
h
ig
h
-
r
is
k
VM
s
,
th
er
eb
y
e
n
h
an
ci
n
g
s
y
s
tem
f
au
lt to
ler
an
ce
a
n
d
s
er
v
ice
co
n
tin
u
ity
.
4
.
4
.
P
er
f
o
r
m
a
nce
m
e
t
rics
T
h
e
p
er
f
o
r
m
a
n
ce
o
f
th
e
p
r
o
p
o
s
ed
f
r
am
ew
o
r
k
was
ass
ess
ed
u
s
in
g
a
s
et
o
f
q
u
a
n
titativ
e
m
etr
ics
th
at
co
llectiv
ely
m
ea
s
u
r
e
en
er
g
y
ef
f
icien
cy
,
r
eliab
ilit
y
an
d
o
p
e
r
atio
n
al
s
tab
ilit
y
.
E
C
r
ep
r
esen
ts
th
e
to
tal
en
er
g
y
u
tili
ze
d
b
y
all
ac
tiv
e
h
o
s
ts
d
u
r
in
g
s
im
u
latio
n
.
SLA
v
io
lat
io
n
d
en
o
tes
th
e
r
atio
o
f
SL
A
b
r
ea
ch
es
s
u
ch
as
m
is
s
ed
r
esp
o
n
s
e
d
ea
d
li
n
es
o
r
laten
cy
v
io
latio
n
s
to
th
e
to
tal
n
u
m
b
er
o
f
e
x
ec
u
ted
task
s
.
T
h
e
Nu
m
b
e
r
o
f
VM
m
ig
r
atio
n
s
r
ef
lects
th
e
f
r
eq
u
e
n
cy
o
f
liv
e
m
ig
r
atio
n
ev
e
n
ts
p
er
s
im
u
latio
n
cy
cle
an
d
s
er
v
es
as
an
in
d
icato
r
o
f
s
y
s
tem
s
tab
il
ity
.
Failu
r
e
p
r
ed
ictio
n
ac
cu
r
ac
y
was
q
u
an
tifie
d
th
r
o
u
g
h
p
r
ec
is
io
n
,
r
ec
all
an
d
F1
-
s
co
r
e
m
etr
ics
o
b
tain
ed
f
r
o
m
th
e
ML
-
b
ased
f
ailu
r
e
p
r
e
d
icto
r
.
Fin
ally
,
av
er
ag
e
C
PU
u
tili
za
tio
n
r
ep
r
esen
ts
th
e
m
ea
n
h
o
s
t
u
tili
za
tio
n
lev
el
af
ter
m
ig
r
atio
n
,
p
r
o
v
id
in
g
in
s
ig
h
t in
to
o
v
er
a
ll lo
ad
b
alan
cin
g
ef
f
icien
c
y
.
4
.
5
.
B
a
s
eline
co
m
pa
riso
n met
ho
ds
T
o
v
alid
ate
t
h
e
ef
f
ec
tiv
e
n
ess
o
f
th
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
,
c
o
m
p
ar
ativ
e
ex
p
er
im
e
n
ts
wer
e
co
n
d
u
cte
d
ag
ain
s
t
f
o
u
r
estab
lis
h
ed
VM
m
ig
r
atio
n
ap
p
r
o
ac
h
es:
s
tatic
t
h
r
esh
o
ld
-
b
ased
m
ig
r
atio
n
[
2
]
,
d
y
n
am
ic
I
QR
-
b
ased
m
ig
r
atio
n
[
8
]
,
co
m
b
in
ed
f
o
r
e
ca
s
t
lo
ad
-
awa
r
e
(
C
F
-
L
A)
mi
g
r
atio
n
[
1
1
]
an
d
th
e
E
AB
F
a
lg
o
r
ith
m
[
3
]
.
T
h
ese
b
aselin
e
m
eth
o
d
s
en
co
m
p
ass
a
s
p
ec
tr
u
m
o
f
s
tr
ateg
ies
r
an
g
in
g
f
r
o
m
s
tatic
an
d
d
y
n
am
ically
ad
ju
s
ted
th
r
esh
o
ld
s
to
p
r
e
d
ictiv
e,
lo
ad
-
awa
r
e
an
d
en
e
r
g
y
-
o
p
tim
ized
m
ig
r
atio
n
p
o
licies,
en
ab
lin
g
a
co
m
p
r
eh
en
s
iv
e
ev
alu
atio
n
o
f
th
e
im
p
r
o
v
e
m
en
ts
ac
h
iev
ed
b
y
t
h
e
p
r
o
p
o
s
ed
p
r
ed
ictiv
e
an
d
f
a
u
lt
-
to
ler
an
t
m
ig
r
atio
n
f
r
am
ew
o
r
k
.
4
.
6
.
T
hrea
t
s
t
o
v
a
lid
it
y
T
h
e
ex
p
er
im
en
tal
s
etu
p
h
as
ce
r
tain
lim
itatio
n
s
.
C
lo
u
d
Sim
a
b
s
tr
ac
ts
lo
w
-
lev
el
s
y
s
tem
f
ac
to
r
s
s
u
ch
as
n
etwo
r
k
laten
cy
,
h
ar
d
war
e
h
eter
o
g
en
eity
an
d
v
ir
tu
aliza
tio
n
o
v
er
h
ea
d
,
wh
ic
h
m
ay
af
f
ec
t
r
ea
l
-
wo
r
ld
p
er
f
o
r
m
an
ce
.
ML
in
f
er
e
n
ce
l
aten
cy
is
n
o
t
e
x
p
licitly
m
o
d
e
led
,
as
p
r
ed
ictio
n
s
ar
e
p
r
e
-
tr
ain
ed
an
d
in
v
o
k
ed
d
u
r
in
g
s
im
u
latio
n
.
Ad
d
itio
n
al
ly
,
th
e
d
atasets
u
s
ed
(
G
o
o
g
le
C
lu
s
ter
an
d
L
ANL
tr
ac
es)
m
ay
n
o
t
ca
p
t
u
r
e
al
l
wo
r
k
lo
ad
v
ar
iatio
n
s
,
p
o
ten
tia
lly
af
f
ec
tin
g
g
en
e
r
aliza
tio
n
.
Scalab
ilit
y
is
ev
alu
ated
at
th
e
s
y
s
tem
lev
el,
as
co
m
p
u
tatio
n
al
o
v
er
h
ea
d
o
f
la
r
g
e
-
s
ca
le
in
f
er
en
ce
a
n
d
m
ig
r
a
tio
n
co
o
r
d
in
atio
n
is
n
o
t
f
u
lly
r
ep
r
esen
ted
in
th
e
s
im
u
latio
n
.
Desp
ite
th
ese
lim
itatio
n
s
,
th
e
s
etu
p
p
r
o
v
id
es
a
co
n
s
is
ten
t
an
d
r
e
p
r
o
d
u
cib
le
b
asis
f
o
r
ev
alu
atin
g
p
r
ed
ictiv
e
an
d
en
er
g
y
-
awa
r
e
VM
m
ig
r
atio
n
.
5.
P
E
RF
O
RM
A
NCE
E
VA
L
U
AT
I
O
N
T
h
e
p
er
f
o
r
m
an
ce
o
f
th
e
PVMM
was
an
aly
ze
d
u
s
in
g
th
e
s
etu
p
d
escr
ib
ed
ab
o
v
e,
with
em
p
h
asis
o
n
en
er
g
y
ef
f
icien
cy
,
m
ig
r
atio
n
f
r
eq
u
en
c
y
,
SLA
co
m
p
lian
ce
an
d
f
ailu
r
e
p
r
ed
ictio
n
ac
cu
r
ac
y
.
T
h
e
d
ata
s
h
o
wn
is
f
r
o
m
5
r
u
n
s
ea
ch
with
th
e
m
ea
n
v
alu
es sh
o
wn
.
5
.
1
.
E
v
a
lua
t
i
o
n
o
bje
ct
iv
es
T
h
e
ex
p
er
im
en
tal
e
v
alu
atio
n
was
d
esig
n
ed
to
c
o
m
p
r
e
h
e
n
s
iv
ely
ass
ess
th
e
ef
f
ec
tiv
en
ess
o
f
th
e
p
r
o
p
o
s
ed
f
r
a
m
ewo
r
k
ac
r
o
s
s
it
s
k
ey
f
u
n
ctio
n
al
co
m
p
o
n
e
n
ts
.
Sp
ec
if
ically
,
th
e
o
b
jectiv
es
wer
e
to
v
ali
d
ate
th
e
ADT
-
EC
m
ec
h
an
is
m
in
m
in
i
m
izin
g
u
n
n
ec
ess
ar
y
m
ig
r
atio
n
s
an
d
s
tab
ilizin
g
h
o
s
t
wo
r
k
lo
a
d
s
.
T
o
ev
alu
ate
t
h
e
ac
cu
r
ac
y
o
f
th
e
GR
U
-
b
ased
f
o
r
ec
asti
n
g
m
o
d
el
u
n
d
er
d
y
n
a
m
ically
v
ar
y
in
g
clo
u
d
wo
r
k
lo
ad
s
.
T
o
m
ea
s
u
r
e
th
e
ca
p
ab
ilit
y
o
f
th
e
f
ailu
r
e
p
r
ed
ic
tio
n
m
o
d
el
in
id
en
tif
y
in
g
h
ig
h
-
r
is
k
VM
s
b
ef
o
r
e
p
o
ten
tial
f
ai
lu
r
e
ev
e
n
ts
.
An
d
to
q
u
an
tify
th
e
im
p
r
o
v
em
en
ts
in
en
er
g
y
e
f
f
icien
cy
a
n
d
SL
A
co
m
p
lian
ce
ac
h
iev
ed
th
r
o
u
g
h
th
e
I
E
AB
F
h
o
s
t
s
elec
tio
n
alg
o
r
ith
m
.
All
r
ep
o
r
ted
r
esu
lts
r
ep
r
esen
t
m
ea
n
v
a
lu
es
o
v
er
f
iv
e
in
d
ep
en
d
en
t
s
im
u
latio
n
r
u
n
s
.
T
h
e
o
b
s
er
v
ed
p
e
r
f
o
r
m
an
ce
tr
en
d
s
r
em
ain
ed
co
n
s
is
ten
t
ac
r
o
s
s
r
u
n
s
with
lo
w
v
ar
ian
ce
,
in
d
icatin
g
s
tab
ilit
y
o
f
th
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
.
5
.
2
.
EC
a
na
ly
s
is
Fig
u
r
e
2
illu
s
tr
ates
av
er
ag
e
EC
ac
r
o
s
s
m
ig
r
atio
n
s
tr
ateg
i
es.
T
h
e
ADT
-
E
C
+I
E
AB
F
co
m
b
in
atio
n
r
ed
u
ce
d
to
tal
en
e
r
g
y
u
s
ag
e
b
y
ap
p
r
o
x
im
ately
1
5
-
1
8
%
r
elati
v
e
to
d
y
n
a
m
ic
I
QR
-
b
ased
m
et
h
o
d
s
an
d
u
p
to
2
2
%
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
P
r
ed
ictive
a
n
d
fa
u
lt
-
to
lera
n
t
virt
u
a
l m
a
ch
in
e
mig
r
a
tio
n
fo
r
en
erg
y
-
efficien
t c
lo
u
d
…
(
R
a
k
s
h
a
n
G.
K
.
)
1521
co
m
p
ar
ed
with
s
tatic
th
r
esh
o
l
d
m
ig
r
atio
n
.
T
h
ese
f
in
d
in
g
s
alig
n
with
Ma
et
a
l.
[
3
]
,
w
h
e
r
e
ad
ap
tiv
e
en
e
r
g
y
th
r
esh
o
ld
s
ac
h
ie
v
ed
a
n
a
v
er
ag
e
1
5
.
4
9
%
r
ed
u
ctio
n
.
T
h
e
im
p
r
o
v
em
e
n
t
s
tem
s
f
r
o
m
d
y
n
am
ic
h
o
s
t
r
ec
lass
if
icatio
n
to
B
L
UE
,
G
R
E
E
N,
an
d
R
E
D
s
tates,
wh
ich
m
in
im
izes
r
ed
u
n
d
a
n
t
m
ig
r
atio
n
s
an
d
co
n
s
o
lid
ates
wo
r
k
lo
ad
s
o
n
ly
wh
e
n
EC
ex
ce
ed
s
th
e
ad
ap
tiv
e
lim
it
o
f
·
max
.
T
h
is
im
p
r
o
v
em
en
t
is
p
r
im
a
r
ily
d
u
e
t
o
p
r
o
ac
tiv
e
co
n
s
o
lid
atio
n
en
a
b
l
ed
b
y
wo
r
k
lo
ad
f
o
r
ec
asti
n
g
,
wh
ich
p
r
ev
e
n
ts
u
n
n
ec
ess
ar
y
h
o
s
t
ac
tiv
atio
n
an
d
r
ed
u
ce
s
en
er
g
y
s
p
ik
es.
Fig
u
r
e
2
.
C
o
m
p
a
r
is
o
n
o
f
av
er
a
g
e
E
C
(
in
k
W
h
)
am
o
n
g
d
if
f
er
e
n
t V
M
m
ig
r
atio
n
s
tr
ateg
ies
5
.
3
.
M
ig
ra
t
i
o
n f
re
qu
ency
a
nd
s
t
a
bil
it
y
As
s
h
o
wn
in
Fig
u
r
e
3
,
th
e
p
r
o
p
o
s
ed
m
et
h
o
d
r
e
d
u
ce
d
th
e
n
u
m
b
e
r
o
f
VM
m
ig
r
atio
n
s
b
y
ap
p
r
o
x
im
ately
8
0
%
co
m
p
ar
e
d
with
d
y
n
a
m
ic
th
r
esh
o
ld
tec
h
n
iq
u
es
an
d
b
y
o
v
er
8
3
%
co
m
p
ar
ed
with
s
tatic
p
o
licies,
clo
s
ely
m
atch
in
g
th
e
8
3
.
3
2
%
r
e
d
u
ctio
n
o
b
s
er
v
ed
b
y
Ma
et
a
l.
[
3
]
.
T
h
e
SS
-
C
AU
co
r
r
elatio
n
-
b
ased
s
elec
tio
n
alg
o
r
ith
m
co
n
tr
ib
u
te
d
s
ig
n
if
ican
tly
b
y
p
r
io
r
itizin
g
VM
s
m
o
s
t
co
r
r
elate
d
with
h
o
s
t
o
v
er
lo
ad
wh
ile
m
in
im
izin
g
d
ata
tr
a
n
s
f
er
o
v
er
h
ea
d
.
T
h
is
r
e
d
u
ce
d
o
s
cillato
r
y
m
ig
r
atio
n
s
a
n
d
im
p
r
o
v
ed
h
o
s
t
u
tili
za
tio
n
s
tab
ilit
y
.
T
h
e
r
ed
u
ctio
n
is
f
u
r
t
h
er
attr
ib
u
ted
to
p
r
e
d
ictiv
e
m
ig
r
atio
n
tim
in
g
,
wh
ich
a
v
o
id
s
r
ea
ctiv
e
o
s
cillatio
n
s
co
m
m
o
n
l
y
o
b
s
er
v
ed
in
t
h
r
esh
o
ld
-
b
ased
a
p
p
r
o
ac
h
es.
Fig
u
r
e
3
.
C
o
m
p
a
r
is
o
n
o
f
th
e
n
u
m
b
er
o
f
VM
m
ig
r
ati
o
n
s
p
er
2
4
-
h
o
u
r
s
im
u
latio
n
p
er
io
d
ac
r
o
s
s
d
if
f
er
en
t
m
ig
r
atio
n
s
tr
ater
g
ies
5
.
4
.
SL
A
v
io
la
t
io
n r
a
t
e
(
SL
AV)
T
h
e
SLA
V
,
d
ef
i
n
ed
as
t
h
e
p
e
r
ce
n
tag
e
o
f
task
s
ex
ce
e
d
in
g
p
er
f
o
r
m
a
n
ce
co
n
s
tr
ain
ts
,
was
r
ed
u
ce
d
b
y
7
%
to
9
%
co
m
p
ar
ed
with
ex
is
tin
g
ap
p
r
o
ac
h
es
as
s
h
o
wn
i
n
F
ig
u
r
e
4
.
T
h
e
im
p
r
o
v
em
en
t
r
esu
lts
f
r
o
m
p
r
o
ac
tiv
e
m
ig
r
atio
n
tr
ig
g
er
s
en
ab
led
b
y
f
o
r
ec
asted
u
tili
za
tio
n
an
d
f
ailu
r
e
r
is
k
s
co
r
es,
en
s
u
r
in
g
t
h
at
h
o
s
ts
r
em
ain
b
elo
w
th
e
s
af
e
en
er
g
y
th
r
esh
o
ld
max
af
ter
r
ea
llo
ca
tio
n
.
T
h
e
I
E
AB
F
h
o
s
t
s
elec
tio
n
p
r
o
ce
s
s
f
u
r
th
e
r
m
ain
tain
ed
SLA
ad
h
er
en
ce
b
y
p
r
ev
e
n
tin
g
o
v
e
r
lo
ad
p
r
o
p
a
g
atio
n
d
u
r
in
g
co
n
s
o
lid
atio
n
.
B
y
an
ticip
atin
g
o
v
er
lo
ad
c
o
n
d
itio
n
s
,
PVMM
r
ed
u
ce
s
d
elay
ed
m
ig
r
atio
n
s
,
wh
ich
ar
e
a
p
r
im
ar
y
ca
u
s
e
o
f
SLA
v
io
latio
n
s
in
r
ea
ctiv
e
s
y
s
tem
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N:
1
6
9
3
-
6
9
3
0
TEL
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
,
Vo
l.
2
4
,
No
.
5
,
Octo
b
e
r
2
0
2
6
:
1
5
1
3
-
1
5
2
5
1522
Fig
u
r
e
4
.
Sh
o
ws th
e
SLA
v
io
l
atio
n
r
ates f
o
r
v
ar
io
u
s
m
i
g
r
atio
n
p
o
licies
5
.
5
.
F
a
ilu
re
predict
io
n per
f
o
rm
a
nce
T
h
e
RF
an
d
XGBo
o
s
t
m
o
d
els,
tr
ain
ed
o
n
th
e
Go
o
g
le,
Mu
s
tan
g
an
d
T
r
in
ity
tr
ac
es,
d
em
o
n
s
tr
ated
s
u
p
er
io
r
p
r
e
d
ictiv
e
ac
cu
r
ac
y
co
m
p
ar
ed
t
o
b
aselin
e
class
if
i
er
s
,
as
s
h
o
wn
in
Fig
u
r
e
5
.
Am
o
n
g
th
em
,
th
e
RF
m
o
d
el
ac
h
ie
v
ed
a
n
ac
cu
r
ac
y
o
f
9
4
.
1
%,
p
r
ec
is
io
n
o
f
9
2
.
8
%,
r
ec
all
o
f
9
4
.
8
%
a
n
d
a
n
F1
-
s
co
r
e
o
f
9
3
.
6
%.
T
h
ese
r
esu
lts
v
alid
ate
th
e
ef
f
ec
ti
v
en
e
s
s
o
f
m
u
lti
-
tr
ac
e
tr
ai
n
in
g
i
n
ca
p
tu
r
in
g
d
iv
er
s
e
w
o
r
k
lo
a
d
b
e
h
a
v
io
r
s
,
en
a
b
lin
g
th
e
m
o
d
el
to
ac
cu
r
atel
y
an
ticip
ate
VM
-
lev
el
f
ailu
r
es.
T
h
e
r
esu
ltin
g
f
ailu
r
e
r
is
k
p
r
e
d
ictio
n
s
f
ac
i
litated
p
r
e
-
em
p
tiv
e
m
ig
r
atio
n
s
,
th
er
e
b
y
r
e
d
u
cin
g
p
o
ten
tial ser
v
ice
in
ter
r
u
p
tio
n
s
a
n
d
im
p
r
o
v
in
g
o
v
e
r
all
s
y
s
tem
r
esil
ien
ce
.
Fig
u
r
e
5
.
Sh
o
ws th
e
co
m
p
a
r
is
o
n
o
f
f
ailu
r
e
p
r
ed
ictio
n
m
o
d
el
p
er
f
o
r
m
an
ce
ac
r
o
s
s
class
if
ier
s
5
.
6
.
Reso
urce
f
o
re
c
a
s
t
ing
a
c
cura
cy
T
h
e
GR
U
-
b
ased
f
o
r
ec
asti
n
g
m
o
d
u
le
o
u
tp
er
f
o
r
m
e
d
tr
ad
itio
n
al
tim
e
-
s
er
ies
m
o
d
els
s
u
ch
as
au
to
r
eg
r
ess
iv
e
in
teg
r
ated
m
o
v
in
g
av
er
a
g
e
(
AR
I
MA
)
an
d
Ho
lt
-
W
in
ter
’
s
ex
p
o
n
en
tial
s
m
o
o
th
in
g
.
T
h
e
m
ea
n
ab
s
o
lu
te
p
er
ce
n
tag
e
er
r
o
r
(
MA
PE)
was
3
.
8
%
f
o
r
C
PU
an
d
4
.
6
%
f
o
r
m
em
o
r
y
u
til
izatio
n
p
r
ed
ictio
n
s
,
co
m
p
ar
ed
with
7
%
to
9
%
f
o
r
AR
I
MA
.
Acc
u
r
ate
s
h
o
r
t
-
ter
m
f
o
r
ec
asti
n
g
f
ac
ilit
ated
tim
ely
ac
tiv
atio
n
o
f
th
e
m
ig
r
atio
n
s
ch
ed
u
ler
,
im
p
r
o
v
in
g
lo
ad
b
alan
ce
an
d
p
r
ev
en
tin
g
o
v
er
lo
a
d
p
r
o
p
a
g
atio
n
.
T
h
is
im
p
r
o
v
e
m
en
t
is
d
u
e
to
th
e
GR
U
m
o
d
el’
s
ab
ilit
y
t
o
ca
p
t
u
r
e
tem
p
o
r
al
d
e
p
en
d
e
n
c
ies
in
wo
r
k
lo
ad
p
atter
n
s
,
en
a
b
lin
g
m
o
r
e
ac
cu
r
ate
s
h
o
r
t
-
ter
m
p
r
ed
ictio
n
s
.
5
.
7
.
Co
m
pa
ra
t
iv
e
perf
o
r
m
a
nce
s
um
m
a
ry
T
ab
le
1
p
r
esen
ts
a
co
m
p
ar
ativ
e
s
u
m
m
ar
y
o
f
th
e
p
r
o
p
o
s
ed
PVMM
f
r
am
ewo
r
k
ag
ain
s
t
th
r
e
e
b
aselin
e
s
tr
ateg
ies
-
Static
T
h
r
esh
o
ld
[
2
]
,
Dy
n
am
ic
I
QR
-
b
ased
m
ig
r
a
tio
n
[
8
]
an
d
C
F
-
L
A
m
ig
r
atio
n
[
1
1
]
.
T
h
e
r
esu
lts
clea
r
ly
d
em
o
n
s
tr
ate
th
e
s
u
p
er
io
r
p
er
f
o
r
m
a
n
ce
o
f
PVMM
ac
r
o
s
s
all
ev
alu
ated
m
etr
ics.
S
p
ec
if
ically
,
PVMM
ac
h
iev
es
th
e
h
ig
h
est
en
er
g
y
r
ed
u
ctio
n
o
f
1
8
.
2
%,
o
u
t
p
er
f
o
r
m
in
g
C
F
-
L
A
an
d
I
QR
-
b
ased
ap
p
r
o
ac
h
es
b
y
4
.
4
%
an
d
6
.
7
%,
r
esp
ec
tiv
el
y
.
T
h
e
f
r
am
ewo
r
k
also
r
ed
u
ce
s
th
e
f
r
e
q
u
en
cy
o
f
VM
m
ig
r
atio
n
s
b
y
8
3
.
3
%
c
o
m
p
ar
e
d
to
s
tatic
th
r
esh
o
ld
m
eth
o
d
s
,
in
d
ic
atin
g
im
p
r
o
v
ed
s
y
s
tem
s
tab
ilit
y
an
d
f
ewe
r
r
ed
u
n
d
a
n
t m
ig
r
ati
o
n
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.