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.
8
,
No.
6
,
D
ece
m
ber
201
8
, pp.
4398
~
44
11
IS
S
N: 20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v8
i
6
.
pp
4398
-
44
11
4398
Journ
al h
om
e
page
:
http:
//
ia
es
core
.c
om/
journa
ls
/i
ndex.
ph
p/IJECE
Profit
Driv
en De
cision As
sist Syst
em to Se
lect Effi
cient
IaaS Pro
viders
Mohan
Mur
thy
MK
1
,
S
an
j
ay H
A
2
,
Supr
ee
th
B
M
3
1,2
Depa
rtment
of
Inform
at
ion
Sci
enc
e
and
Engi
n
e
eri
ng,
Nitte
Me
e
nakshi
Insti
tut
e
of
Technol
og
y
,
I
ndia
3
Mis
y
s Fina
nc
ial
Software
(India) Pvt
L
td, Indi
a
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Oct
8
, 2
01
7
Re
vised
Jan
2
, 201
8
Accepte
d
Ja
n 1
6
, 2
01
8
Iaa
S
provid
ers
provide
infra
str
uct
ure
to
the
en
d
users
with
v
ar
ious
pric
ing
sche
m
es
and
mode
ls.
They
pr
ovide
diffe
r
ent
t
y
pes
of
virt
u
al
m
ac
hines
(sm
al
l,
m
edi
um
,
la
rge
,
etc.
)
.
Since
e
ac
h
Ia
aS
provide
r
uses
the
ir
own
pric
in
g
sche
m
es
and
m
odel
s,
pri
ce
v
ari
es
from
one
provide
r
to
th
e
oth
er
fo
r
the
sam
e
req
uire
m
ent
s.
T
o
sele
c
t
a
b
est
I
aa
S
provide
r
,
th
e
end
users
ne
e
d
to
consider
var
ious p
ar
amet
ers
such
as
SLA,
pric
ing
m
odel
s/
sche
m
es,
VM
he
te
roge
n
ei
t
y
,
et
c
.
Since
m
an
y
par
amete
rs
ar
e
i
nvolve
d,
sel
ectin
g
an
eff
icie
nt
I
a
aS
provide
r
is
a
cha
l
le
nging
job
for
an
end
user.
To
addr
ess
thi
s
issue,
in
th
is
work
we
have
design
ed,
i
m
ple
m
ent
ed
and
te
sted
a
d
ec
isio
n
-
assist
s
y
stem
which
assists
the
end
users
to
sele
ct
eff
icient
Iaa
S
provide
r(s).
Our
dec
ision
-
as
sist
sy
st
em
consis
ts
of
an
an
aly
tica
l
m
odel
to
ca
l
culate
the
co
st
and
de
ci
sion
s
tra
t
egi
es
t
o
assist
the
end
user
in
sele
cting
the
eff
i
ci
en
t
IaaS
provide
r(s).
The
dec
isi
o
n
assist
sy
st
em
conside
rs
var
ious
rel
eva
n
t
par
amete
rs
such
as
VM
conf
iguration,
pr
ic
e
,
av
ai
l
abi
l
ity
,
et
c
.
to
d
ecide
th
e
eff
i
cient
I
aa
S
provide
r(s).
Rigorous
expe
ri
m
ent
s
have
bee
n
conduc
t
ed
by
emulating
va
rious
Iaa
S
provide
rs,
and
we
have
observe
d
tha
t
our
DA
S
succ
essfull
y
suggests
the
eff
icient
Iaa
S
pr
ovide
r/
prov
ide
r
s
b
y
consid
eri
ng
the
inp
ut
pa
ramet
ers
giv
e
n
b
y
the user.
Ke
yw
or
d:
Cl
oud
Decisi
on assist
IaaS
pro
vid
e
rs
IaaS s
el
ect
io
n
Copyright
©
201
8
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
:
Moh
a
n
M
ur
t
hy MK,
Dep
a
rtm
ent o
f Info
rm
at
ion
Sc
ie
nce and
En
gi
neer
i
ng,
Nitt
e Me
enak
s
hi Insti
tute
of
Tech
no
l
og
y,
Go
ll
ahall
i,
Yel
ahan
a
ka, B
en
ga
lur
u,
Karnata
ka 560
064, I
nd
ia
.
Em
a
il
:
m
aakem
@g
m
ail.co
m
1.
INTROD
U
CTION
Cl
oud
com
pu
ti
ng
is
the
dis
tribu
te
d
com
pu
ti
ng
m
od
el
wh
ic
h
pro
vid
e
s
com
pu
ti
ng
f
aci
li
ti
es
an
d
resou
rces
to
th
e
us
e
rs
in
an
on
-
dem
and
pa
y
-
as
-
you
-
go
m
od
el
[1
]
.
Cl
ou
d
c
om
pu
ti
ng
pro
vi
des
the
facil
it
y
to
acce
ss sh
a
red
r
eso
ur
ces a
nd
c
omm
on
inf
rast
ru
ct
ur
e,
offer
i
ng
ser
vices on de
m
and
over
t
he
n
et
w
ork
to
pe
rfor
m
op
e
rati
ons
that
m
eet
chan
gi
ng
busi
ness
nee
ds
[
2].
Users
a
re
m
ov
ing
to
w
ard
s
cl
oud
bec
ause
it
offe
rs
s
ever
al
ben
e
fits
suc
h
as
el
ast
ic
it
y,
m
ai
ntenan
ce
f
ree,
c
os
t
ef
fec
ti
ven
ess
,
et
c.
It
pro
vid
es
a
higher
Q
oS
t
han
a
tradit
ion
al
soft
war
e
m
od
el
with
le
ss
init
ia
l
i
nv
e
stm
ent.
Based
on
the
ty
pe
of
ser
vices
pro
vid
e
d
in
the
cl
oud
par
a
dig
m
,
thre
e
i
m
po
rtant
se
rv
ic
e
m
od
el
s
are
de
fine
d:
S
of
t
war
e
as
a
Ser
vice
(S
aa
S),
I
nfrastr
uctu
re
as
a
Ser
vice
(I
aaS
),
and
Plat
f
or
m
as
a
Serv
ic
e
(
P
aaS).
I
n
Iaa
S,
i
nfrastr
uctu
re
s
uch
as
c
om
pu
ti
ng
resou
rces
(
Virtual
Ma
chines
),
st
orage
sp
ace
,
net
work
,
et
c.
a
re
giv
e
n
as
se
r
vices.
VM
sel
ect
ion
is
a
c
om
plica
te
d
ta
sk
in
cl
oud
com
pu
ti
ng
e
nvir
on
m
ent
bec
ause
t
her
e
a
re
m
any
al
te
rn
at
iv
e
VMs
with
var
yi
ng
capa
c
it
ie
s
[
3
]
.
Si
nc
e
IaaS
pro
vid
er
s
us
e
t
heir
own
pri
ci
ng
sc
hem
es
and
m
od
el
s,
pri
ce
va
ries
f
r
om
on
e
pro
vid
e
r
to
the
ot
her
f
or
th
e
sam
e
requirem
ents.
We
ha
ve
c
onduct
ed
a
detai
le
d
s
urvey
[
4
]
of
diff
e
re
nt
IaaS
pro
vid
er
s.
T
he
nex
t
few
pa
ra
gr
a
ph
s
br
ie
f
ou
r
im
po
r
ta
n
t fin
dings
of
the s
urvey.
IaaS
pro
vid
e
rs
pro
vid
e
i
nfras
tructu
re
to
the
custom
ers
wit
h
var
i
ou
s
pri
ci
ng
opti
ons.
For
ex
am
ple,
pay
-
as
-
you
-
go
m
od
el
;
in
this
m
od
el
,
the
us
e
r
will
be
payi
ng
th
e
m
on
ey
f
or
w
hat
he
has
us
e
d.
A
var
ia
ti
on
of
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
& C
om
p
Eng
IS
S
N: 20
88
-
8708
Profit
D
riv
e
n Deci
sio
n
Assist
S
yst
e
m
to
S
el
e
ct
…
(
Moha
n Murthy
MK
)
4399
the
pay
-
as
-
you
-
go
m
od
el
i
s
also
avail
able
in
wh
ic
h
if
the
use
r
is
interest
ed
in
the
long
-
t
erm
util
iz
ation
of
the
resou
rce,
the
n
init
ia
l
ly
a
one
-
tim
e
su
bs
cri
ption
fee
is
colle
ct
ed
from
the
us
er
with
reduce
d
hourl
y
us
age
charge.
T
his
pr
ic
ing
opti
on
is
cal
le
d
as
su
bs
c
riptio
n
base
d
pr
ic
ing.
W
e
ha
ve
ob
se
rv
e
d
this
m
od
el
in
Am
a
zo
n
EC2.
The
subs
cripti
on
-
base
d
pr
ic
in
g detai
ls
of
Am
azon
EC
2
a
re
giv
e
n
in
Table
1.
Table
1.
Subsc
riptio
n
Fee
D
et
ai
ls
in
Am
azon
Ins
tan
ce ty
p
e
Initial Fee
Linu
x
/Un
ix
us
ag
e
p
er
h
o
u
r
W
in
d
o
ws
u
sag
e per ho
u
r
1
-
y
ear
ter
m
3
-
y
ears
ter
m
S
m
all
$
2
2
7
.5
$350
$
0
.03
$
0
.05
Lar
g
e
$910
$
1
4
0
0
$
0
.12
$
0
.20
Extra L
arge
$
1
8
2
0
$
2
8
0
0
$
0
.24
$
0
.40
IaaS
prov
i
der
s
pro
vid
e
diff
e
r
ent
ty
pes
of
vi
rtual
m
achines.
For
insta
nc
e,
Am
azon
E
C2
pro
vid
e
s
sm
a
ll
,
la
rg
e,
e
xtra
-
la
rg
e
ty
pe
s
of
VMs
(
Virt
ual
Ma
chines).
The
pr
ic
in
g
de
ta
il
s
of
these
VM
ty
pes
are
giv
en
in
Table
2.
Few
I
aaS
pr
ov
i
der
s
offer
a
disc
ount
on
the
total
bill
ed
am
ou
nt.
F
or
instanc
e,
t
he
disc
ount
detai
ls
of
the
IaaS
P
r
ov
i
der
Cl
ou
d
Sig
m
a
is
giv
en
in
Table
3.
So
m
e
IaaS
pr
ov
i
de
rs
gi
ve
the
op
t
ion
to
e
nd
us
e
rs
to
config
ur
e
the
VMs
wh
il
e
cre
at
ing
them
.
I
n
this
case,
e
nd
us
ers
ca
n
c
onf
igure
the
RA
M,
CPU
,
a
nd
Stor
a
ge
Sp
ace
of
t
he
VM.
T
his
ty
pe
of
c
onfig
urabl
e
VM
opti
on
i
s
obser
ved
i
n
Cl
oudS
igm
a.
In
su
c
h
cases
,
pr
ic
in
g
will
b
e
at
t
he
m
or
e
gran
ular
level.
Table
2
.
Pr
ic
in
g detai
ls of A
m
azon
EC
2 V
Ms
Ins
tan
ce ty
p
e
Linu
x
/Un
ix
us
ag
e
W
in
d
o
ws u
sag
e
S
m
all
$
0
.08
5
per ho
u
r
$
0
.12
per ho
u
r
Lar
g
e
$
0
.32
per ho
u
r
$
0
.48
per ho
u
r
Extra L
arge
$
0
.64
per ho
u
r
$
0
.96
per ho
u
r
Table
3
.
Disco
un
t
detai
ls i
n
C
loud Si
gm
a
Du
ration
in
m
o
n
th
s
1
3
6
12
24
36
% d
isco
u
n
t
0
3
10
25
35
45
IaaS
pro
vid
e
rs
li
ke
Ra
ck
Sp
ac
e,
Am
azon
EC2
,
pro
vid
e
fixe
d
VMs
wh
e
re
the
capa
ci
ty
of
the
VM
i
s
pr
e
def
i
ned
,
a
nd
t
he
e
nd
us
er
will
not
ha
ve
a
ny
opti
on
t
o
c
hange
it
.
D
ue
to
t
he
v
ast
div
e
rsity
of
the
ava
il
able
c
loud se
rv
ic
es
,
f
r
om
the cu
sto
m
er’
s poi
nt of
view, i
t
is ve
ry
dif
ficult
to dec
ide wh
os
e se
rvi
ces they
sho
uld
use
and
wh
at
is
th
e
basis
for
the
ir
sel
ect
ion
[
5
].
Sele
ct
ing
ef
fici
ent
IaaS
pro
vid
e
rs
is
a
te
dio
us
j
ob
f
or
t
he
en
d
us
ers
since
h
e/
sh
e m
us
t co
ns
ider va
rio
us
para
m
et
ers
li
ke
S
LA,
pri
ci
ng m
od
el
s/sc
hem
es, and
diff
e
ren
t t
ypes of
VM
insta
nces.
A
decisi
on
assi
st
syst
e
m
wh
ic
h
assist
s
the
use
r
to
sel
ect
e
ff
i
ci
e
nt
IaaS prov
iders
m
akes
th
e
en
d
us
er
jo
b
easi
er
.
In
this
w
ork,
we
ha
ve
desig
ned,
im
ple
m
ented
an
d
te
ste
d
a
decisi
on
-
assi
st
syst
e
m
(D
AS
).
T
he
DAS
consi
sts
of
an
a
naly
ti
cal
m
od
el
to
com
pu
te
the
cost
an
d
dec
isi
on
strat
egies
to
assist
the
end
us
er
i
n
sel
ect
ing
th
e
ef
fici
ent
I
a
aS
pr
ov
i
der(s).
The
D
AS
has
the
us
er
inter
face
to
capt
ur
e
the
e
nd
us
er
requirem
ents.
Af
te
r
captu
rin
g
the
requirem
ents,
us
in
g
the
anal
yt
ic
al
m
od
el
and
decisi
on
strat
egies
the
syst
e
m
wil
l
s
uggest
eff
ic
ie
nt
Iaa
S
pro
vid
er
(s
)
bas
ed
on
the
us
e
r
req
uirem
ents.
W
e
ha
ve
c
onsidere
d
the
fo
l
lowing
par
am
et
ers
to
dev
el
op
analy
ti
cal
m
od
el
a
nd
decisi
on strate
gies
a.
The req
uirem
ents
s
uch as m
em
or
y, CPU,
st
or
a
ge,
et
c.
b.
Ten
ur
e,
which
play
s an i
m
po
rtant role
in se
le
ct
ing
the
prici
ng sc
hem
e.
c.
VM h
et
e
roge
ne
it
y.
d.
Diff
e
re
nt prici
ng sc
hem
es su
ch
as
pay
-
as
-
y
ou
-
go
,
s
ub
s
cri
ption,
et
c.
e.
QoS
par
am
et
ers
su
c
h
as
se
rv
e
r
a
vaila
bili
ty
a
nd V
M i
niti
at
i
on tim
e.
f.
Locati
on
of
t
he
d
at
a ce
nter.
The
decisi
ons
of
the
D
AS
will
be
accurate
on
ly
if
the
Iaa
S
prov
i
der
in
f
or
m
at
ion
is
up
-
to
-
dat
e.
A
ny
changes
i
n
the
par
am
et
ers
(which
a
re
go
i
ng
to
af
fect
the
de
ci
sion
of
s
el
ect
ing
ef
fici
ent
IaaS
pro
vid
er
)
at
the
IaaS
prov
i
der
s
shou
l
d
b
e
ref
l
ect
ed
in
real
ti
m
e
at
DAS
.
O
ther
wise
,
the
D
AS
will
us
e
ob
so
le
te
data
to
decid
e
the
eff
ic
ie
nt
Ia
aS
pro
vid
e
r
w
h
ic
h
re
su
lt
s
in
an
inacc
ur
at
e
de
ci
sion
.
T
o
ad
dr
ess
t
his
issue
,
we
ha
ve
de
ve
lop
e
d
the
we
bs
e
rv
ic
e
APIs
w
hic
h
a
re
ex
pose
d
by
the
D
AS.
Th
e
se
APIs
ar
e
use
d
to
receive
t
he
in
form
at
ion
and
helps
the
D
AS
to
be
i
n
sy
nchronizat
ion
wit
h
the
i
nfor
m
at
ion
a
vaila
ble
at
t
he
IaaS
pro
vide
rs.
By
a
greei
ng
t
o
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
20
88
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
2
01
8
:
4398
-
4411
4400
pro
vid
e
the
i
nfor
m
at
ion
ab
out
the
VM
insta
nce
ty
pes
a
nd
the
pri
ci
ng
det
ai
ls
,
an
Iaa
S
pro
vid
e
r
ca
n
at
tract
a
la
rg
e
nu
m
ber
of
cu
stom
ers
if
qu
al
it
y
ser
vice
is
pro
vid
e
d
at
the
rea
sona
ble
cost.
Using
th
e
DAS,
e
nd
us
e
rs
ca
n
set
th
ei
r
pr
i
or
it
ie
s
on
di
ff
e
ren
t
par
am
et
ers
(p
r
ic
e,
avail
abili
ty
,
et
c.)
to
sear
ch
the
m
os
t
su
it
able
IaaS
pro
vi
der
s.
The
DAS
al
so
pro
vid
es
a
n
opti
on
to
relax
the
searc
h
c
rite
ria
on
diff
e
re
nt
pa
ram
et
ers.
For
e
xam
ple,
if
the
us
ers
are
lo
ok
i
ng
f
or
a
VM
with
m
e
m
or
y
in
a
ce
rtai
n
range
(
rathe
r
t
han
a
fixe
d
num
ber
),
they
ca
n
use
the
op
ti
on
of
m
e
m
or
y
var
ia
ti
on
wh
ic
h
is
pr
ov
i
ded
by
DA
S
.
The
op
ti
ons
of
set
ti
ng
pri
ori
ti
es
and
va
riat
ion
s
on
diff
e
re
nt p
a
ra
m
et
ers
m
ake
th
e DAS
flexi
ble and
us
er
-
f
rien
dly
.
To
te
st
our
D
AS
,
we
ha
ve
e
m
ulate
d
var
i
ou
s
IaaS
pr
ov
i
der
s
,
an
d
dif
fe
ren
t
po
pu
la
r
s
cenari
os
are
te
ste
d.
I
n
t
he
te
ste
d
sce
nar
i
os,
we
ha
ve
fou
nd
that
t
he
D
A
S
pro
vid
e
d
m
os
t
eff
ic
ie
nt
Iaa
S
pro
vid
e
r
/
pro
vid
e
rs
consi
der
i
ng
t
he
diff
e
ren
t
in
put
par
am
et
ers.
Re
st
of
th
e
pa
per
is
orga
nize
d
as
fo
ll
ows
.
S
ect
ion
2,
br
ie
fs
about
the
relat
ed
wor
k,
sect
ion
3
giv
es
an
over
vie
w
of
the
decisi
on
assist
syst
em
,
sect
ion
4
and
sect
io
n
5
exp
la
in
s
the
analy
ti
cal
m
od
el
and
dec
isi
on
strat
e
gies
resp
ect
iv
el
y,
sec
ti
on
6
gi
ves
the
detai
ls
of
the
ex
per
im
ents
an
d
resu
lt
s,
foll
owed by a
c
on
cl
usi
on
.
2.
RELATE
D
W
ORK
The
w
ork
by
S
.
K.
Ga
rg
et
al
[5
]
pr
ese
nts
a
fr
am
ewo
r
k
(SM
ICloud)
to
r
ank
cl
oud
ser
vi
ce
pro
vid
er
s
base
d
on
pe
rfo
rm
ance
m
et
ric
s
li
ke
su
sta
ina
bili
ty
,
su
it
ability
,
sta
bili
ty
,
etc.
auth
ors
ha
ve
desig
ned
A
na
ly
ti
ca
l
Hierarc
hical
P
ro
ces
s
(
A
HP)
base
d
ra
nkin
g
m
echan
is
m
to
c
om
par
e
di
f
fer
e
nt
cl
ou
d
s
erv
ic
es.
T
he
work
by
Mi
chael
Sm
i
t
et
al
[
6
]
prese
nts
a
m
et
ho
dolog
y
a
nd
an
im
plem
entat
ion
of
a
se
rv
ic
e
-
ori
ented
a
ppli
cat
ion
that
pro
vid
es
rele
va
nt
m
e
ta
data
inf
or
m
at
ion
de
scribin
g
offere
d
cl
oud
se
r
vi
ces
via
a
un
i
form
RESTfu
l
web
serv
ic
es
.
This
work
co
nce
ntr
at
es
on
ly
in
fet
chin
g
the
in
f
orm
at
ion
us
in
g
web
s
er
vices.
Sele
ct
ing
the
be
st
IaaS
pro
vid
er
s
acco
rd
i
ng
to
the
use
r
requirem
ent
is
no
t
add
res
s
ed.
The
work
by
Dh
aval
Lim
b
ani
et
al
[
7
]
proposes
a
ser
vice
br
ok
e
r
f
or
the
sel
ect
ion
of
data
ce
nter
ba
sed
on
t
he
la
te
ncy
of
the
us
er
re
quest
s
.
The
w
ork
c
on
s
ider
s
thec
os
t
only
wh
en
m
or
e
than
on
e
datace
nters
ha
ve
the
lo
w
est
la
te
ncy
wit
h
in
a
reg
i
on
.
I
n
this
work,
only
the
pro
blem
o
f
sel
ect
ing
a
n
e
ff
ec
ti
ve
datace
nter
is ad
dr
esse
d.
The
w
ork
by
Stel
la
Gatzi
u
Gr
i
vas
et
al
[8
]
pr
opose
s
a
cl
oud
b
roker
w
hich
ha
s
know
le
dg
e
of
the
su
pp
or
te
d
busi
ness
processes
,
the
e
xisti
ng
serv
ic
e
offer
i
ngs
f
r
om
the
m
ark
et
p
la
ce,
t
he
cu
rrent
r
el
at
ion
s
betwee
n
the
business
proces
ses
and
t
he
cl
oud
se
rv
ic
es
.
Cl
oud
brok
e
r
m
a
nag
e
s
a
reposit
or
y
of
al
l
provi
ders
and
se
rv
ic
es
w
hich
are
releva
nt
to
the
value
chain
of
a
co
m
pan
y.
In
this
work,
d
if
fer
e
nt
VM
heteroge
neity
,
pr
ic
in
g
m
od
el
s
,
an
d
sc
hem
es,
VM
init
ia
ti
on
tim
e
are
no
t
consi
der
e
d
m
or
eo
ver
the
wor
k
is
in
t
he
pro
po
s
al
sta
ge
,
im
ple
m
entat
ion
is
not
done.
T
he
w
ork
by
S
rij
it
h
K.
Nai
r
et
al
[9
]
descr
i
bes
th
e
con
ce
pts
of
cl
oud
burstin
g,
cl
ou
d
bro
ke
rag
e
a
nd
discu
sses
t
he
secu
rity
issues
as
s
ociat
ed
with
the
tw
o
m
od
el
s.
T
he
cl
oud
bro
ker
a
ge
m
odel
do
e
s
no
t
ha
ve
the
a
bili
ty
to
giv
e
ef
fici
ent
cl
oud
pro
vid
e
r
s
by
c
onside
rin
g
us
er
re
quire
m
ents
since it
is only
serv
ic
in
g base
d on sto
rag
e
and c
om
pu
ti
ng
use
case sce
nar
i
os
.
T
he
w
ork
by
Ma
y
Al
-
Roo
m
i
et
al
[10
]
fo
c
us
es
on
com
par
in
g
m
any
e
m
p
loye
d
an
d
pro
pose
d
pr
ic
in
g
m
od
el
s
te
chn
iq
ues
a
nd
highli
gh
ts
t
he
pr
os
a
nd
c
ons
of
eac
h.
T
he
c
om
par
ison
is
base
d
on
m
any
aspects
su
c
h
as
fairn
es
s,
pri
ci
ng
a
ppro
ac
h,
and
u
ti
li
zat
io
n
pe
rio
d.
I
n
th
is
wo
r
k
,
t
he
com
par
ison
of
t
he
pri
ci
ng
m
od
el
s
is
m
ad
e
.
They
ha
ve
not
consi
de
red
the
pr
ic
in
g
schem
es,
VM
heter
og
e
neity
and
QoS
of
m
ulti
ple
prov
i
de
rs.
T
he
work
by
Hyu
n
Jin
Mo
on
et
al
[1
1
]
analy
zes
the
pe
rfo
rm
ance
of
resou
r
ce
sche
du
li
ng
po
li
ci
es.
T
hey
hav
e
consi
der
e
d
se
ve
ral
m
od
el
s
and
sche
duli
ng
po
li
ci
es
,
wh
ic
h
are
pro
fit
m
od
el
,
SL
A
m
od
el
,
a
nd
SL
A
-
base
d
sche
du
li
ng.
T
hi
s
wo
r
k
c
on
ce
ntrates
on
opti
m
iz
at
ion
of
co
st.
The
w
ork
by
Linli
n
Wu
et
al
[12
]
has
de
fine
d
m
app
in
g
strat
egy b
y i
nterp
ret
ing
cu
st
om
er r
equ
i
rem
ents to
inf
rastr
uctu
re level
par
am
et
e
rs.
I
t al
so
d
esi
gns a
nd
i
m
ple
m
ents
scheduli
ng
m
echan
ism
s
to
m
axi
m
iz
e
SaaS
prov
i
der’s
prof
it
by
re
duci
ng
t
he
in
fr
a
struct
ure
c
os
t
and
m
ini
m
izing
SL
A
vi
olati
on
s
.
T
his
w
ork
co
nce
ntrates
on
Saa
S
pro
vi
der
s.
Dif
fer
e
nt
VM
heter
ogeneit
y,
QoS
par
am
et
ers,
pri
ci
ng m
odel
s and
pr
ic
in
g schem
es are no
t consi
der
e
d.
In
work
[1
3
]
,
auth
or
s
a
ddr
ess
the
pro
ble
m
of
ser
vice
request
sche
duli
ng
i
n
cl
oud
com
pu
ti
ng
syst
e
m
s.
They
con
si
der
a
t
hr
ee
-
ti
er
cl
ou
d
str
uctur
e
,
w
hich
c
on
sist
s
of
in
fr
a
struct
ure
ve
ndors,
s
erv
ic
e
pro
vid
er
s
,
a
nd
consum
ers.
Th
ey
def
ine
t
he
s
cheduli
ng
strat
egies
to
sat
isfy
the
obj
ect
ive
s
of
se
rv
ic
e
pro
vi
der
s
and
c
onsu
m
ers.
This
w
ork
does
not
ad
dr
es
s
the
issue
of
sel
ect
ing
an
ef
fici
ent
IaaS
pr
ov
i
der
wh
e
n
the
e
nd
us
er
w
a
nts
t
o
a
ppr
oach the
Ia
aS pr
ov
i
der
s
fo
r
the
ser
vice
directl
y
.
Plan
F
or
Cl
oud
[
14
]
is
a
fr
ee
cl
oud
c
os
t
cal
culat
or
w
hich
giv
es
c
os
t
re
po
rts
f
or
de
plo
y
m
ent
op
ti
ons
.
It
sel
ect
s
a
server
base
d
on
RAM
and
CP
U
count
an
d
li
sts
the
resu
lt
ing
s
erv
e
r
of
only
a
few
pro
vid
e
rs
.
W
e
hav
e
obser
ve
d
that
cost
repo
rt
g
ene
rati
on
opti
on
is
a
vaila
ble
only
fo
r
3
ye
ars
durati
on.
The
cha
nges
at
the
IaaS
pr
ov
i
der
s
will
no
t
ref
le
ct
i
m
m
ediat
ely
at
the
PlanFo
rClo
ud
we
bs
i
te
.
The
Q
oS
par
am
et
ers
li
ke
VM
init
ia
ti
on
tim
e and av
ai
la
bili
t
y are
no
t c
onsidere
d.
In
w
ork
[15
]
,
a
uthors
prese
nt
a
Cl
oud
se
rv
ic
e
sel
ect
ion
fr
a
m
ewo
r
k
that
use
s
a
rec
omm
e
nd
e
r
syst
em
(RS)
wh
ic
h
he
lps
a
us
e
r
to
s
el
ect
the
serv
i
ces
from
diff
er
ent
Cl
oud
pro
vid
e
rs
(CP
).
T
he
RS
rec
omm
ends
a
serv
ic
e
base
d
on
the
netw
ork
QoS
a
nd
Virt
ual
Ma
chin
e
(
VM)
platfo
rm
factors
o
f
diff
e
ren
t
CPs
.
T
he
ranki
ng
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
& C
om
p
Eng
IS
S
N: 20
88
-
8708
Profit
D
riv
e
n Deci
sio
n
Assist
S
yst
e
m
to
S
el
e
ct
…
(
Moha
n Murthy
MK
)
4401
m
et
ho
d
pro
pos
ed
by
aut
hors
on
ly
c
onsider
s
erv
ic
es
’
insi
de
at
tribu
te
sa
nd
igno
re
the
relat
ion
s
be
twee
n
c
on
te
xt
pro
vid
er
s
an
d
co
ns
um
ers.
J
unping
D
ong
et
al
propose
serv
ic
es
rec
om
m
end
at
ion
s
yst
e
m
[1
6]
ba
sed
on
heter
og
e
ne
ou
s
netw
ork
a
naly
sis
in
cl
ou
d
c
om
pu
ti
ng
.
Au
t
hors
pr
opos
e
ser
vice
rec
omm
e
nd
at
io
n
syst
e
m
base
d
on
hete
roge
ne
ou
s
ser
vice
net
work
ra
nkin
g
and
cl
us
te
ri
ng.
In
th
is
w
ork
,
QoS
par
am
et
ers
li
ke
a
vaila
bili
ty
and
VM
init
ia
ti
on
tim
e
are
no
t
c
on
si
der
e
d.
I
n
work
[17
]
aut
hors
pro
pose
a
hiera
rch
ic
al
i
nfor
m
at
ion
m
od
el
f
or
integrati
ng
heteroge
neous
cl
oud
i
nfor
m
at
ion
from
diff
eren
t
prov
i
der
s
a
nd
a
correspo
nd
i
ng
cl
oud
i
nfor
m
at
ion
colle
ct
ing
m
ec
han
ism
.
Also
,
auth
or
s
pro
pos
e
a
prefe
ren
ce
-
awa
re
s
olu
ti
on
e
valuati
on
m
od
e
l
fo
r
e
val
uatin
g
and
rec
omm
en
ding
s
olu
ti
ons
accor
ding
t
o
th
e
prefe
ren
ces
of
ap
plica
ti
on
pro
vid
er
s.
In
thi
s
w
ork
,
a
uthors
us
e
web
pag
e
pars
ing
an
d
web
AP
I
s
invocat
io
n
to
colle
ct
th
e
inform
at
ion
in
real
tim
e.
These
ope
rati
ons
are
trigg
e
red
w
hen
the
us
e
r
re
que
sts
cl
oud
ser
vi
ce.
Parsi
ng
we
bp
a
ge
of
t
he
e
xisti
ng
pro
vid
e
rs
in
realt
im
e
i
n
the
ocean
of
inte
rnet
is
virtu
al
ly
i
m
po
ssible
,
an
d
it
is
err
or
-
pro
ne
.
A
s
the
aut
hors
rig
htly
po
inted
out
in
the
pap
e
r,
on
ly
a
fe
w
Iaa
S provi
der
s
pr
ov
i
de
webser
vice
APIs t
o pro
vid
e c
os
t a
n
d VM in
form
at
io
n.
The
w
ork
do
e
s
no
t
co
ns
i
der
QoS
pa
ram
et
er
s
,
an
d
the
opti
on
of
e
xpos
i
ng
the
webser
vice
AP
Is
i
n
th
e
b
ro
ker
in
g
syst
e
m
is
no
t
c
onsidere
d.
Als
o,
t
he
w
ork
doesn
’t
co
ns
ide
r
the
di
ff
ere
nt p
rici
ng
schem
es
and
pri
ci
ng
m
od
el
s
of
fer
e
d
by
IaaS
prov
i
de
rs
.
M.
W
hai
du
zzam
an
et
a
l
[1
8]
ta
lks
ab
out
m
ulti
-
crit
eria
based
cl
oud
serv
ic
e
sel
ect
ion
.
The
auth
or
s
hav
e
done
a
sur
vey
on
dif
fer
e
nt
m
ul
ti
crit
eria
m
e
tho
ds
wh
ic
h
can
be
us
e
d
to
sel
ect
the
cl
oud
ser
vices.
The
work
doe
sn
’t
ta
lk
a
bout
the
VM
heter
og
e
neity
,
dif
fe
ren
t
pr
ic
in
g
s
ch
em
es
,
and
m
od
el
s
.
Also
,
c
ollec
ti
ng
i
nfor
m
at
ion
from
IaaS
pro
vid
e
rs
is
not
addresse
d
i
n
t
his
work.
T
he
w
ork
[
19
]
is
ab
out
exp
l
oit
ing
perf
or
m
ance
heter
og
e
neity
by
se
le
ct
ing
a
pr
op
e
r
VM
i
n
an
Ia
aS
pro
vid
e
r.
I
n
this
wor
k,
m
ulti
ple
IaaS
pro
vid
e
rs
an
d
diff
e
re
nt
pr
ic
in
g
sc
hem
es
offe
red
by
t
he
sam
e
prov
i
der
are
not
c
onside
red
.
I
nwo
r
k
[20
]
auth
or
s
propos
e
a
br
okera
ge
base
d
arch
it
ect
ur
e
f
or
e
ff
ic
ie
nt
serv
ic
e
sel
ect
ion
.
I
n
this
m
o
del,
the
cl
oud
bro
ke
r
colle
ct
s
and
in
dex
e
s
the
ser
vi
ce
prov
i
der’s
pro
per
ti
es.
T
he
ind
e
x
is
us
e
d
to
ide
ntify
the
best
-
m
at
ched
serv
ic
e
wh
e
n
a
r
e
quest
is receive
d fro
m
the cu
st
om
e
r.
3.
DECISIO
N A
SSIST
SYST
EM (D
AS
)
F
igure
1
s
how
s
the
ov
e
rv
ie
w
of
our
decisi
on
assist
syst
e
m
.
En
d
us
e
rs
will
interact
with
DAS
us
in
g
thin
cl
ie
nts
(
brow
se
rs
).
T
hro
ugh
t
he
web
i
nterf
ace
,
us
e
rs
c
an
lo
gin
a
nd
pr
ov
i
de
their
re
quirem
ents.
The
DAS
has
t
he
f
ollo
wing im
po
rtant c
om
po
ne
nts
a.
Fr
ont
c
ontr
oller
:
All
the
use
r
re
quest
s
a
r
e
receive
d
by
the
f
ront
c
ontr
oller.
It
do
es
t
he
first
le
ve
l
of
screeni
ng.
A
fter
pr
e
-
processi
ng
the
us
e
r
re
qu
e
sts,
it
fo
r
w
ard
s
the
reque
sts
to
the
decis
ion
m
aker
.
Th
e
fron
t c
ontr
oller
al
so
receives
the
resu
lt
s
from
the
decisi
on m
aker
an
d
for
wa
rd
s
it
to the
br
owser
.
b.
Decisi
on
m
aker
:
This
m
odul
e
has
the
bu
siness
lo
gic
(
analy
ti
cal
m
o
del
an
d
decisi
on
strat
egies
).
It
receives
the
pre
-
proce
ssed
re
qu
e
sts
from
th
e
fron
t
co
ntr
oller.
Usi
ng
the
a
naly
ti
cal
m
od
el
,
it
co
m
pu
te
s
the
total
cost
f
or
the
us
er
re
qu
i
re
m
ents.
This
c
om
pu
te
d
cost
is
us
e
d
byd
eci
si
on
strat
egies
t
o
decide
t
he
be
st
IaaS
pr
ov
i
der
.
Decisi
on
st
ra
te
gies
al
so
co
ns
ide
r
the
use
r
input
pa
ram
et
ers
to
deci
de
the
best
Iaa
S
pro
vid
er
.
c.
DB
:
Up
-
to
-
dat
e
inf
or
m
at
ion
of
t
he
IaaS
pr
ov
i
der
s
s
uch
a
s
pr
ovide
r
i
de
ntit
y,
locat
ion,
pr
ic
i
ng
inf
o,
an
d
VM d
et
ai
ls a
re
stor
e
d
i
n
the
DB.
Sy
nc
hron
i
zer a
nd D
eci
si
on m
aker
m
od
ules
directl
y i
nt
eract wit
h DB.
d.
Synchr
on
iz
er
:
The
sy
nchroni
zer
ex
poses
t
he
we
bs
er
vice
A
PI
s.
The
se
A
PIs
m
ake
su
re
a
ny
update
do
ne
at
the
pr
ov
i
der
sit
e
is
autom
at
i
cal
ly
ref
le
ct
ed
in
t
he
syst
em
’s
database
.
P
ub
li
s
hed
we
b
serv
ic
e
APIs
a
r
e
util
iz
ed
by
the I
aaS
pr
ov
i
der
s to
update
a
ny
change
in
t
he
re
le
van
t
in
form
a
ti
on
. A
lso
,
t
he
DAS
has
a
we
b
in
te
rf
ace
opti
on, u
si
ng whic
h
an
a
dm
in can
e
nt
er th
e
IaaS
pro
vid
e
r deta
il
s
m
anu
al
ly
.
Fo
ll
owin
g
ar
e
web
ser
vice
A
PI
s
w
hich
we
hav
e
de
velo
pe
d
f
or
prov
i
der
s
to
update
a
ny
change
i
n
th
e
releva
nt in
for
m
at
ion
.
a.
sen
dS
ta
ti
cVMDetai
ls()
-
T
his
AP
I
is
us
ed
to
sen
d
in
form
at
i
on of
ne
wly cr
eat
ed
sta
ti
c VM
te
m
plate
s.
b.
sen
dP
rice
Detai
ls()
-
This
API i
s u
se
d
t
o
se
nd
pr
ic
e
detai
ls f
or e
xisti
ng
sta
ti
c V
M t
em
plate
s
.
c.
sen
dConfi
gura
ti
on
VM
Detai
ls()
-
This
A
P
I
is
us
e
d
to
send
pr
ic
e d
et
ai
ls f
or
c
onfig
ur
a
ble
V
M t
e
m
plate
s.
d.
delet
eVMDeta
il
s
()
-
T
his
API i
s u
se
d
t
o
se
nd i
nfor
m
at
ion
of t
he dele
te
d
sta
ti
c V
M t
em
plates.
e.
sen
dD
isc
ount
Detai
ls()
-
This
AP
I
is
us
ed
to
sen
d
the
d
isc
ount
detai
ls.
f.
sen
dIniti
al
feeD
et
ai
ls()
-
This
AP
I
is
us
ed
to
sen
d
the
init
ia
l fee
detai
ls.
g.
sen
dAvail
abili
ty
Detai
ls()
-
T
his
A
P
I
is
us
e
d
t
o sen
d
the
av
ai
la
bili
ty
d
et
ai
ls.
h.
sen
dIniti
at
ion
ti
m
eDetai
ls()
-
T
his API
is used
to
se
nd the
VM
init
ia
ti
on
ti
m
e d
et
ai
ls.
The D
AS
als
o pro
vid
es a
sim
ple w
e
b
i
nterfa
ce to m
anu
al
ly
en
te
r
t
he deta
il
s of the
V
M
, pr
ic
ing
sc
hem
es
,
et
c.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
20
88
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
2
01
8
:
4398
-
4411
4402
Figure
1.
Decis
ion
a
ssist
syst
em
4.
ANALYTI
C
A
L MO
DEL
We
ha
ve
dev
e
lop
e
d
a
n
a
nal
yt
ic
al
m
od
el
to
cal
c
ulate
th
e
cost
by
c
on
siderin
g
us
e
r
r
equ
i
rem
ents,
durati
on
of
t
he
ser
vice
re
quir
ed,
VM
heter
ogeneit
y,
pr
ic
in
g
sc
hem
es
,
an
d
m
od
el
s.
In
c
onfig
ur
a
ble
V
Ms
the
pr
ic
e
de
pends
on
the
pr
ic
e
pe
r
unit
of
res
ource
an
d
nu
m
ber
of
un
it
s
of
res
ource
requ
ired
by
the
use
r.
F
or
instance,
i
f
the
pr
ic
e
per
unit
(1
GB)
of
RA
M
is
$0
.
02,
t
he
n
f
or
2
GB
of
RAM
total
pr
i
ce
would
be
$0.
04.
I
n
gen
e
ral, t
he
c
onfi
gurab
le
VM
s co
st ca
n be c
al
culat
ed by
usi
ng
belo
w f
orm
ula.
P
=
Cost
cpu
+C
ost
memory
+Cost
sto
rage
wh
e
re;
Co
st
cpu
=
CPU
N
*CPUP
RICE
U
Cost
memory
=
MEM
N
*MEMP
RI
CE
U
Cost
storage
=
ST
G
N
*S
T
GPR
ICE
U
wh
e
re;
P
=P
ric
e of
VM
CPU
N
=N
um
ber
of
unit
s of C
PU
CPUPR
ICE
U
=Price
per
unit
for
CPU
MEM
N
=N
um
ber
of
unit
s of
m
e
m
or
y
MEMPRICE
U
=Price
per
unit
for
m
e
m
or
y
STG
N
=N
um
ber
of
un
it
s
of sto
r
age
STGPR
ICE
U
=Price
per
unit
for st
orage
Fo
r
f
i
xed (s
ta
ti
c)
VMs the
price
can be
cal
cu
la
te
d
as
P
=
VM
price
wh
e
re;
VM
price
=
Pr
ic
e
of
VM
In
fi
xed
VMs,
the
pr
ic
e
de
pe
nd
s
on
the
pr
ic
e
of
the
VM.
F
or
e.
g.
,
in
Am
a
zon
EC
2,
f
or
s
ta
nd
a
rd
on
-
dem
and
sm
al
l
instance
with
windows
OS
,
t
he
pr
ic
e
is
$0.
12
pe
r
hour.
In
so
m
e
cases,
if
the
us
er
is
i
nt
erested
in
long
te
rm
ut
il
iz
at
ion
of
a
re
so
urce
,
the
n
in
it
ia
lly
a
su
bs
cr
ipti
on
cha
r
ge
will
be
colle
ct
ed
from
the
us
er,
la
te
r
the
hourl
y
us
age
cha
rg
e
will
be
re
du
ce
d.
I
n
few
cases
,
we
hav
e
obser
ved
that
cl
oud
pro
vi
der
s
offe
r
a
dis
count
on the t
otal bil
le
d
am
ou
nt.
T
he
final
pr
ic
e
is
giv
e
n by the
be
low
a
nal
yt
ic
al
m
od
el
.
FP
=I+
(
P*
T)
-
D
wh
e
re;
FP
=Fi
nal pric
e
I
=I
niti
al
f
ee
P
=
Pr
ic
e
of VM
.
T
=D
ur
at
io
n (T
enure)
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
& C
om
p
Eng
IS
S
N: 20
88
-
8708
Profit
D
riv
e
n Deci
sio
n
Assist
S
yst
e
m
to
S
el
e
ct
…
(
Moha
n Murthy
MK
)
4403
D
=Disc
ount
offer
e
d
The
a
naly
ti
cal
m
od
el
co
ns
id
ers
the
diff
e
re
nt
pri
ci
ng
sc
hem
es
and
pr
ic
i
ng
m
od
el
s.
It
al
so
co
vers
th
e
cases
wh
e
re
init
ia
l
fee
and
disco
un
t
com
es
into
the
picture
.
VM
heter
ogen
ei
ty
is
i
m
plicitl
y
con
sidere
d
wh
e
n
cal
culat
ing
t
he
pri
ce
of
t
he
V
M.
The
value
of
P
is
cal
cula
te
d
diff
e
ren
tl
y
base
d
on
the
ty
pe
of
VM
(f
i
xed
or
config
ur
a
ble)
.
5.
DECISIO
N
S
TRATEGIE
S
We
ha
ve
dev
el
op
e
d
al
gorithm
s
to
sel
ect
ef
fici
ent
IaaS
prov
i
der
s
ba
sed
on
the
us
er
r
e
quire
m
ents
a
nd
SLA
par
am
eter
s.
In
t
he
fi
rst
ste
p,
requ
irem
ents
of
the
use
r
a
re
colle
ct
ed
f
rom
the
web
i
nterf
ace
.
The
re
quirem
e
nts
inclu
de
RAM,
CPU
,
sto
rag
e
s
pace,
O
S,
durati
on,
lo
cat
ion
,
a
nd
pr
i
or
it
ie
s
for
the
cost
an
d
QoS
pa
ram
et
er
s
(av
ai
la
bili
ty
&
init
ia
ti
on
ti
m
e).
The
use
r
can
set
the
pri
or
it
ie
s
as
per
hi
s
need
s
.
For
e
xam
ple,
if
the
us
e
r
is
i
nterested
on
ly
in
getti
ng
the
lowe
st
co
st
pro
vid
e
r,
with
out
gi
ving
m
uch
im
po
rtance
t
o
QoS
par
am
et
ers
the
n
he/she
ca
n
s
e
t
the
cost
pri
ori
ty
to
highest.
If
t
he
us
e
r
is
i
nterested
i
n
QoS
,
the
n
he
ca
n
s
et
the
pr
i
or
it
ie
s acco
r
dingly
.
T
he
m
a
in alg
or
it
hm
wi
ll
take u
se
r
in
pu
ts
to deci
de
the
best I
aa
S
prov
i
der
s
.
Algori
th
m
1:
Ma
in
A
l
go
ri
thm
1.
I
npu
ts
: R
A
M, CPU
, S
t
or
a
ge,
OS,
Durati
o
n, L
ocati
on, C
os
t Va
riat
ion,
Pc, Pa,
Pi
2.
Set VM
m
ain
=
[
]
, VM
sublist
=
[
]
,
VM
potential
=
[
]
, V
M
avail
=
0,
V
M
init
=
0,
V
M
cost
=
0,
IaaS
providers
=
[ ]
3.
VM
m
ain
=
read
from
d
at
abas
e as
per
t
he
in
puts R
AM,
CP
U
, S
to
ra
ge,
OS,
Durati
on, L
oca
ti
on
4.
i
f
VM
m
ain
=
∅
5.
I
ndic
at
e this to
the
us
e
r a
nd ex
it
.
6.
e
ndif
7.
f
or
i=
1 t
o n
um
ber
of
VMs
pr
ese
nt in
V
M
m
ai
n
8.
VM
m
ain
[i]
.co
st
=
I
i
+
(
P
i
*T
i
)
-
D
i
9.
e
ndfor
10.
if
[
(P
c
> P
i
) &&
(P
c
> P
a
)]
11. create
V
M
sublist
| VM
sub
list
VM
m
ain
and
VM in VM
sublist
, VM
cost
=M
IN(V
M
cost
)
1
2.
if
|
V
M
sublist
|
=
1
13.
VM
potential
[1]
=
VM
sublist
[1]
14.
el
se
15.
V
M
potential
=
ti
eB
reak
er
1( P
a
,
P
i
,
VM
sublist
)
16.
en
dif
17.
el
sei
f
[(P
i
>
P
c
)
&
&
(P
i
>
P
a
)]
18. VM
m
ain
=
read
F
ro
m
DB(VM
main
[i]
, 2
)
19. c
reate V
M
sublist
| VM
sub
list
VM
m
ai
n
and
V
M i
n
VM
sublist
,
V
M
init
=M
IN(
VM
init
)
20.
if
|
V
M
sublist
|
=
1
21.
VM
potential
[1]
=
VM
sublist
[1]
22.
el
se
23.
VM
potential
=
tieBreake
r2(P
c
, P
a
, VM
sublist
)
24.
e
ndif
25.
el
sei
f
[(P
a
> P
c
) &&
(P
a
>
P
i
)]
26. VM
m
ain
=
read
F
ro
m
DB(VM
main
[i]
, 1
)
27.
c
reate
VM
sublist
| VMsub
li
st
VM
m
ain
and
VM i
n VM
sublist
, V
M
avail
=M
AX(V
M
avail
)
28.
if
| VM
sublist
|
=
1
29.
VM
potential
[1]
=
VM
sublist
[1]
30.
el
se
31.
VM
potential
=
ti
eB
reak
er
3(
P
c
,
P
i
,
VM
sublist
)
32.
end if
33.
el
se if
[
(P
c
=
P
a
&&
P
c
> P
i
)
||
(
P
c
=
P
i
&&
P
c
> P
a
)
||
(P
c
=
P
i
=
P
a
)]
34. c
reate
V
M
sublist
| VM
sub
list
VM
m
ain
and
VM in
V
M
subli
st
, V
M
cost
M
I
N(VM
cost
)
cos
t var
ia
ti
on
35.
if
| VM
sublist
|
=
1
36.
VM
potential
[1]
=
VM
sublist
[1]
37.
el
se
38.
VM
potential
=
tieBreake
r1(P
c
, P
a
, P
i
,
VM
sublist
)
39.
endif
40.
en
dif
41. IaaS
providers
[ ]
=
getIaa
SPro
vid
e
rD
et
ai
lsFr
om
DB(V
M
potential)
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
20
88
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
2
01
8
:
4398
-
4411
4404
The
pur
pose
of
the
m
ai
n
al
go
rithm
is
to
prov
i
de
the
best
IaaS
pro
vid
e
r/
s
based
on
the
us
er
in
pu
ts.
First,
it
co
m
pute
s
the
total
cost
us
ing
the
a
na
ly
ti
cal
m
od
el
(step
8).
T
hen,
based
on
the
pr
i
or
it
ie
s
set
by
the
us
er
f
or
c
os
t,
avail
abili
ty
and
VM
i
niti
at
ion
ti
m
e
it
decides
the
be
st
Iaa
S
pro
vid
e
rs.
T
he
m
ai
n
al
go
ri
thm
is
assist
ed
by
s
ub
-
al
gorithm
s
t
o
decide
th
e
be
st
IaaS
pro
vid
er/s.
Wo
r
king
pr
inci
ple
of
the
sub
-
al
gorit
hm
s
is
sam
e.
Su
b
-
al
gorithm
1
(tie
Br
eaker
1)
is
cal
le
d
from
the
m
a
i
n
al
go
rithm
wh
en
the
c
os
t
ha
s
the
highest
pri
or
it
y
and if
we ha
ve m
or
e than
one
VM
with
the
s
a
m
e cost.
Ther
e
is
a
pos
sibil
it
y
that
us
er
m
a
y
be
no
t
su
re
a
bout
th
e
pr
io
riti
es.
H
e
m
ay
set
al
l
t
he
pri
ori
ti
es
sam
e,
or
the
pr
iority
of
the
c
os
t
is
equ
al
to
the
pr
io
rity
of
any
one
of
the
oth
e
r
pa
ra
m
et
er
(av
ai
la
bi
li
t
y
or
init
ia
ti
on
tim
e)
in
these
cases
we
hav
e
gi
ven
an
opti
on
of
s
et
ti
ng
the
C
os
t
Var
ia
ti
on.
I
n
s
uc
h
cases,
the
VMs
with the
m
ini
m
um
cost
c
os
t
var
ia
ti
on, a
re c
on
si
der
e
d
t
o de
ci
de
the
pote
nt
ia
l VMs (
Steps
33 a
nd 34
).
Sub alg
orit
hm
1
Functi
on
V
M
potenti
al
=
tieB
reaker1(Pa
, Pi,
V
M
sublist
)
1.
I
npu
ts
: P
a
, P
i
, VM
sublist
2.
i
f
[
(P
a
> P
i
)
||
(
P
a
=
P
i
)
]
3.
VM
sublist
=
readFr
om
DB(V
M
sublist
[i]
, 1
)
4.
c
reate V
M
secondlist
| VM
s
econdlist
VM
sublist
an
d
VM i
n V
M
secondlist
, V
M
avail
=M
AX
(
VM
a
vail
)
5.
if
| VM
sec
ondlist
|
=
1
6.
VM
potential
[1]
=
VM
secondlist
[1]
7.
el
se
8.
VM
sublist
=
readFro
m
DB(V
M
su
blist
[i]
, 2
)
9.
create
V
M
thirdlist
| VMt
hir
dlist
VM
secondlist
and
VM i
n VM
thirdlist
, V
M
init
=M
IN
(
VM
init
)
10.
VM
potential
=
getPo
te
ntial
VMs
(
VM
thirdlist
)
11.
en
d if
12.
el
se if
(P
a
<
P
i
)
13. VM
sublist
=
read
F
r
om
DB(V
M
sublist
[i]
, 2
)
14. creat
e
V
M
secondlist
| VM
secondlist
VM
sublist
an
d
VM i
n V
M
secondlist
, V
M
in
it
=M
IN
(
VM
init
)
15.
i
f
| VM
se
condlist
|
=
1
16.
VM
potential
[1]
=
VM
secondlist
[1]
17.
el
se
18.
VM
sublist
=
readFro
m
DB(V
Ms
ub
li
st[i]
, 1)
19.
create
V
M
thirdlist
| VM
thirdlist
VM
secondlist
and
VM
in
VM
thirdlist
, V
M
avail
=M
A
X(VM
avail
)
20.
VM
potential
=
getPo
te
ntial
VMs
(
VM
thirdlist
)
21.
en
d if
22.
en
d if
Sub
-
al
gorithm
2
(tie
Breake
r2)
is
cal
le
d
fr
om
the
m
ai
n
alg
ori
thm
wh
en
the
VM
init
ia
t
ion
tim
e
has
the
highest
pr
i
or
it
y
and
if
w
e
hav
e
m
or
e
t
han
on
e
VM
with
the
sam
e
VM
init
ia
ti
on
tim
e.
Su
b
-
al
go
rithm
3
(tie
Breake
r3)
i
s
cal
le
d
f
r
om
t
he
m
ai
n
al
gorithm
wh
en
t
he
avail
abili
ty
has
the
highest
pr
i
or
it
y
an
d
if
w
e
ha
ve
m
or
e
than
one
VM
with
the
s
a
m
e
avail
abili
t
y.
The
m
ai
n
alg
ori
thm
al
on
g
with
the
s
ub
-
al
gorithm
s
1,
2,
and
3
are
the
co
re
pa
rt
of
the
deci
sion
strat
e
gies.
Ap
a
rt
from
t
hese
al
gorithm
s
we
ha
ve
w
ritt
en
util
it
y
fu
nctions
wh
ic
h
se
rv
e
as
helpe
r
f
unct
io
ns t
o finali
ze the
b
est
IaaS
pro
vi
der
s.
Sub alg
orit
hm
2
Functi
on
V
M
potenti
al
=
tieB
reaker2(P
c
, P
a
, V
M
sublist
)
1.
I
npu
ts
: P
c
, P
a
, P
i
,
V
M
sublist
2.
if
[
(P
c
> P
a
)
||
(
P
c
=
P
a
)
]
3.
c
reate V
M
secondlist
| VM
s
econdlist
VM
sublist
an
d
VM i
n V
M
secondlist
, V
M
c
ost
=M
IN
(
VM
cost
)
4.
if
| VM
sec
ondlist
|
=
1
5.
VM
potential
[1]
=
VM
secondlist
[1]
6.
el
se
7.
VM
sublist
=
read
F
ro
m
DB(V
M
sublist
[i]
, 1
)
8.
create
V
M
thirdlist
| VM
thirdlist
VM
secondlist
and
V
M i
n
VM
thirdlist
, V
M
avail
=M
A
X(V
M
avail
)
9.
VM
potential
=
getPo
te
ntial
VMs(
V
M
thirdlist
)
10.
en
dif
11.
el
sei
f
(P
c
<
P
a
)
12. VM
sublist
=
readFr
om
DB(V
M
sublist
[i]
, 1)
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
& C
om
p
Eng
IS
S
N: 20
88
-
8708
Profit
D
riv
e
n Deci
sio
n
Assist
S
yst
e
m
to
S
el
e
ct
…
(
Moha
n Murthy
MK
)
4405
13. creat
e
V
M
secondlist
| VM
secondlist
VM
sublist
and
VM in V
M
secondlist
, V
M
avail
=M
AX
(
VM
a
vail
)
14.
if
| VM
se
condlist
|
=
1
15.
VM
potential
[1]
=
VM
secondlist
[1]
16.
el
se
17.
create
V
M
thirdlist
| VM
thirdlist
VM
secondlist
and
VM in VM
thirdlist
, V
M
cost
=M
I
N(VM
cost
)
18.
VM
potential
=
getPo
te
ntial
VMs
(
VM
thirdlist
)
19.
en
d if
20.
en
d if
Sub alg
orit
hm
3
F
unct
i
on
V
M
potenti
al
=
tieB
reaker3(P
c
, P
i
,VM
sublist
)
1.
I
npu
ts
: P
c
, P
a
, P
i
,
V
M
sublist
2.
if
[
(P
c
> P
i
)
||
(
P
c
=
P
i
)
]
3.
c
reate V
M
secondlist
| VM
s
econdlist
VM
sublist
an
d
VM i
n V
M
secondlist
, V
M
c
ost
=M
IN
(
VM
cost
)
4.
if
| VM
sec
ondlist
|
=
1
5.
VM
potential
[1]
=
VM
secondlist
[1]
6.
el
se
7.
VM
sublist
=
readFro
m
DB(V
M
su
blist
[i]
, 2
)
8.
create
V
M
thirdlist
| VM
thirdlist
VM
secondlist
and
VM in VM
thirdlist
, V
M
init
=
MIN
(V
M
init
)
9.
VM
potential
=
getPo
te
ntial
VMs
(
VM
thirdlist
)
10.
en
d if
11.
el
se if
(P
c
<
P
i
)
12. VM
sublist
=
readFr
om
DB(V
M
sublist
[i]
, 2)
13. creat
e
V
M
secondlist
| VM
secondlist
VM
sublist
an
d
VM i
n V
M
secondlist
, V
M
in
it
=
MIN(VM
init
)
14.
if
| VM
se
condlist
|
=
1
15.
VM
potential
[1]
=
VM
secondlist
[1]
16.
el
se
17.
create
V
M
thirdlist
| VM
thirdlist
VM
secondlist
and
VM in VM
thirdlist
, V
M
cost
=M
IN(V
M
cost
)
18.
VM
potential
=
getPo
te
ntial
VMs(
V
M
thirdlist
)
19.
en
d if
20.
en
d if
The
util
it
y
fu
nc
ti
on
1
(
getP
ot
entia
lVMs)
stores
VM
detai
ls
fr
om
on
e
li
st
to
ano
the
r.
The
util
it
y
functi
on
2
(
rea
dF
r
om
DB)
rea
ds
the
avail
a
bili
ty
or
init
ia
ti
on
ti
m
e
of
a
V
M
base
d
on
th
e
value
of
th
e
input
‘k
ey
’
.
T
he
util
it
y
fu
nctio
n
3
(
getIaa
SP
rovide
rD
et
ai
lsFr
om
DB)
gets
t
he
Ia
aS
pro
vid
er
detai
ls
f
r
om
the
database
b
ase
d o
n
the
V
M
k
e
y.
Ut
il
ity
functi
on 1
Functi
on
V
M
pot
ential
=
get
Po
te
nt
ial
VM
s
(
VM
thirdli
st
)
1.
Set VM
potential
=
[ ]
2.
f
or
i=
1 t
o n
um
ber
of
VMs
pr
ese
nt in
V
M
t
hirdlist
3.
VM
potential
[i]
=
VM
thirdlist
[i]
4.
e
nd fo
r
Ut
il
ity
functi
on 2
Functi
on
V
M
li
st
=
readFr
om
D
B(V
M
list
, k
e
y)
1.
if
key=1
2.
f
or
i=
1 t
o nu
m
ber
of V
M
s prese
nt in V
M
list
3.
VM
list
[i]
.av
ai
la
bili
ty
=
read
A
va
il
abili
ty
Fr
o
m
DB(V
M
sublist
[i]
.id)
4.
e
nd f
or
5
. else
if
key
=
2
6.
f
or
i=
1 t
o nu
m
ber
of V
M
s prese
nt in V
M
list
7.
VM
list
[i]
.init
tim
e
=
read
I
nitTi
m
eFro
m
DB(VM
sublist
[i]
.id)
8.
en
d for
9.
e
nd
i
f
Ut
il
ity
functi
on 3
Functi
on
IaaS
provider
s
[ ]
=
get
I
aaSPr
ovid
erDet
ailsFro
mD (V
M
potenti
al
)
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
20
88
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
2
01
8
:
4398
-
4411
4406
1.
Set Iaa
Sprov
iders [ ]
=
[
]
2.
for
i=
1
t
o num
ber
o
f VMs
present i
n VMp
otentia
l
3.
IaaS
prov
i
der
s
[
i]
=
read
IaaS
Det
ai
lsFro
m
DB(VM
potential
[i
]
.k
ey
)
4.
e
nd fo
r
wh
e
re;
VM
m
ain
=First
li
st
of
t
he
VMs
w
hich
m
at
ches
the
f
irst
le
vel
of
se
arch
c
rite
ria
(
RAM,
CP
U,
S
tora
ge,
OS
,
Dur
at
i
on, L
ocati
on)
VM
sublist
=Op
ti
m
al
li
st of
the
VMs afte
r
a
pp
l
yi
ng
a
no
t
her le
vel of sea
rc
h
cr
it
eria
(prio
riti
es of c
os
t,
avail
abili
ty
and
VM init
ia
ti
on
ti
m
e).
VM
potential
=Pos
sible l
ist
o
f
VM
s which m
atch
us
er
crite
ria
.
VM
secondlist,
VM
thirdlist
=
List
o
f
the
VMs a
fter
app
ly
in
g
th
e
ti
ebr
ea
kers
base
d on the
pri
or
it
ie
s set by the
us
ers
on c
os
t,
a
vaila
bili
ty
an
d VM init
ia
ti
on
t
i
m
e.
IaaS
providers
=Final
li
st of
VMs
wh
ic
h
is s
how
n
to
the
us
e
r.
P
c
=Prio
rity
o
f
cost set
by the
us
er
.
P
a
=
Pr
io
rity
o
f
avail
abili
ty
set
b
y t
he
us
er
.
P
i
=Prio
rity
o
f
VM init
ia
ti
on
t
i
m
e set by the
use
r
.
VM
avail
=Av
ai
la
bili
ty
o
f
the
VM
.
VM
init
=In
it
ia
ti
on tim
e o
f
t
he VM.
VM
cost
=C
os
t o
f
the
VM.
6.
RESU
LT
S
A
ND
DI
SCUS
S
ION
The
decisi
on
a
ssist
syst
e
m
is
i
m
ple
m
ented
usi
ng
Java
an
d
Java
relat
ed
te
chnolo
gies.
Th
e
syst
e
m
is
ho
ste
d
on
T
omc
at
serv
e
r.
We
hav
e
us
e
d
J
DK
1.7.0_2
5
an
d
Ap
ac
he
T
om
ca
t
7.
0.4
2.
My
S
QL
5.5
is
us
e
d
as
the
database
to
sto
re
the
IaaS
pro
vid
e
r
and
VM
inf
or
m
at
ion
.
A
pach
e
An
t
1.9
.
2
has
bee
n
us
e
d
to
de
velo
p
buil
d
scripts.
The
u
s
er
inter
face
is
desig
ne
d
usi
ng
JSP.
Fron
t
c
ontr
oller
desi
gn
patte
rn
has
be
en
use
d
betw
e
en
the
end use
rs
and t
he DA
S.
T
he
dec
isi
on
st
rategi
es are im
ple
m
e
nted usi
ng core
j
a
va.
We
ha
ve
us
e
d
publisher
,
sub
scribe
r
desig
n
patte
rn
betwee
n
the
IaaS
pr
ovide
rs
an
d
our
DA
S
.
Our
syst
e
m
ac
ts
as
publisher
by
publishi
ng
the
AP
Is
,
a
nd
th
e
IaaS
prov
i
de
rs
are
the
sub
scribe
rs.
To
de
velo
p
web
s
er
vice
API
s
,
we
ha
ve
us
ed
A
xis
2
f
ra
m
ewo
r
k.
Deskt
op
m
achines
with
intel
cor
e
i3
process
or,
4
GB
RAM
an
d
50
0
GB
ha
rd
dis
k
are
us
e
d
t
o
e
m
ula
t
e
IaaS
pr
ov
i
der
s
.
All
th
e
m
achines
ar
e
connecte
d
th
rou
gh
LAN.
W
e
have
e
m
ulate
d
IaaS
pro
vid
er
s
to
te
st
diff
ere
nt
scenari
os
.
I
niti
al
ly
,
we
hav
e
te
ste
d
our
Web
s
erv
i
c
e
AP
I
s
for
their
functi
onal
it
y.
Fo
ll
owin
g
are
so
m
e
of
the
op
erati
ons
w
hic
h
are
exec
uted
at
the
e
m
ulated
IaaS
pro
vid
er
s.
a.
In
se
rtin
g
a
ne
w VM.
b.
In
se
rtin
g
a
ne
w pr
ic
in
g
sc
he
m
e fo
r
a
VM.
c.
Updati
ng the
di
scount
detai
ls.
d.
Updati
ng the
prici
ng d
et
ai
ls
of a
VM.
e.
Updati
ng the
VM con
fig
ur
at
ion
detai
ls.
f.
Delet
ing
a
V
M
.
g.
Delet
ing
a
pric
ing
schem
e.
We
ha
ve
obser
ved
t
h
at
as
soo
n
as
the
operat
ion
s
a
re
com
plete
d
at
the
em
ulate
d
Iaa
S
pr
ov
i
der
s
,
the
changes
m
ade ar
e re
flect
ed
in
the
DASs
data
base.
The
u
ser
c
an
a
ccess
our
deci
sion
as
sist
syst
e
m
via
a
we
b
browser.
Th
e
pag
e
s
w
hich
a
re
us
e
d
t
o
coll
ect
the
us
er
i
nputs
are
desig
ne
d
us
i
ng
JSP
te
c
hnol
ogy.
Us
ers
ca
n
nav
i
gate
f
ro
m
on
e
pa
ge
to
a
nothe
r
via
hype
rlink
s
.
User can
en
te
r t
he
val
ues for t
he follo
wing
pa
ram
et
ers
a.
Locati
on
: L
oca
ti
on
of the
data
center
.
b.
Durati
on
: Sta
rt
and E
nd
date
durin
g w
hich u
ser need
s the
VM
.
c.
Op
e
rati
ng Syst
e
m
: OS
of t
he VM.
d.
Mem
or
y
(
GB):
RAM size
i
n GB.
e.
Mem
or
y
va
riat
ion
(%)
:
Som
et
i
m
es
fixed
tem
plate
V
Ms
does
not
exactl
y
m
at
c
h
with
t
he
use
r
requirem
ent. Users ca
n
s
pecif
y how m
uch
va
riat
ion
t
hey ca
n
tole
rate.
f.
CPU
: C
P
U
s
pe
ed
in
GHz.
g.
CPU
var
ia
ti
on
(%)
: T
he va
riat
ion
i
n
CP
U
ca
pa
ci
ty
u
sers
ca
n t
olerate.
h.
Stor
a
ge
(G
B
):
Stor
a
ge
s
pace
of V
M i
n GB.
i.
Stor
a
ge variat
i
on
(%
): T
he va
riat
ion
i
n
st
or
a
ge
s
pace
us
e
rs ca
n
tole
rate.
j.
Pr
io
riti
es
:
Pr
io
riti
es o
f
the cost
, av
ai
la
bili
ty
and
VM init
ia
ti
on
ti
m
e. 1
0
is
the h
i
ghest
pri
or
it
y
,
an
d
1
is t
he
lowest.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
& C
om
p
Eng
IS
S
N: 20
88
-
8708
Profit
D
riv
e
n Deci
sio
n
Assist
S
yst
e
m
to
S
el
e
ct
…
(
Moha
n Murthy
MK
)
4407
k.
Cost
Va
riat
ion
(%)
:
Use
r
s
hould
set
t
his
perce
ntage
only
wh
e
n
t
he
use
r
c
us
tom
iz
es
thre
e
pr
i
or
it
ie
s
e
qual
or
w
he
n
t
he
use
r
giv
es
pri
ori
ty
of
cost
an
d
one
SLA
para
m
et
er
eq
ual
,
and
pr
io
rity
of
a
no
t
her
S
L
A
par
am
e
te
r
is l
ess t
ha
n
the
o
t
he
r
tw
o.
User
s
will
sub
m
it
their
re
qu
i
rem
ents
by
cl
i
ckin
g
on
the
s
ub
m
it
bu
tt
on.
Pr
essin
g
t
he
s
ub
m
it
bu
tt
on
trigg
e
rs
our
de
ci
sion
strat
egie
s.
Ba
sed
on
the
inp
ut
giv
e
n
by
the
us
er
the
corres
pondin
g
de
ci
sion
strat
eg
y
will
be
exec
uted
,
a
nd
t
he
res
ult
s
are
publis
hed
t
o
the
us
er
.
I
n
the
res
ult
pa
ge
,
the
pro
vid
e
r
nam
e
is
hyper
li
nk
e
d
t
o
the
act
ual
pro
vid
e
r’
s
VM
s
el
ect
ion
an
d
pa
ym
ent
pag
e,
wh
e
re
the
us
e
r
can
sel
ect
th
e
VM
a
nd
m
ake
the
paym
ent.
Ba
sed
on
our
s
urvey
of
dif
fe
r
ent I
aaS
pro
vide
rs
[
4
]
w
e
had
m
ock
ed
the d
at
a to test
o
ur DAS.
Part of
the
data
wh
ic
h
is
rele
van
t
to
the
do
c
ume
nted
te
st
sce
nar
i
os
in
t
he
nex
t
s
ubsect
io
n
is
ta
bula
te
d
in
th
e
T
ables
4,
5,
6,
7.
Eac
h
VM
instance
ty
pe
is
assigne
d
a
un
iqu
e
ide
ntifie
r
(which
is
not
sh
ow
n
he
re)
i
n
the
database
wh
ic
h
help
s
in
ide
ntifyi
ng
the
c
orres
pondin
g
pr
ov
i
der,
pr
ic
i
ng
schem
es
/
m
od
el
s
,
disco
un
t
de
ta
il
s
,
et
c.
Table
4
shows
the
data
f
or
the
fi
xed
V
M
tem
plate
.
We
can
obse
rve
diff
e
ren
t
cat
egories
of
VM
(S
m
all,
Me
diu
m
,
et
c.)
with
fi
xed
c
on
fig
ur
at
io
ns
.
Al
l
the
pr
ic
es
a
re
in
U
SD.Table
5
s
hows
t
he
da
ta
for
co
nfi
gu
rab
le
VMs
.
Ta
ble
6
sh
ows
the
disc
ount
detai
ls
offer
e
d
by
the
pr
ov
i
der
s
on
t
he
total
bill
ed
a
m
ount.
Iaa
S
pro
vid
e
r
s
pro
vid
e
a
high
er
disc
ount
w
hen
t
he
dur
at
ion
of
te
nure
i
s
long.S
om
e
IaaS
pro
vid
e
rs
offer
the
V
Ms
at
a
disco
unte
d
rat
e
for
the
use
rs
who
are
inte
r
est
ed
in
lo
ng
t
erm
bu
siness
if
us
e
rs
are
rea
dy
to
pay
so
m
e
init
ia
l
fee.
Table
7
s
hows
the
init
ia
l
fee
detai
ls.
Rig
or
ou
s
te
sti
ng
has
bee
n
co
nducted
by
r
unni
ng
se
ver
al
sce
nar
i
os
with
dif
fer
e
nt
req
uirem
ents
and
f
ound
that
in
each
scena
ri
o
our
D
AS
s
uggests
best
IaaS
pr
ovide
r/s
bas
ed
on
the user i
nput. Few test
e
d
sce
nar
i
os
a
nd
thei
r
re
su
lt
s a
re tab
ulate
d
in
Ta
ble 8
a
nd T
a
ble
9 resp
ect
ively
.
Table
4
.
Fixe
d VM
Tem
plate
D
et
ai
ls
VM
Ty
p
e
CPU
(GHz
)
RAM
St
o
rage
VM
in
itiatio
n
ti
m
e
(
m
in
u
tes)
Av
ailab
ility
Locatio
n
OS
Price
p
e
r
m
o
n
th
Prov
id
er
S
m
all
1
.2
2
25
15
99
Ch
icag
o
Linu
x
500
Prov
id
er
1
S
m
all
1
.5
2
20
30
9
9
.5
Ban
g
alo
re
W
in
d
o
ws
550
Prov
id
er
3
Mediu
m
2
.4
4
50
30
99
Ch
icag
o
Linu
x
750
Prov
id
er
2
Lar
g
e
3
.6
6
75
30
9
9
.5
New York
win
d
o
ws
1200
Prov
id
er
4
Extra
Lar
g
e
4
.8
8
100
60
99
New York
win
d
o
ws
1400
Prov
id
er
1
0
T
a
bl
e
5
.
C
on
f
i
gu
r
a
bl
e
V
M
D
e
t
a
il
s
IaaS
p
rov
id
er
RAM pric
e
/GB/
m
o
n
th
CPU p
rice
/GHz/
m
o
n
th
Sto
rage
p
rice/GB/
Mon
th
VM
Initiatio
n
Ti
m
e
(
m
in
u
tes)
Av
ailab
ility
in
%
Locatio
n
OS
Prov
id
er
1
50
75
10
30
99
Sy
d
n
ey
W
in
d
o
ws
Prov
id
er
2
40
70
12
45
9
9
.5
Lon
d
o
n
Linu
x
Prov
id
er
3
40
85
15
45
9
9
.9
Ban
g
alo
re
W
in
d
o
ws
Prov
id
er
4
75
90
10
15
9
9
.9
Ban
g
alo
re
Linu
x
Prov
id
er
5
50
65
20
15
99
Ban
g
alo
re
W
in
d
o
ws
Table
6
.
Disco
un
t
Detai
ls
on t
he
T
otal B
il
le
d
Am
ou
nt
IaaS
p
rov
id
er
VM
#
VM
Ty
p
e
Disco
u
n
t on
the to
t
al bill in
%
Fo
r
3
-
6
m
o
n
th
s
Fo
r
7
-
1
2
m
o
n
th
s
Fo
r
1
y
e
ar
+
Prov
id
er
1
1
S
m
all
10
20
30
Prov
id
er
1
Co
n
f
i
g
u
rable
10
15
20
Prov
id
er
2
1
Mediu
m
15
20
25
Prov
id
er
3
2
S
m
all
5
10
15
Prov
id
er
4
1
Lar
g
e
0
20
30
Prov
id
er
5
3
S
m
all
0
10
20
T
a
bl
e
7
.
D
e
t
a
i
l
s
of
t
he
R
e
d
uc
e
d
P
r
i
c
e
a
f
t
e
r
I
n
i
t
ia
l
F
e
e
IaaS
Prov
id
er
VM
#
VM
Ty
p
e
Initial f
ee
Price
af
ter
initial f
ee
Reg
u
lar
p
rice/
m
o
n
th
Fo
r
3
-
6
m
o
n
th
s
Fo
r
7
-
12
m
o
n
th
s
Fo
r
1
y
e
ar
an
d
m
o
re
Fo
r
3
-
6
m
o
n
th
s
Fo
r
7
-
12
m
o
n
th
s
Fo
r
1
y
e
ar
+
Prov
id
er
1
2
S
m
all
100
150
250
425
400
375
450
Prov
id
er
6
1
Mediu
m
200
300
400
825
815
800
850
Prov
id
er
7
1
S
m
all
0
100
200
475
450
400
475
Prov
id
er
8
2
Mediu
m
0
0
250
850
850
750
850
Evaluation Warning : The document was created with Spire.PDF for Python.