TELKOM
NIKA
, Vol.14, No
.1, March 2
0
1
6
, pp. 280~2
8
5
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v14i1.2321
280
Re
cei
v
ed
Jul
y
13, 201
4; Revi
sed
De
ce
m
ber
20, 201
5; Acce
pted Janua
ry 6, 20
1
6
The Analyses on Dynamic an
d Dedicated Resource
Allocation on Xen Server
Mardha
ni Riasetia
w
a
n
1
, Ahm
a
d
Asha
ri
1
, Ir
w
a
n Endra
y
anto
2
1
Departme
n
t of Computer Sci
ences a
nd El
e
c
tronics,
F
a
cult
y of Mathem
ati
cs and Natur
a
l
Scienc
es,
Univers
i
tas Ga
dja
h
Mada, Bu
l
a
ksumur, Yog
y
akarta
2
Departme
n
t of Mathematics, F
a
cult
y
of
Mathematics a
nd
Natura
l Scienc
es,
Univers
i
tas Ga
dja
h
Mada, Bu
l
a
ksumur, Yog
y
akarta
e-mail: mardha
ni@u
gm.ac.id
A
b
st
r
a
ct
Data c
enter to
day c
hal
len
ges
is n
o
t o
n
ly s
e
r
v
e the
users,
i
n
sa
me
ti
me
n
eed
to esta
bl
is
h scal
a
b
l
e
resourc
e
s. Dat
a
C
enter
man
age
the
reso
u
r
ces suc
h
as process
o
r,
sto
r
age, netw
o
rk,
an
d memory
in
appr
opri
a
te w
a
y to handl
e to l
oad. In the bi
g data era,
lo
ad
w
ill increas
e a
nd co
me i
n
rap
i
d w
a
y w
i
th large
volu
me data,
ma
ny type
of d
a
ta, can b
e
strea
m
a
nd
b
a
tch
data, an
d u
n
k
now
n sourc
e
s. Reso
urces n
e
e
d
to
ma
na
ge w
i
th
compre
hens
ive
strategies to
face the ch
ara
c
teristic of big
data
lo
ad. D
a
ta Ce
nter ha
ve
capa
bilti
e
s on
alloc
a
te the re
osourc
e
in dy
na
mic a
nd d
e
d
icate
d
w
a
ys. T
he researc
h
i
n
vestig
ate in t
he
perfor
m
a
n
ce o
f
dedyc
a
ted a
nd dyna
mic
re
source
al
lo
cati
on to
defi
n
e
the r
e
li
abl
e str
a
tegi
es o
n
D
a
ta
Center. T
h
e r
e
searc
h
w
o
rk
on
Xe
nServe
r pl
atform as
Data
Cent
er.
T
he res
earch
defin
e
18 V
i
rtual
Machi
ens
both
on
d
edic
a
ted
and
dy
na
mic
s
t
rategies,
use
the s
hare
d
stor
age
mech
an
is
m, a
n
d
res
our
ce
poo
ls. T
he r
e
s
earch
an
aly
z
e
on
CPU
perfor
m
a
n
ces
on
Xe
nServer
1
a
n
d
XenServ
e
r2 t
h
at des
ign
as
cl
uster
Data Ce
nter.T
he test has ru
n on Xe
nServ
e
r and r
e
su
lti
n
g the 2 p
has
e
of process w
hen D
a
ta C
e
n
t
er
alloc
a
te th
e r
e
sources, th
ere
are
inti
ation
phas
e a
n
d
pr
ocess phas
e. T
he
res
ear
ch
show
n
that in
the
intiati
on
ph
ase
both
dy
na
mic
an
d
ded
icate
d
strateg
i
es
sti
ll
not ru
nn
ing,
and
us
e th
e i
n
itial
reso
urces
to
establ
ish D
a
ta
Center. T
h
e
process
ph
a
s
e show
n
th
a
t
dyna
mic
an
d de
dicat
ed s
t
rategies r
un
and
gen
eratin
g th
e
lo
ad
proc
ess. In th
e pr
oces
s ph
ase
it s
h
o
w
n the
use
of
me
mo
ry and
CPU
P
e
rfor
ma
nce
stream li
ne i
n
t
o
the
bal
anc
e
positi
ons
. T
h
e
researc
h
res
u
lt can
use for
alloc
a
tin
g
res
o
urces is
ne
ed
to
defin
e differe
nt strategies in
i
n
itition a
nd pr
oc
ess phas
e.
Ke
y
w
ords
: Da
ta Center, Virtu
a
li
z
a
ti
on, Virtu
a
l Mach
ines, C
P
U Index, Me
mory Us
ag
e
Copy
right
©
2016 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
Clou
d Data
Cente
r
i
s
ne
eded fo
r
co
mpany that
has ch
allen
g
e
in the
re
so
urces an
d
servi
c
e
s
[1]. Data Cente
r
has
ca
pabiliti
e
s
to han
dl
e
huge t
r
affic
of se
rvice
s
, la
rge volum
e
of
data
transfe
r and
scalabl
e
reso
urce
m
ana
ge
ment
[16
]. Data
Center
combines
the in
fr
as
tu
c
t
ur
es
,
platform
and
appli
c
atio
n
into services that
can
b
e
a
c
cess th
rough
the
ne
tworks.
Seve
ral
approa
che
s
,
Data
Ce
nter d
e
velop
with sta
nda
rd and
security paramete
r
s.
Data
Ce
nter
infrast
r
u
c
ture
s also need t
o
meet with the hardw
are
stand
ard
s
an
d also the scalable ha
rd
ware
manag
eme
n
t. In the hard
w
are
confu
guration, its
also
need to e
a
ssy to mana
g
e
the re
so
urces.
The te
chn
o
lo
gy to man
a
g
e
the h
a
rdwa
re from
clu
s
ter, gi
rd, hig
h
perfo
rma
n
ce
com
puting
a
n
d
clou
d esta
blish for add
re
ss
the hard
w
a
r
e
probl
em
s [13].
The technol
o
g
y to easy m
anag
e hardware in
Data
Center i
s
Virtu
a
lizatio
n. Virtualization
provide
s
t
he t
e
ch
nolo
g
y for clo
ud
co
mpu
t
ing an
d h
o
st
ed
se
rvice
s
cl
oud
co
mputin
g [2, 1
4
]. Wh
en
applie
d to th
e
data
ce
nter,
then virtuali
z
ation
can
im
p
r
ove IT
efficie
n
cy of
an
ent
erp
r
ise [17]. I
n
this
ca
se, virt
ualization
ca
n re
du
ce th
e
need
for
phy
sical serve
r
so
that
it can
save
th
e co
st
of
pro
c
u
r
eme
n
t and mai
n
ten
ance of co
m
puter inf
r
a
s
tructure. In ad
dition to these advantag
es, the
virtual data center h
a
s a p
r
oble
m
that is not necessa
rily
the capa
ci
ty of t
he available resource
s
can
serve
th
e ne
ed
s
of the d
a
ta
ce
nter. If a
dat
a
ce
nter reso
urce
sh
ortag
e
s, the
exi
s
ting
solutio
n
s
are likely
to add
hard
w
a
r
e. Where
a
s
the
co
nce
p
t of virtu
a
lizatio
n i
s
try
i
ng to
streaml
i
ne
existing re
so
urces
with growin
g inform
ation nee
ds [
3
, 4].
Data
ce
nter
physi
cal a
b
ili
ty to store
in
co
min
g
d
a
ta
co
ntinuou
sly
ha
s the limi
t
ations.
Hardware is
own
ed by th
e data
ce
nter has a lif
e
span of
about
3-5 ye
ars. Data ce
nters a
l
so
requi
re
a
st
able
so
urce
of ele
c
tri
c
p
o
we
r
and
a
n
en
orm
o
u
s
investm
ent
co
sts to b
u
il
d it.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
The Anal
yses on Dynam
ic and Dedi
cate
d Re
sou
r
ce
Allocatio
n
on…
(Mard
hani
Ri
asetia
wa
n)
281
Substantio
nal
co
st red
u
cti
on mu
st be
made
wit
hou
t redu
cing th
e efficien
cy of data ce
nter
perfo
rman
ce.
One
meth
od
to improve th
e effectiven
e
s
s an
d effi
cie
n
cy of
the
dat
a cente
r
i
s
u
s
ing
the Virtual
Data Ce
nter [
5
]. Each server ha
s ma
n
y
virtual machine
s
ru
nnin
g
on laye
r VMM
(Virtual
Ma
chine M
onitor). With the
VMM, multiple
virtual m
a
chine
s
can
ru
n on
a p
h
ysical
serve
r
. Ea
ch
virtual ma
chi
ne
contai
ns the o
perat
ing
system
and
a
pplication
s
th
at are gove
r
n
e
d
by Virtual
Center M
ana
g
e
ment Se
rver. Wi
th
the
virtual data
cente
r
m
a
nagem
ent a
nd
maintena
nce can ma
ke the
data cente
r
more effici
ent
[13].
Re
sou
r
ce all
o
catio
n
in th
e data center can
b
e
int
e
r
p
ret
ed i
n
sev
e
ral t
e
r
m
s.
R
e
so
ur
ce
manag
eme
n
t is relate
s to manage resource
s well to cre
a
te virtual machi
n
e
s
(VM) to multipl
e
resou
r
ce pool
s and
combi
n
e them in a VM grou
p mad
e
[6, 7]. This will re
sult in a more o
r
de
rl
y
VM and f
a
cilitate the
distrib
u
tion
of acce
ss rig
h
ts to
a parti
cul
a
r VM g
r
o
up.
Re
sou
r
ce Po
ols u
s
e
d
to
set up the
ha
rdwa
re
whi
c
h
is the
n
u
s
ed
by the virtual
machine
in t
he
data
cente
r
. In the
re
sou
r
ce pool
s
allo
ws allo
catin
g
u
nused
re
sou
r
ce
s to the
V
M
that re
qui
res
the su
ppo
rt o
f
resource
s
much
large
r
. VM prio
ri
tize
certai
n p
r
o
c
e
s
ses
ca
n al
so be
done
ea
sily.
Built-in SSO i
s
p
r
ovide
d
to
manag
e u
s
e
r
s
who
ca
n a
c
ce
ss the d
a
ta
ce
nter fa
cilit
y to set
spe
c
i
f
ic
permi
ssion
s
.
Users Poli
cy/Rule
s
, case manag
em
e
n
t acce
ss
ba
sed on g
r
o
u
p
s
of op
eratio
nal
function
s [12]
.
The re
se
arch
works on g
e
nerate
a virtu
a
l dat
a cente
r
archite
c
ture
implementati
on and
Comp
ari
ng n
eed
s CP
U a
nd mem
o
ry reso
urce
s on
the virtual da
ta cente
r
(X
e
n
Server) u
s
in
g
dedi
cated
an
d dyna
mic all
o
catio
n
of
re
sou
r
ces to
organi
ze. T
he
pape
r
will
explain i
n
seve
ra
l
se
ction
s
. Re
search m
e
tho
d
will explai
n
about d
e
sig
n
and im
ple
m
entation. Result
s sectio
n will
explain the result of rese
arch. Concl
u
si
on will sum
m
ar
ize the resear
ch works.
2. Rese
arch
Metho
d
In this se
ction
,
it is explained the re
sea
r
ch
method con
s
ist of Data
Center Proj
ect
s
and
works, sched
ul hira
rchy on
data
cente
r
, and the re
se
a
c
h process.
2.1. Data
Ce
nter Projec
ts
Some data
center d
e
velop
m
ent proje
c
ts resu
lted
in some
p
r
od
uct appli
c
ation
s
are
q
u
ite
diverse seep
rti Open
Ne
b
u
la, Xtreem
OS, Future
G
r
id, DIET, BEinGRID, ScienceFo
r
ge,
Dala
Proje
c
t, and
Gama
Clou
d
.
Open
Neb
u
l
a
and Extre
e
mOS al
rea
d
y discu
s
sed
in the secti
on
virtualiasi.
Future
Grid i
s
a test-bed
for grid a
n
d
cloud
comp
uting whi
c
h i
s
ke
rjam
a b
e
twee
n
Grid'5000 and TeraGrid.
FutureGrid
[8] built with a lot of
c
l
o
ud infras
truc
ture with a
broad
demog
ra
phic.
At this infra
s
tru
c
ture is
d
one rese
a
r
ch
on auth
entication,
autori
a
sai, sch
eduli
ng,
virtualizatio
n, and clo
ud-based comp
uting. Future
Grid integ
r
at
e with so
me
tools like X
en,
Nimbu
d
, Vine
and Ha
doo
p [1].
INRIA in 200
0 introd
uced
DIET that implem
ent
s a d
i
stribute
d
sch
edulin
g in gri
d
an
d
clou
d infra
s
tructures.
DIET is a muli
-agent sy
ste
m
to the middle
w
are. Th
is infra
s
tru
c
t
u
re
provide
s
sp
e
c
ific
eleme
n
ts of th
e
clou
d an
d the
all
o
catio
n
of
re
sou
r
ces ba
sed o
n
d
e
ma
nd,
combi
ned
wi
th economi
c
facto
r
s cl
ou
d computin
g
[8]. BeinG
r
id is a
re
se
a
r
ch
p
r
oje
c
t t
hat
provide
s
a
grid infrast
r
u
c
ture fo
r th
e purpo
se
s of the real
busi
n
e
ss
scenari
o
s. At this
infrast
r
u
c
ture
done some
resea
r
ch on
co
st red
u
ctio
n, improved
perfo
rman
ce
on the stren
g
th
pemo
r
sesan,
bu
sine
ss
model
devel
opment, a
n
d
run
n
ing
Sa
aS. Scien
c
e
F
org
e
[9] is an
infrast
r
u
c
ture
developm
en
t cluste
r aim
ed at collab
o
rative resea
r
ch
data. Bu
ilding research-
based
storag
e cl
uste
rs th
at ru
n o
n
PC-ba
s
ed
cl
uste
r. Research
use
s
Ap
plication F
r
ame
w
o
r
k
that provide
s
SaaS se
rvices to
the a
p
p
licatio
n laye
r. Applicatio
n
Layer i
s
buil
t
applicatio
ns to
perfo
rm data
colle
ction, the
data pro
c
e
s
sing and d
a
ta archives.
DALA Project [10] is a
co
ntinuation
of
Scie
n
c
eF
org
e
infra
s
tructu
re d
e
velop
e
d
for the
purp
o
se of prese
r
vation of
data. The re
sea
r
ch focu
ses on the p
r
e
s
ervatio
n
pro
c
e
ss d
a
ta usi
ng
clu
s
ter infrast
r
uctu
re. Th
e resea
r
ch re
sulted in
the pro
c
e
ss of d
a
ta manag
e
m
ent model
that
con
s
i
s
ts
of an inp
u
t lay
e
r, laye
r re
membe
r
, p
r
e
s
ervatio
n
lay
e
r a
nd
outp
u
t layer, the
data
manag
eme
n
t layer, and a
r
chival sto
r
a
g
e
layer. Ga
m
a
Clo
ud [11] impleme
n
ting
the infra
s
tru
c
ture
of scie
n
tific resea
r
ch wit
h
the model
grids.
Ga
m
a
Clo
ud co
nstructed by a
rra
ngin
g
fabric
comp
one
nts
of the data center, net
work ma
n
agem
e
n
t and sto
r
a
ge clu
s
te
r. G
a
maCl
oud
use
menjem
bani
Globu
s mi
dd
lewa
re
se
rvice
s
for t
he
physi
cal laye
r with
the l
a
yer a
bove
it.
Gama
Clou
d run the d
a
ta
manag
eme
n
t se
rvice
s
th
a
t
previou
s
ly
built by DAL
A
proje
c
t a
s
an
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 1, March 2
016 : 280 – 2
8
5
282
engin
e
. Setting resource
s is th
rou
gh
worklo
ad m
a
nagem
ent
wi
th acce
ssible
throu
gh
ce
rtain
sci
entific ap
pl
ication
s
.
2.2. Scheduller on Da
ta Center
As the
key chara
c
te
risti
c
s of re
sou
r
ce
manag
eme
n
t, sched
uling
servi
c
e m
a
ke
s cl
oud
comp
uting di
fferent from
other
com
put
ing pa
r
adi
gm
s. Ce
ntrali
ze
d sche
dule
r
on the
clu
s
ter
system aims to
improve overall
system
perfo
rman
ce;
while th
e schedul
er i
s
di
stributed in
a
gri
d
system ai
ms
to improve th
e perfo
rma
n
ce of sp
e
c
ific
end u
s
e
r
s.
Compa
r
ed to t
he othe
rs, th
e
sched
uling in
clou
d co
mp
uting is m
u
ch more co
m
p
licate
d
. On
the one h
a
n
d
, a ce
ntrali
zed
sched
uler i
s
requi
re
d, sin
c
e each cl
oud
provide
r
,
wh
ich p
r
omi
s
e
s
to provide
se
rvice
s
to use
r
s
without
reference to the i
n
frast
r
u
c
ture
of the ho
st, h
ad an
individ
ual data
cent
er. On
the ot
her
hand, a
di
stri
buted
sched
u
l
er i
s
al
so ve
ry nece
s
sa
ry, becau
se
com
m
ercial p
r
o
p
e
r
ty determi
ne
s
that
clou
d co
mputing ha
s to
deal with
t
he
QoS
re
qu
ireme
n
ts of
custome
r
s di
stributed
a
c
ro
ss
dunia.Suatu i
m
porta
nt poi
nt of th
is di
scussion i
s
to o
u
tline t
he sch
edulin
g
probl
ems asso
ciat
ed
with
clou
d co
mputing. Be
cause
cloud
service
s
ar
e a
c
tually virtual
pro
d
u
c
t in th
e supply
chai
n,
sched
uling
se
rvice
s
ca
n be
classified int
o
two
ba
sic
categori
e
s: u
s
er-l
evel and
system-level [
8
].
User-level
scheduli
ng de
al
s with
th
e issues
rai
s
ed
by the provi
s
ion
of se
rvice
s
b
e
twee
n
the provider
and th
e cust
omer. T
h
is m
a
inly refe
rs
to e
c
on
omic i
s
sue
s
su
ch a
s
the
bala
n
ce of
sup
p
ly and
d
e
mand,
com
petition bet
ween the
con
s
umer and
th
e minimi
zatio
n
of cost to
the
con
s
um
er is
elasti
c. Sch
e
duling
of sy
stem-lev
el deal
s with
the ma
nagem
ent
of resou
r
ces
in
t
h
e
data cente
r
. From
the cu
stomer stan
dp
oint,
the
d
a
ta
ce
nt
er
is
t
h
e
sy
st
e
m
inte
g
r
ation, providi
ng
uniform
se
rvice
s
. Actually, the data ce
n
t
er co
nsi
s
ts
o
f
many physi
cal ma
chi
n
e
s
, homoge
neo
us
or h
e
tero
gen
eou
s. After receivin
g man
y
tasks of
dif
f
erent u
s
e
r
s,
the pla
c
eme
n
t of a task to
a
physi
cal ma
chine have a
significa
nt impact on the p
e
r
forma
n
ce of the data ce
nter. In addition
to
increasing the utilization
of the system,
certai
n condi
tions hav
e to be consi
dered
, such as real-
time satisfa
c
ti
on, resou
r
ce sha
r
ing, an
d other fault toleran
c
e [8].
2.3. Process
Data Ce
nter and
Vi
rtual Data
Center
is requi
re
d a
s
t
he fou
ndation
of this re
se
a
r
ch
is in
the form
of Data Ce
nter i
n
frast
r
u
c
ture
is
functi
oni
ng a
s
a virtu
a
l dat
a ce
nter. Virt
ual Data Cen
t
er
itself will
hav
e the
charact
e
risti
cs/mi
nim
u
m param
e
ters, such as
has a
phy
sical server resources
manag
eme
n
t in reso
urce
manage
men
t
based IaaS
. It is need to have a ph
ysical resou
r
ce
allocation ma
nagem
ent into Virtual Ma
chin
es. It is ave the use
r
manag
emen
t with use
r
a
n
d
grou
p p
r
ivileg
e
s
pa
sa u
s
e
r
. It is ha
s th
e ha
ndling
of
the ha
nd
-ov
e
r m
e
chani
sms fo
r mig
r
at
ion
manag
eme
n
t serve
r
overl
a
y within the Virtual Data
Ce
nter.
Stages of Virt
ual
Data
Ce
n
t
er a
nalysi
s
h
a
s
bee
n
carri
ed o
u
t p
r
elimi
nary
re
sea
r
ch
on th
e
allocation of reso
urce
s to the clu
s
ter inf
r
ast
r
u
c
ture a
nd virtual ma
chin
e enviro
n
m
ent. Re
sea
r
ch
analysi
s
on t
he perfo
rma
n
ce of the virtual se
rver
i
s
focu
se
d o
n
the isolatio
n of resources,
esp
e
ci
ally Virtual M
a
chin
e Ma
nag
em
ent [12].
T
h
is re
sea
r
ch comp
ares
th
e
data
cent
er
environ
ment
that is Pro
x
mox, Open
Stack and
Eu
c
a
lyp
t
us
.
T
h
e
r
e
s
e
ar
ch
r
e
s
u
lted
in
th
e
allocation of resou
r
ces to
the isolation
and vi
rtualization model
s. This pro
c
e
ss p
r
od
uced
a
perfo
rman
ce
dedi
cated for
every job that
is run in a
clustered envi
r
onment.
In the Data
Center
req
u
ire
d
distri
bution
patte
rn d
a
ta center that
can
handl
e the n
eed
s o
f
resource Alocation to adapt with the load. De
termination of the
burd
en
will be based on the
availability of
data
ce
nter re
so
urce
s,
para
m
eters a
nd m
e
tadata
pr
eservation
, as well
a
s
the
categ
o
ry of big data. Determinat
ion of data cente
r
re
sou
r
ces i
s
de
sign
ed to follow the pattern of
the ne
ed
s o
f
the data.
Re
sou
r
ces t
hat w
ill
ha
n
d
le
comp
utin
g ne
ed
s in
clude
hardwa
r
e
para
m
eters t
hat can
be set in IaaS data cente
r
servi
c
e
s
, namely the proc
esso
r, stora
ge me
d
i
a,
netwo
rk.
3. Results a
nd Analy
s
is
In this sectio
n, it is explai
ned the
re
sul
t
s of
re
sea
r
ch and
at the
same tim
e
is
given the
comp
re
hen
si
v
e
disc
us
sio
n
.
3.1. Implementa
tion
R
e
s
e
ae
ch
us
e
d
in
fr
as
tr
uc
tu
r
e
o
f
the
PC
-b
as
e
d
c
l
us
te
r
.
PC
c
l
u
s
ter
ea
c
h
be
operationali
z
ed op
eratin
g
system
sof
t
ware
that suppo
rts Vi
rt
u
a
l Data
Ce
nter. Research
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
The Anal
yses on Dynam
ic and Dedi
cate
d Re
sou
r
ce
Allocatio
n
on…
(Mard
hani
Ri
asetia
wa
n)
283
prep
ared
Ce
ntOS ope
rati
ng sy
stem a
nd Wi
ndo
ws
Server
201
2 R2
Data
Cente
r
. Cent
OS
operating
system ha
s b
e
e
n
known to
b
e
quite
goo
d
in compatibili
ty with multip
le Dat
a
Cent
er
softwa
r
e, like
Open
Ne
bul
a, OpenSta
ck, and Othe
r.
While Win
d
o
w
s Data Ce
nter
is a
de
dica
ted
operating
system for the d
a
ta ce
nter. T
o
su
ppo
rt
the
operation of
the infra
s
tru
c
t
u
re
will be
used
as PC
se
rve
r
hardware with the foll
owin
g sp
ecifi
c
ation
s
: Xeo
n
Processo
r,
4GB Memo
ry,
Network Spe
ed Giga
bytes, Storage 50
0GB. PC se
rvers will b
e
p
r
epa
re
d individually as
server
and cl
uste
r configuration
will melalu
ui Data C
ente
r
softwa
r
e resp
ectively. Mechani
sm goe
s by
Data
Center
will use the approach
to
handling large
data. Data Cent
er
will be
set multiple Virt
ual
Machi
n
e
s
an
d ope
rated
several o
p
e
r
ati
ng sy
stem
s therei
n. Virtua
l Machi
n
e
s
o
n
ea
ch
clu
s
te
r will
be configu
r
e
d
so
as to
al
low e
a
ch oth
e
r to
divide i
t
s re
sou
r
ce
s. Mech
ani
sm
of manufa
c
tu
re
cluster
will follow the
nature of the
soft
ware exis
ting data center. At the Data
Center, this
has
been impl
em
enting the clu
s
ter, and the
n
implemente
d
the mecha
n
i
sm of allocati
on of resources
from the
de
sign
of dyna
mic resou
r
ce
allocation.
Impleme
n
tatio
n
is pl
aced
at the level
of
appli
c
ation la
yer data ce
nter to facilitate
t
he setting of dynamic resource allo
cati
on.
3.2. Test Res
u
lts
System testin
g condu
cted
to mea
s
u
r
e t
he p
e
rf
o
r
man
c
e
of the
system, in this case
the
XenServe
r. T
h
is p
e
rfo
r
ma
nce
pa
ramet
e
r i
s
the p
e
rcenta
ge of
CPU
re
so
ur
ce
s an
d mem
o
ry
.
Testing i
s
do
ne by run
n
ing
19 VM gra
d
u
a
lly from
initiation 3 VM, runnin
g
8-d
edi
cated VM to run
8 VM-dynami
c
and CP
U u
s
ag
e cal
c
ulat
ion usin
g
dat
a averag
es from 4 core CPUs ea
ch h
o
s
t
and use me
mory whe
n
the VM is executed. In
these discu
s
sion
s, displaye
d an avera
ge u
s
ag
e
of CPU
re
sou
r
ce
s eith
er
when initiation
VM,
VM-dedi
cated run
n
ing
and
runni
ng VM-dynami
c
as
well as the u
s
e of CPU re
sou
r
ces a
s
a
whole. As fo
r the memo
ry to be display
ed VM memo
ry
use
d
in
ea
ch
test sce
n
a
r
io,
ch
ang
e the
memory
in X
en, a
s
well
as the ove
r
all
u
s
e
of mem
o
ry
for
925 minute
s
.
Table 1 . CP
U
usage
Xen
S
erver1 d
an
XenServe
r2 (%)
Data Cente
r
Pro
c
ess
CPU Usage aver
ages
CPU
usage gap
XenServer1
XenServer2
Initiation
4,38
0,85 3,53
Dedicated1
4,10
0,93 3,17
Dedicated2
3,52
1,83 1,69
D
y
namic1
3,16
1,76 1,4
D
y
namic2
3,63
2,45 1,18
Averages
3,76
1,56 2,20
Table 1
sh
o
w
s th
e avera
ge CP
U u
s
a
ge on Xe
nServer1 and X
enServe
r2 fo
r all test
scena
rio
s
.
It is sho
w
th
at CPU usage
on
XenS
er
ve
r1 al
way
s
hi
g
her t
han t
he
CPU
usage
on
XenServe
r2
well a
s
the in
itiation of the VM, r
un-d
edi
cated VM a
n
d VM-dynami
c
ru
n. CPU u
s
e
s
grap
h on Xe
nServer1 fluctuations, unli
k
e the
CP
U usa
ge on Xe
nServer2 whi
c
h tend to ri
se.
Cha
r
ts the a
v
erage diffe
rence in CPU usag
e and
XenServe
r2 XenServe
r1 are
in
crea
sin
g
ly
smaller.
Table
2
sho
w
s the
ch
an
ges in m
e
m
o
ry at
th
e X
en VM
wh
en
there i
s
a
run. Th
at
indicates
ch
a
nge
s in me
m
o
ry wh
en the
scena
ri
o initi
a
tion Xen V
M
, run-dedi
cated VM an
d
VM-
dynamic ru
n.
Whe
n
a VM
i
s
not th
ere ru
n (idl
e)
either on o
r
XenS
e
r
ver2 X
enSe
r
ver1, Xen
ha
s a
default mem
o
ry of 953 MB
on ea
ch Xe
n
S
erver1
and
XenServe
r2.
Whe
n
there i
s
a VM runni
ng
on XenSe
r
ve
r1 a
nd Xe
nServer2, Xen
requires the
a
ddition of
re
source
s resulting in i
n
crea
sed
memory.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 1, March 2
016 : 280 – 2
8
5
284
Table 2. Mem
o
ry Test
Scenario
Me
mo
ry
XenServer1
XenServer2
Idle 953
953
Initiation 1733
953
Dedicated1
2773
953
Dedicated2
2773
2000
D
y
namic1
2773
3040
D
y
namic2
3558
3308
Based
on the
test data, th
e avera
ge val
ue of CP
U u
s
age o
n
ea
ch
host XenS
erv
e
r1 a
n
d
XenServe
r2
are at 3.7
6
% and 1.56%.
The maxi
mu
m value of th
e CPU
usag
e on XenSe
r
ver1
and X
enServ
e
r2
re
sp
ectiv
e
ly 46.93%
a
nd 3
2
.71%
,
while
the
min
i
mum valu
e o
f
the
CPU
usage
on XenSe
r
ve
r1 an
d XenS
erver2 re
sp
e
c
tively
by 2.16% and 0.72
%. Based on
test data, 4055
MB of total memory u
s
ed
memory u
s
a
g
e
on ea
ch
h
o
s
t XenServe
r1 and XenS
erver2 of 355
8
MB
and 382
0 MB. While the use of unuse
d
memory o
r
memory that is available on
XenServe
r1 a
n
d
XenServe
r2 e
a
ch for
497 M
B
and 235 M
B
.
In
term
s
of CPU and me
mory
pe
rform
ance,
XenSe
r
ver1
and
Xe
nServer2 h
a
ve goo
d
balan
cing
proce
s
s, whi
c
h
is indicated
by t
he decli
ne in XenSe
r
ver1
CPU p
e
rform
a
n
c
e a
n
d
increa
sing
pe
rforma
nce of CPU in Xe
nServer2,
in oth
e
r word
s the
differen
c
e b
e
twee
n the CP
U
usa
ge an
d XenServe
r2 X
enServe
r1 ha
s narro
we
d.
The mo
re virtual machine
(VM) on a server
and the more memory re
quire
d impa
ct on CPU pe
rf
orman
c
e. It is sh
own fro
m
the use of
the
CPU
when the VM initiation sc
enario, dedica
ted VM
and VM dynamic
changes
in the average
value of both the XenServe
r1ma
upu
n XenServer2 with
incre
a
si
ng m
e
mory.
4. Conclusio
n
The re
se
arch
has sho
w
n that the re
sou
r
ce
allocation
use dynami
c
and de
ciate
d
have
several p
a
te
rn, the
r
e
are initiation
pha
se
wh
en
the resou
r
ce configu
r
e
and
set
up
for
establi
s
ihin
g the servi
c
e
s
. The next proce
s
s is
the
contin
uo
s ph
ase e
s
tabli
s
h
i
ng the se
rvice
s
bith dynamic
and de
dicate
d. Base do
n the CP
U i
nde
x perform
an
ce we can
see
n
that along
with
the pro
c
e
s
s both Xen Se
rver 1 a
nd Xen Serve
r
2
strea
m
line int
o
the balan
ce positio
n. it is
sho
w
n that th
e bload b
a
lan
c
ing throug
h the dynami
c
a
nd dedi
cate
d can
works into the se
rvers.
In the mem
o
ry utilization
can be
se
en t
hat
usability of memo
ry in
cre
a
si
ng th
ro
ugh the
pro
c
e
ss.
A
l
l
of
se
rv
er
s
ca
n allo
cat
e
t
h
e mem
o
ry
fo
r all th
e p
r
o
c
ess, it can
b
e
proven
by
the
increa
sing of
memory usi
ng both on d
y
namic an
d dedi
cated. In the technica
l view, it is very
importa
nt to allocate the re
sou
r
ce match
with the strat
egie
s
, both d
y
namic o
r
de
dicate
d.
Based
on
the
highli
g
th, we
ca
n
rea
c
h
a
n
con
c
lu
sion
that man
age
the resource
on d
a
ta
cente
r
with d
e
fine the re
source into dy
namic
and d
e
d
icate
d
both
will sig
n
ifica
n
t
have impact
on
the se
rver pe
rforma
nces.
The re
so
urce
have 2 pha
se to establi
s
hi
ng the re
sou
r
ce
s, the initiation
pha
se and th
e pro
c
e
ss ph
ase. A long with the pr
o
c
ess, both servers will st
re
amline into the
balance
resources. T
he number
of V
M
s even increase
still
can handle
with
the resources
manag
eme
n
t. The VMs itself will be pla
c
e a
s
task
/jo
b
s in the
re
source
s allo
ca
tion. Even that,
the re
sea
r
ch still need to
continue
with the VM with d
a
ta load on it
. The load wil
l
more la
rge i
n
volume an
d i
n
vestigatin
g
on it will b
e
t
he future dire
ction
s
to e
s
ta
blish th
e relia
ble cl
oud
data
cente
r
.
In the future,
the research
will discuss m
o
re
in
held further
resear
ch on the perf
ormance
of the CPU with a given load the data on each VM. The addition of VM for eac
h dedicated and
dynamic testi
ng in
order to produ
ce
da
ta that is mo
re va
ried. In
the future
ad
d serve
r
s in t
h
e
resou
r
ce pool
so that a larg
er amo
unt of memory an
d can g
ene
rate
a lot of VM.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
The Anal
yses on Dynam
ic and Dedi
cate
d Re
sou
r
ce
Allocatio
n
on…
(Mard
hani
Ri
asetia
wa
n)
285
Referen
ces
[1] Robertazzi
T
.
Data C
enters
.
Basic of C
o
mp
uter Net
w
o
r
kin
g
Spri
nger Br
ie
fs in Electrica
l
and
Comp
uter
Engi
neer
in
g. 2012: 69-
72.
[2]
Carlso
n M. S
ystems and
Virt
ualiz
atio
n Ma
n
agem
ent: Stan
dards
an
d the
Clo
ud (A r
e
p
o
rt on SVM
201
0).
Journ
a
l
of Netw
ork and
System Man
a
g
e
m
e
n
t.
Spring
er. 2011; 1
9
(4)
:
536-54
2.
[3]
Gao W
,
Jin H, W
u
S, Shi X,
Yuan J. Effectivel
y De
pl
o
y
in
g Services
on
Virtualis
atio
n Infrastructure
.
F
r
ontiers of Co
mp
uter Scie
nc
e.
Spring
er. 20
12; 6(4): 39
8-4
08.
[4]
Z
h
ixia
ng J,
Ro
ngch
a
n
g
Y, L
i
xin
L, F
angc
hu
n
,
T
ao
X, Ji
ao
D. Desi
gn
an
d
Ana
l
ysis
Unifi
ed
Reso
urce
Mana
geme
n
t P
l
atform of Grid
Dispatc
h
in
g S
yst
em Based
on
Virtualis
atio
n
T
e
chnolog
y.
TE
L
K
O
M
N
I
K
A
Indon
esi
an Jou
r
nal of Electric
al Eng
i
ne
eri
ng.
2014; 1
2
(3): 2
014-
202
0.
[5]
Joshi SC, Siv
a
lin
gam KM. F
ault T
o
leranc
e Mecha
n
isms
for Virtual Data Centre Ar
chitectures.
Photon
ic Netw
ork Co
mmu
n
ic
ations.
Sp
rin
ger
.
2014; 28(
2): 154-
164.
[6]
Che
n
X, Z
a
ng
J, Li J.
Reso
ur
ce Ma
nag
eme
n
t
F
r
ame
w
o
r
k f
o
r Co
ll
abor
ativ
e C
o
mputi
n
g
S
y
stems Over
Multiple Virtual Machines.
Se
rvice
Orient
ed Co
mp
uting an
d
App
licati
ons.
Sprin
ger
.
20
1
1
; 5(4): 22
5-
243.
[7]
Ma WLX
,
Shi Y, Guo Y.
A
Virtua
l Mac
h
in
e C
l
on
in
g
Appro
a
ch
Bas
ed
on T
r
usted C
o
mputi
n
g
.
T
E
LKOMNIKA Indon
esi
an Jou
r
nal of Electric
al Eng
i
ne
eri
ng.
2011; 1
1
(11):
693
5-69
42.
[8]
T
eng F
.
Mana
geme
n
t D
e
s D
onn
ees
Et Ord
i
nn
nanc
em
e
n
t
Des T
a
ches
S
u
r Arch
itecture
s Distrib
utes.
Desertati
on. C
entral
e
Paris: Ecole
C
enra
l
e
Paris Et Manuf
actures; 201
2.
[9]
Riaseti
a
w
a
n
M,
Mahmood AK.
Science-F
o
rge: A coll
ab
or
ative sci
entif
ic framew
ork
.
2010 IEEE
S
y
mp
osi
u
m on
Industrial El
ec
tronics & Appl
i
c
at
ions (ISIEA). Penang M
a
la
ysi
a
. 201
0: 665
-668.
[10]
Riaseti
a
w
a
n
M
,
Mahmood A
K
.
DALA Project: Digital ar
chive syste
m
for long ter
m
access,
201
0
Internatio
na
l
C
onfere
n
ce on Distribut
ed
F
r
a
m
e
w
ork
an
d A
pplic
atio
ns (DF
m
A). Yog
y
ak
ar
ta Indo
nesi
a
.
201
0: 1-5.
[11] Riaseti
a
w
a
n
M
.
Gama
Clo
ud:
T
he Dev
e
l
o
p
m
ent of C
l
uster
and
Grid M
o
d
e
ls b
a
se
d Sh
a
r
ed
me
mory
and MPI
. CIT
E
E 2012. Yo
g
y
a
k
arta Indon
esi
a
. 2012.
[12]
Riaseti
a
w
a
n
M
,
Ashari A. Resource Isol
at
io
n Anal
ys
is on Virtual Serv
er Performanc
e.
International
Journ
a
l of Scie
ntific & Engin
e
e
rin
g
Rese
arch
. 2014; 5(1): 22
9-55
18.
[13]
Varasteh A, G
oud
arzi M. Ser
v
er Cons
ol
idati
on T
e
chniq
ues
in Virtua
lise
d
Data Ce
nter, A
Surve
y
.
IEEE
System
s Journal
. 201
5; 9: 19
32-8
184.
[14]
Sun G, Lia
o
D
,
Anand V, Z
h
ao D, Yu
H. A Ne
w
T
e
chni
q
ue for Efficie
n
t Live Mi
grati
o
n of Multi
p
l
e
Virtual Mac
h
in
es.
F
u
ture Generatio
n Co
mp
u
t
er Systems,
2
016; 5.
[15]
Ilkhech
i
AR, Koreo
g
lu
e I,
Uluso
y O. Netw
o
r
k-a
w
a
r
e Vi
rtual Machi
nes
Placeme
n
t in Cloud D
a
t
a
Centers
w
i
t
h
Multipl
e
T
r
affic-intensive C
o
mpo
nents.
Co
mp
uter Netw
orks
. 2015; 91: 1
389-
128
6.
[16]
Al-A
y
y
oub M,
Wardat M, Jar
a
r
w
e
h
Y, Khr
e
i
s
hah
AA. Opti
mizing
E
x
p
ans
ion
Strategi
es
for Ultrasca
l
e
Clou
d
Data C
e
nters.
Simulati
on Mod
e
ll
in
g Practice an
d T
heory.
201
5; 58:
1569-
19
0
X
.
[17]
Hag
h
ih
at M, Z
ono
uz S, Ab
de
l-Mottale
b M. Clou
d
ID T
h
rust
w
o
rth
y
Cl
oud-
base
d
a
nd
Cro
ss-Enterpris
e
Biomentric Id
entification.
Expert System w
i
th Appl
icatio
ns.
201
5; 42(2
1
): 0957-
414
7.
Evaluation Warning : The document was created with Spire.PDF for Python.