TELKOM
NIKA
, Vol.13, No
.2, June 20
15
, pp. 739 ~ 7
4
4
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v13i2.1741
739
Re
cei
v
ed
De
cem
ber 8, 20
14; Re
vised
Ma
rch 14, 20
15; Accepted
April 10, 201
5
High Available VoIP Server Failover Mechanism in
Wide Area Network
Hery
Dian Septam
a
*
1
, Ar
dian Ulv
a
n
2
, Rober
t
Bestak
3
, Jiri Hla
v
acek
4
1,2
Department of Electrical En
gin
eer
i
ng, Un
iv
ersit
y
of L
a
mp
ung,
Jln. Soema
n
tri Bojo
neg
oro N
o
.
1 Bandar L
a
m
pun
g 35
145,
Indon
esi
a
3,4
Department of
T
e
lecommu
nicati
on En
gin
eeri
ng, Czech
T
e
chnical Un
iv
ersit
y
,
T
e
chnicka 2, 166 27 Pra
g
u
e
, Czech R
epu
bli
c
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: her
y@e
ng.un
ila.ac.i
d
1
, ardia
n
.ulva
n
@e
ng.u
n
ila.
a
c.id
2
,
robert.bestak
@
fel.cvut.cz
3
, hlav
aji
1
@fel.cv
u
t.cz
4
A
b
st
r
a
ct
Voice
over Int
e
rnet Protoc
ol
(VoIP) is Inter
net Protoc
ol
(IP) based
t
e
chn
o
lo
gy for
the next
gen
eratio
n n
e
tw
ork. VoIP has bee
n d
e
vel
o
ved to ov
erco
me f
u
ture tel
e
pho
ny d
e
m
an
d
.
How
e
ver, so
me
issues
i.e. how
to provi
de th
e
serv
ices
avai
la
bility
and r
e
li
ab
ility e
qua
lly to
circuit b
a
sed t
e
lep
hony,
nee
d
to
overco
me b
e
fo
re VoIP r
epl
ac
ed th
e circ
uit
b
a
sed
tele
ph
on
y. Live
migr
atio
n i
n
virtu
a
li
z
a
t
i
on
envir
on
men
t
is
used to
provi
d
e hi
gh av
ail
a
b
i
l
i
ty service, b
u
t the fa
il
over
me
chan
is
m over
W
i
de Are
a
Net
w
ork (W
AN) nee
d
to be
i
m
pr
ove
d
.
In this
pa
per,
the a
b
il
ity of v
i
rtuali
z
at
ion
fai
l
o
v
er
mec
han
is
m over
W
A
N a
n
d
vo
ice
qu
ality
of
VoIP servic
e in High Avail
ability system
is st
udi
ed
and ex
am
ined. Both
objectiv
e analys
is by
using quality
of services
(QoS) attribut
es i
s
cond
ucted
a
s
w
e
ll
as th
e
subj
ective
ana
l
ysis usi
ng M
e
an Opi
n
i
on Sc
ore
(MOS). T
he
w
o
rk utili
z
e
s Xe
n
®
Hypervisor w
i
th mo
difi
ed Re
mus ext
ensi
o
n
s
to provide th
e Hig
h Avail
abi
lity
envir
on
me
nt w
i
th GRE tun
nel
li
ng a
n
d
new
tw
ork virtual
i
z
a
t
io
n
.
Re
mus
ap
pro
a
ch us
in
g ch
ec
kpoi
nt bas
ed
is
dep
loye
d to co
py the pr
i
m
ary
server to the
b
a
ckup
s
e
rver.
A rang
e of 4
0
m
s –
90
0
m
s h
a
s be
en
app
lie
d as
time
interv
al
of checkp
oi
nt. T
he res
u
lts sh
o
w
that t
he the f
a
ilov
e
r d
o
w
n
ti
me
is 1.4
s a
n
d
mea
n
j
i
tter is
9,98
ms, pack
e
t los
s
3,12% a
nd
MOS 3.61 for Re
mus 4
0
0
m
s
checkpo
i
nt. MOS w
i
th different checkp
oint time
interva
l
is also
prese
n
ted
Ke
y
w
ords
:
Vo
IP, High Avail
a
bility, Virtua
li
z
a
tion, Tunne
li
ng
, W
A
N
1. Introduc
tion
Traditio
nal tel
epho
ny ca
rri
e
r
s
usi
ng
circu
i
t based
swit
ching h
a
s be
e
n
de
sign
ed
a
s
voi
c
e
carrie
r. It is a
very good technolo
g
y in early co
mmu
nication sin
c
e u
s
ers o
n
ly need
voice se
rvice
.
User’
s
nee
ds in communi
cation techn
o
logy are th
en increa
se
and varying,
such a
s
vo
ice
con
n
e
c
tion, v
i
deo
call,
em
ail, insta
n
t m
e
ssagi
ng,
we
b b
r
o
w
sin
g
a
nd oth
e
rs m
u
l
t
imedia
se
rvice
s
.
These n
eed
s, however
are not
suitabl
e to be
impl
emented
u
s
i
ng tra
d
itional
circuit
ba
se
d
swit
chin
g. Th
us, p
a
cket
ba
sed
switching
usi
ng Inte
rn
et Proto
c
ol
(IP) be
com
e
m
o
re
attra
c
tive to
fulfill this requirement.
Telecommu
ni
cation te
chno
logy is also i
m
pr
ove
d
, in whi
c
h the IP based techno
logy had
been
cho
s
en
as n
e
xt gen
eration
net
work.
Voice o
v
er Internet
Protocol
(VoIP) introdu
ce
d t
o
overcome fut
u
re telep
hon
y demand. Public
or p
r
i
v
ate networks have bee
n
utilised as
the
backb
one of
VoIP service. The
utilisation of esta
blish
ed inte
rnet netwo
rk
is one of V
o
IP
advantag
e, therefo
r
e telecommuni
catio
n
provide
r
s d
o
not need to build anothe
r speci
a
l network
to run
VoIP
se
rv
ice.
Ho
wev
e
r, th
es
e
rapi
d
cha
n
ges are h
a
ving p
r
obl
em
s, hen
ce
vario
u
s
resea
r
che
s
have been do
ne to make
sure IP tel
eph
ony could p
r
ovide availabi
lity and reliab
ility
equally to circuit base
d
tele
phony.
One of the most sig
n
ifica
n
t challen
g
e
s
in
next generati
on network is how to provi
de high
carrie
r-gra
de
for VoIP to overco
me the
publi
c
teleph
ony availabilit
y and fault to
lerant
system
.
Availability mean
s the a
b
il
ity of server
to be in
state to perfo
rm
requi
re
d services th
at cli
ent
requested at any given time inte
rval. T
e
lephony industries dema
nd the availability that have to
rea
c
h 99.
999
percent of the time, well
kno
w
n a
s
five nine rule
s, whi
c
h mea
n
s those net
wo
rks
must h
a
ve t
he maxim
u
m
of 5 mi
nute
s
d
o
wntime
over a
yea
r
[1]. VoIP as ne
w em
ergi
ng
techn
o
logy i
n
telepho
ny use
s
no
n rel
i
able prot
oco
l
since IP is firstly developed for p
a
cket
deliverin
g. Howeve
r, to be
able to re
pla
c
e the tr
aditio
nal teleph
ony
, this availabil
i
ty require
me
nt
should be full
filled.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 2, June 20
15 : 739 – 74
4
740
The
wo
rk o
n
voice
q
uality for V
o
IP ha
s b
een
do
ne
in
seve
ral
o
c
casi
on
s. It h
a
s
bee
n
repo
rted
in [
2
] that the t
r
ansmi
ssion
o
f
VoIP
can
g
enerate n
e
twork
cong
estio
n
du
e to
we
ak
sup
e
rvisi
on o
f
the traffic incomi
ng p
a
cket, quie
ng
and
sched
uling. The
con
gestio
n
affecte
d
several qualit
y of service (QoS) pa
ram
e
ters e.g,
un
stable voi
c
e packet delive
r
y, packet jitter,
packet lo
ss
and e
c
h
o
, etc. To
cop
e
with this
i
s
su
e, the autho
rs introdu
ced
a ne
w qu
eu
ing
sched
uling
al
gorithm
ba
se
d on
combin
e
d
PQ al
go
rith
m and
fuzzy l
ogic. T
he
pro
posed
algo
rithm
cla
ssifie
d
differentiate
d pa
cket from in
comin
g
pa
cket and also redu
ce
d re
cu
rsive loo
p
a
n
d
starving
as o
c
curred in n
o
r
mal PQ alg
o
r
ithm. At the
end it red
u
ce
d delay in Vo
IP network. In
addition, the
autho
rs i
n
[3] used
no
d
e
disj
oint
m
u
ltipath routi
ng p
r
oto
c
ol t
o
provide fa
ult-
toleran
c
e
sy
stem an
d hi
gh QoS to
alert pa
ck
ets cau
s
ed
by unu
sual a
n
d
critical eve
n
ts.
More
over, th
e FELGo
s
sipi
ng protocol is also d
epl
oye
d
to route the
routine d
a
ta
packet
s
, upd
ate
the neigh
bo
rs table and d
e
t
ect the failed
node
s. T
he result
s sh
owe
d
that the net
work laye
r ca
n
achi
eve sh
ort
average e
nd
to end delay
and very
low
packet lo
ss rate for alert p
a
ckets.
Re
cent d
e
vel
opment
s in th
e field of virtu
a
lizatio
n hav
e led to a
ren
e
we
d interest
in ho
w
to take adva
n
tage
s of virtualize
d
tech
nology
in order to su
ppo
rt high availability and fault
tolerant sy
ste
m
. Virtualize
d
environm
e
n
t offer hi
gh availability and or fault tolerant for server
usin
g
contin
u
ous live mig
r
ation of vi
rtu
a
l ma
ch
in
e b
e
twee
n p
r
ima
r
y serve
r
a
n
d
ba
ckup
se
rver.
No
wad
a
ys, virtualizatio
n
techni
que mostly
u
s
ed
by enterpri
s
e fo
r ha
rd
ware effici
en
cie
s
.
Therefore, t
he hig
h
ava
ilable
servi
c
e in virt
uali
z
ation is also
studie
d
. Li
ve migratio
n
in
virtualizatio
n is able to rep
licate ru
nning
server
into its backu
p se
rver. When fai
l
ure
s
occu
r the
backu
p se
rve
r
took ove
r
the
servi
c
e with
out disruption
.
The work in [
4
] studied
high available
abilit
y in Local
Area Network (LA
N
)
using Xen®
Hypervi
so
r. It is
an x8
6
virtual ma
chi
ne m
onitor t
hat allo
ws m
u
ltiple
comm
odity ope
rati
ng
system
s to
share
conventi
onal h
a
rdwa
re in a
sa
fe
a
nd resource
manag
ed fa
shion, but
with
out
sa
crifici
ng
either
pe
rform
a
nce
or fun
c
tio
nality. Th
is i
s
achieved
by
providin
g a
n
i
deali
z
ed
virtu
a
l
machi
ne ab
stractio
n
to whi
c
h
o
p
e
r
ating system
su
ch
as Li
nux, BSD an
d
Wind
o
w
s XP, can
b
e
ported
with minimal effort.
By allowing
100
op
eratin
g sy
stem
s to
run
on
a
sin
g
le
serve
r
i
n
the
experim
ental
system, Xe
n
®
Hype
rv
isor system
redu
ced
the asso
ciated
cost
s by
two ord
e
rs
of
magnitud
e
. F
u
rthe
rmo
r
e,
by turning
th
e setu
p
an
d
config
uratio
n
of each OS
into a
software
concern, it facilitated mu
ch smaller granularity
timescales of hosting.
The appli
c
ati
on tran
spa
r
e
n
t solution u
s
i
ng cont
in
uou
s live migratio
n in LAN is prese
n
ted
in [5]. A novel system
call
ed Re
mu
s was exploite
d for ret
r
ofitting high availa
bili
ty onto existing
softwa
r
e run
n
ing on
com
m
odity hard
w
are by usi
n
g virtualizatio
n to encap
sulate a prote
c
ted
virtual ma
chi
ne.The
freq
u
ent whole
system
ch
e
ckpoints was
perfo
rmed
to
asyn
ch
ro
no
us
repli
c
ate the
state of a singl
e
sp
ecul
ative executi
ng virtual m
a
chi
ne. Ho
wever, sin
c
e this
approa
ch i
s
n
o
t develop
ed
for real time
servi
c
e,
it g
e
nerate
d
ve
ry high jitter, l
a
tency
and
pa
cket
loss. Cla
ssifi
cation of packet propo
se
d by paper [6
] is used in ord
e
r to decrea
s
e those i
ssu
e
s
.
Wo
rk in [7] expand the re
sea
r
ch to si
mulate the
re
al serve
r
loa
d
by using
call gene
rator.
This
work
sh
owed
the me
asure
m
ent of the
o
b
jective
voi
c
e
quality an
alysis u
s
ing th
e
QoS attrib
ute
s
i.e., jitter, delay and pa
cket loss and
a
l
so utilises th
e su
bjective
analysi
s
u
s
in
g Mean
Opin
ion
Score (MOS
).
In orde
r to p
r
ovide
scalab
ility and better serv
ice, p
r
ovide
r
s of V
o
IP serve
r
m
a
y not
locate its server only in one location of data c
ente
r
. Even though establi
s
hi
ng several serve
r
s in
several lo
cati
ons i
s
hig
h
ly recomme
nde
d, con
s
e
quen
tly continuou
s live migration over th
e
wid
e
area e
n
viron
m
ent is req
u
i
r
ed. Wid
e
area cha
r
a
c
teri
stics e.g., lack of band
wid
t
h, higher del
ay,
jitter and
pa
cket lo
ss cau
s
e the repli
c
ati
on of
lin
ks m
o
re
unreliabl
e
and
be
come
new challen
ges
to overcom
e
. The failover mech
anism
of Xen with Remus exte
nsions utilise the
gratuitous
Addre
s
s Resolution Proto
c
ol
(ARP)
broad
ca
st. Ho
wever, thi
s
mech
ani
sm i
s
not
suitabl
e for
Wide Area
Network (WA
N
) sin
c
e A
R
P broa
dcast i
s
distrib
u
ted o
n
l
y on its own
local n
e
two
r
ks of
origin
Th
e im
pact
of WA
N
failover
fo
r
se
rver live
migration
wa
s ex
plaine
d in thi
s
p
ape
r [8]. T
h
is
pape
r also propo
sed tun
n
e
ling an
d the
dynamic Dy
namic
Nam
e
System (DNS) for wid
e
a
r
ea
packet
re-dire
c
tion, when t
he ba
ckup
se
rver
will
have
a ne
w IP ad
dre
s
s after fai
l
over. Seaml
e
ss
live migratio
n
over Met
r
op
olitan Are
a
Networ
k
(MAN) and
Wi
de
Area
Network (WAN) i
s
a
l
so
con
d
u
c
ted in
y this pap
er
[9]. Howeve
r,
those
pap
ers sce
n
a
r
io o
n
ly studie
d
the o
r
iginal li
ve
migratio
n inst
ead of co
ntin
uou
s live mig
r
ation. The
r
e
is no be
st sol
u
tion right n
o
w
sin
c
e a l
o
t o
f
researcher still working
on thei
r enhancement. T
he p
aper in [10] prov
ided work
i
n
the
enha
ncement
of contin
uou
s live mig
r
atio
n to study
the
probl
em of
continuo
us live
migratio
n ov
er
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
High Available VoIP Serv
er Failov
e
r Mec
h
anis
m
in Wide Area
Net
w
ork
.... (Hery
Dian Septama)
741
WAN
enviro
n
ment. Th
e
works p
r
o
p
o
s
ed
u
s
ing
B
o
rde
r
Gate
way
Proto
c
ol (BGP) upd
ate
to
anno
un
ce the
new path to the se
rver afte
r failover.
In this pa
per, the failove
r me
cha
n
ism
in
WAN u
s
ing G
R
E tu
n
neling
an
d n
e
twork
virtualizatio
n is propo
sed.
Do
wntime wh
en the failo
ve
r occu
rs a
nd t
he impa
ct to the voice q
uali
t
y
is investig
ate
d
and examin
ed.
2. High Av
ailable VoIP Challenges
Availability is the ability of the server to
be
able to serve the client
request at any given
time interval.
Circuit
based telephony i
s
desi
gned
to deliver high
av
ailability,
reliable connection
and hi
gh qu
a
lity of voice. Since VoIP relies
on IP, packeti
zation
and e
n
codin
g
pro
c
e
s
s m
a
y
cau
s
e
d
high
delay, jitter and packet lo
ss. Therefore,
to provide hig
h
available a
nd fault toleran
t
VoIP service
to met the telepho
ny stand
ard is
chall
e
n
g
ing.
The g
ene
ral
model
of hig
h
availabl
e V
o
IP se
rvi
c
e i
s
de
picte
d
in
figure
1. On
e se
rver
acts a
s
prim
a
r
y serve
r
that
se
rve th
e
cli
ents
re
q
u
e
s
t. T
h
e pr
imar
y s
e
r
v
e
r
s
t
a
t
es
ar
e fr
eq
u
entl
y
copi
ed to
the
ba
ckup
se
rver. Ba
ckup
server too
k
ov
e
r
th
e jo
b to
s
e
r
v
e th
e
c
lien
t
r
e
qu
es
t if th
e
prima
r
y se
rv
er fail with th
e last p
r
ima
r
y state copi
e
s
with mi
nim
a
l downtime
well kno
w
n a
s
fail
over me
cha
n
i
s
m.
Figure 1. Hig
h
availability model of VoIP Server
Virtualization technology offers
not only hardw
are efficiencies
but
also high availability
solutio
n
by continuo
us liv
e migratio
n o
f
its r
unnin
g
virtual machi
ne between t
he prim
ary server
and b
a
ckup
server. Live mi
gration
usi
ng
Xen is e
nh
a
n
c
ed
by the work i
n
[5] to provide a soluti
on
for transparent continuous live mi
gration using Rem
u
s.
Since Remus i
s
a lowe
r layer availability
prote
c
tion, th
erefo
r
e it
is
not requi
red
to modi
fy a
n
y
appli
c
ation
that ru
nnin
g
on top
of it.
The
Remu
s
wo
rk
by se
nt copie
s
of
all p
r
ima
r
y ma
ch
in
e st
ates to
the p
aused virtu
a
l
machi
ne
at the
backu
p serve
r
. Wh
eneve
r
f
a
ilure
s o
c
cu
r
on the p
r
ima
r
y serve
r
the
n
the pau
se
d
virtual ma
chi
n
e
at the ba
cku
p
serve
r
is resum
ed
with
the la
st
che
c
kpointin
g st
ates th
erefo
r
e do
wntime
of the
servi
c
e is mi
n
i
mized.
This failove
r
is tra
n
sparen
t from user
point
of view, backup
ma
chin
e sent g
r
atuitous
ARP b
r
oa
dca
s
t to it
s LA
N
segm
ent to
a
nnou
nce the
positio
n of vi
rtual ma
chin
e
s
. Thi
s
gratuitous
ARP notified
the cli
ent to
redire
ct the
p
a
cket
requ
est
to the
ne
w
b
a
ckup
ma
chi
ne
sin
c
e vi
rtu
a
l
machi
ne o
n
the prima
r
y
serve
r
ha
s b
een failed.
Ho
wever, thi
s
failover m
e
ch
ani
sm is not
accepta
b
le in
a wide
are
a
environm
ent
. The grat
uit
ous A
R
P bro
adcast may
not re
ceived
by
client in different net
wo
rk i.e., WAN conn
ec
tio
n
con
s
ist
s
of numbe
rs di
fferent network.
Therefore, th
e failover me
cha
n
ism to g
uara
n
t
ee the
minimal do
wn
time in wide a
r
ea environm
ent
sho
u
ld b
e
im
proved
to me
et WAN requ
ireme
n
ts. In t
h
is
pape
r th
e
tunneli
ng m
e
ch
ani
sm a
n
d
netwo
rk virtu
a
lizatio
n to overcome the p
r
oble
m
is con
s
ide
r
ed.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 2, June 20
15 : 739 – 74
4
742
3. Simulation and Re
sults
In this pape
r,
the prop
ose
d
idea of fail
over me
cha
n
ims over
WA
N and the im
pact of
high availa
bl
e se
rvice i
n
VoIP calls q
u
a
lity is
studie
d
and exa
m
i
ned. Packet
s durin
g VoIP call
usin
g Hig
h
Available VoIP servi
c
e is
ca
ptured
fo
r further obje
c
tive
voice quality
analysi
s
usi
ng
both the
QoS
attribute i.e.,
jitter, delay
a
nd p
a
ck
et lo
ss, an
d the
su
bjective
analy
s
is u
s
ing
Me
an
Opinio
n Score (M
OS). Th
e ITU
– T
st
anda
rd i
n
[11
]
descri
b
e
s
p
a
ram
e
ters for good
telep
h
ony
comm
uni
cati
on i.e., the averag
e MOS 4
.
4, delay <
25
0 ms, jitter < 30 ms an
d pa
cket loss < 5
%.
Tunneli
ng is
a comm
on way that used
in WAN to
provide a tra
n
sp
are
n
t net
work lin
k
betwe
en two
or more d
e
v
ices in diffe
rent netwo
rk segme
n
t or remote are
a
. Therefo
r
e,
b
y
cre
a
ting the t
unnel b
e
twe
e
n
prima
r
y to the ba
ckup
se
rver a
nd al
so
the client, it is en
su
ring th
e
migratio
n and
access traffic tran
spa
r
e
n
t. IP t
unneling
also ena
bled
virtual machi
ne to keep th
e
same IP add
ress. Ou
r propo
sed id
ea,
as depi
cted
on figure 2,
sho
w
s a hi
g
h
level mode
l of
failover mo
de
l in WA
N. Th
e wo
rks
pro
p
o
se tu
nnel
switching to
re
direct th
e net
work pa
cket when
failover o
c
curs. Only 1 Ge
neri
c
Routing
Enca
p
s
ul
ation (G
RE) tu
nnel o
r
acce
ss lin
k bet
we
en
client an
d ph
ysical
se
rver i
s
active/up, d
epen
ds
o
n
where th
e prot
ected virtu
a
l machi
ne i
s
re
lied
on.
`
Figure 2. Hig
h
level of test bed archite
c
ture.
If the protect
ed virtual
ma
chin
e i
s
runni
ng o
n
the
pri
m
ary
se
rver t
he G
R
E T
u
n
nel 1
i
s
activated. Thi
s
GRE Tu
nn
el 1 enable
d
client to
acce
ss the
serve
r
while the Re
mus p
r
o
c
e
ss
is
runni
ng throu
gh GRE T
u
n
nel 0 or
WA
N migrati
on
chann
el. The
measurement
on this pa
pe
r is
done
by lo
ad
ing u
p
the
Hi
gh Availabl
e
VoIP se
rver
t
hat ru
n in
p
r
i
m
ary
serve
r
with
call
s u
s
i
ng
VoIP call ge
nerato
r
with
minimum
CPU load
of 50
%, repre
s
e
n
ting the real server
wo
rk l
o
ad.
Then,
anoth
e
r
call i
s
mad
e
u
s
ing
SIP
Phone
for se
veral mi
nute
s
and
the
pa
cket i
s
sniffed
to
captu
r
e th
e d
a
ta pa
ckets f
o
r fu
rther obj
ective an
d
su
bjective voi
c
e
quality an
alyse
s
. All call i
s
usin
g SIP signalling p
r
oto
c
ols an
d G.71
1 cod
e
c
si
n
c
e
it is widely used by VoIP system
s.
After 1 minutes call, virtual machin
e on the prim
ary se
rver is
shuttin
g
down to trigger th
e
failover. Whe
n
failover
occured,
ba
ckup
machine
s
e
n
t
a notificatio
n to the
clien
t
to redi
re
ct the
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
High Available VoIP Serv
er Failov
e
r Mec
h
anis
m
in Wide Area
Net
w
ork
.... (Hery
Dian Septama)
743
traffic to the backu
p ho
st machi
ne. Thi
s
notif
icatio
n
drop
s the G
R
E Tunnel 1 a
nd GRE T
u
n
nel 2
are e
s
tabli
s
h
ed to gui
de t
he traffic into
the ba
ckup
machi
ne to
minimize the
downtime. T
he
voice call
should not dropped since our system is
high available but the voice will be interupted
as lo
ng
as th
e se
rver do
wntime. All ca
p
t
ured
pa
cket
is an
alyze
d
u
s
ing
pa
cket sniffer for fu
rth
e
r
qualitative an
alysis a
nd u
s
i
ng perce
ptual
voic
e quality measurement
for subje
c
tive analysi
s
.
In this
re
sult,
jitter calcul
ation i
s
con
s
id
e
r
ed
to p
r
e
s
en
t the imp
a
ct
of Re
mu
s
co
ntinuou
s
live migration
in VoIP communication
s. Real
-time
Tra
n
sp
ort Proto
c
ol (RTP
) is a
proto
c
ol defin
ed
by IETF RFC 3550 [12] that is respon
sible to prov
id
e real time d
a
ta delivery servi
c
e. Jitter is
cal
c
ulate
d
using interarrival jitter (J) an
d mean
d
e
viation of the d
i
fference (D)
defined fo
r p
a
irs
packet a
s
sh
own o
n
equ
ation 1.
(1)
Figure 3
shows the impact
in High Availability
VoIP service. T
he de
lay, jitter and packet
loss are still
accepta
b
le. Maximum jitter is
17,02 m
s
, the mean jitter only 9,98 ms, packet
loss
also
only
3,1
2
%. Furth
e
r
analysi
s
of subje
c
tive attri
bute u
s
in
g M
O
S is e
s
tima
ted u
s
ing
voi
c
e
quality analy
s
er. Thi
s
an
alyser
wo
rks by sendin
g
the reco
rd
e
d
voice du
ri
ng the call
and
comp
ared the echo
ed voi
c
e with o
r
igin
al voice
and
estimate
s the
MOS. This work al
so proof
that use
r
defi
ned
che
c
kpoi
nting interval
has
a si
g
n
ificant imp
a
ct i
n
voice q
ualit
y but doe
s n
o
t
impact the
d
o
wntime i
n
terval, the resul
t
is sho
w
n
i
n
table 1. Accordin
g to the
result, the most
suitabl
e ch
eckpoi
nt interva
l
is every 400
– 600 ms.
Figure 3
.
Jit
t
e
r lev
e
l in mod
i
f
i
ed Remu
s
HA
sy
st
em
Table 1. Ch
e
c
kpoint interv
al impact
No
Remus
checkpoint
i
n
terv
al
(ms
)
Do
w
n
time (s)
MOS
1 40
1.3
2.73
2 70
1.3
2.84
3 100
1.2
2.95
4 150
1.4
3.12
5 200
1.3
3.23
6 300
1.4
3.40
7 400
1.4
3.42
8 500
1.3
3.42
9 600
1.5
3.49
10 700
1.5
3.48
11 800
1.4
3.48
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 2, June 20
15 : 739 – 74
4
744
4. Conclus
s
ion and Futur
e Work
s
This work sh
ows that tunnelling and virt
ual net
wo
rk is promi
s
ing for failover mechani
sm.
The
result
show that the downti
me i
s
still comparabl
e to the grat
itous A
R
P. The impact of
hi
gh
available VoI
P
Server in
WAN to th
e ji
tter, delay
, p
a
cket lo
ss
an
d MOS of voi
c
e q
uality is
still
accepta
b
le
a
c
cordi
n
g
to t
he g
u
ideli
n
e
s
of ITU-T.
Th
e results
sh
o
w
s that th
e
mean
jitter i
s
9,98
ms, pa
cket loss 3,12% a
nd MOS 3.4
2
for Rem
u
s 400m
s ch
e
c
kpoint. Furt
her
work sh
ould
con
s
id
er
the
optimizatio
n of
packet co
mpre
ssion du
e to network
cha
r
a
c
teri
stics i.e., high
de
lay,
jitter and pa
cket loss in WAN
Referen
ces
[1]
Gra
y
J, Sie
w
i
o
rek DP.
Hig
h-a
v
aila
bi
lit
y C
o
m
puter S
y
stems.
IEEE Comput
er Maga
z
i
nes.
199
1;
2
4
(9):
39–
48, ISSN: 001
8-91
62.
[2]
Suard
i
nat
a, Abu BK. A F
u
zz
y
L
ogic C
l
as
sification of Incomi
ng Packet for VoIP.
TELKO
M
NI
KA
T
e
leco
mmunic
a
tion C
o
mputi
n
g Electron
ics a
nd Co
ntrol.
20
10; 8(2): 16
5–1
74.
[3]
Mekki K, Z
oui
n
k
hi A, Abd
e
lkri
m MN. F
ault-tole
ra
nt and QoS
Net
w
ork L
a
y
er
for Securit
y
M
ana
geme
n
t.
T
E
LKOMNIKA T
e
leco
mmunic
a
tion C
o
mputi
n
g Electron
ics a
nd Co
ntrol.
20
13; 11(2): 3
63–
372.
[4]
Barham P, Dra
govic B, F
r
aser K, Hand S, Harri
s
T
,
Ho A, N
eug
eb
auer R, Pratt I, W
a
rfield A.
Xen an
d
T
he Art of Virtuali
z
a
t
io
n.
Pro
c
eed
ings
of the nin
e
tee
n
th
ACM S
y
m
posi
u
m on Oper
ati
ng S
y
stems
Princip
l
es. 20
0
3
: 164–
17
7.
[5]
Cully
B, Lefebvre G, Mey
e
r
D, F
eele
y
M, Hutchi
nson
N, W
a
rfield A.
R
e
mus: Hig
h A
v
aila
bi
lity via
Asynchro
no
us Virtua
l Mac
h
ine
Re
plic
atio
n.
Proce
e
d
i
ng
s of the
5th
USENI
X
S
y
mposi
u
m
o
n
Net
w
ork
ed S
y
s
t
ems Desig
n
a
nd
Implem
enta
t
ion. 200
8: 161
–17
4.
[6]
Hlavac
e
k J, Bestak R.
Live
Repl
icati
on of
Virtuali
z
e
d
V
o
IP Servers
. T
h
e Eight
h Intern
ation
a
l Mu
lti
-
Confer
ence
on
Computi
ng in t
he Glob
al
Infor
m
ation T
e
chno
log
y
(ICCGI). 2013.
[7]
Septama
HD,
Ulvan A, H
l
a
vacek JB.
V
o
ice Qu
ality o
f
VoIP in Hig
h Avail
a
b
ility
Enviro
n
m
ent
.
Internatio
na
l C
onfere
n
ce o
n
Electrical
E
ngi
ne
erin
g and C
o
m
puter Scie
nce.
201
4
[8]
Bradford
R, Ko
tsovinos E, F
e
l
d
man
n
A, Sch
i
¨ob
e
rg H.
L
i
ve
W
i
de-Are
a
Mi
gratio
n of Virtu
a
l Mac
h
in
es
Includ
ing
Loca
l
Persistent Sta
t
e
. VEE '07 Proceedings of the 3rd
in
ternational conf
er
enc
e on Virtual
executi
on e
n
vir
onme
n
ts. 2007
.
[9]
T
r
avostino F
,
Daspit P, Gom
m
ans L, Jo
g C
,
de
La
at C, Mambretti J, Mo
nga I, van Ou
d
ena
ardec B,
Rag
hun
atha S,
W
ange P. Se
amless l
i
ve mi
gratio
n of virtu
a
l mach
in
es o
v
er the MAN/
W
A
N.
Fu
tu
re
Generati
on C
o
mp
uter Syste
m
s 22
. 200
6.
[10]
Raja
go
pal
an
S
., Cull
y B,
O'
C
onn
or R, W
a
rfi
e
ld A.
S
e
co
nd
Site: dis
a
ster t
o
ler
ance
as
a
service
. VEE
'
12 Procee
di
ng
s of the 8th ACM SIGPLAN/SIGOPS
conference o
n
Virtu
a
l Ex
ecuti
on E
n
viro
nments.
201
2.
[11]
One-
w
a
y tra
n
s
m
ission
time. I
T
U.T
Recomm
end
atio
n G.11
4
. Internati
o
n
a
l
T
e
lecommuni
cation
Un
ion
T
e
lecommunic
a
tion Stan
dar
di
zation Sector (I
T
U
.T
). 2003.
[12]
Schulzri
n
n
e
H,
Casn
er S, F
r
ederick
R, Ja
c
obso
n
V. RT
P: A
T
r
anspor
t Protocol for
Real-T
ime
Appl
icatio
ns. IET
F
RF
C 3550
. Internet
Engin
eeri
ng T
a
sk F
o
rce (IET
F
)
. 200
3.
Evaluation Warning : The document was created with Spire.PDF for Python.