TELKOM
NIKA
, Vol.14, No
.4, Dece
mbe
r
2016, pp. 15
34~154
4
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v14i4.3295
1534
Re
cei
v
ed
De
cem
ber 1
8
, 2015; Re
vi
sed
Septem
ber 4, 2016; Accept
ed Septem
be
r 19, 2016
A Methodology for Characterizing Re
al-Time
Multimedia Quality of Service in Limited Bandwidth
Network
Yoanes Ban
dung*
1
, Idi S
u
mardi
2
1,2
School of Ele
c
trical Eng
i
ne
e
r
ing a
nd In
form
atics, Institut
T
e
kno
l
og
i Ban
d
ung,
Jala
n Ganes
a 10 Ban
d
u
ng, In
don
esia, Ph.: +
62-2
2
-25
0
0
9
8
5
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: band
un
g@st
ei.itb.ac.id
1
, idi
s
umardi
@gm
a
i
l
.com
2
A
b
st
r
a
ct
T
h
is pap
er pre
s
ents how
to character
i
z
e
the
qual
it
y of mu
l
t
ime
d
i
a
w
h
ich
consists of au
dio a
n
d
vide
o that ar
e
transmitted i
n
real-t
i
m
e c
o
mmu
n
ic
ation thr
oug
h the Inter
net w
i
th li
mite
d ba
ndw
idth.
W
e
deve
l
op
ed a method
ol
ogy of
character
i
z
i
n
g
t
he mu
lti
m
ed
ia Quality-of-Ser
v
i
ce
(QoS) by measur
ing
netw
o
rk
para
m
eters (i.e
., bandw
idth ca
pacity, pack
e
t loss rate (PLR),
and e
nd-to-
e
n
d
del
ay) of testbed n
e
tw
ork a
n
d
simulati
ng
the
aud
io-v
ide
o
deliv
ery
accor
d
in
g to th
e
measur
ed
netw
o
rk par
a
m
eter
s. T
he an
alysi
s of
netw
o
rk para
m
eters w
a
s aime
d to describ
e the netw
o
rk
cha
r
acteristics. Multimedi
a QoS w
a
s character
i
z
e
d
by cond
uctin
g
a simul
a
tion
u
s
ing d
a
ta w
h
ic
h w
a
s collecte
d
from the
pre
v
ious n
e
tw
ork character
i
z
a
tio
n
. A
simulati
on n
e
t
w
ork mod
e
l w
a
s bui
lt usin
g
OMNet+
+
r
epresenti
ng a
del
i
v
ery of aud
io-
v
ide
o
in re
al-ti
m
e
w
h
ile a b
a
ckg
roun
d traffic w
a
s gener
ate
d
to repr
esent
a real c
o
n
d
iti
on of the
net
w
o
rk. Apllyin
g
the
meth
od
olo
g
y in
a netw
o
rk testbed i
n
Indo
nes
ia
’
s
rur
a
l ar
e
a
, the si
mul
a
tio
n
results show
e
d
that aud
io-vi
d
e
o
coul
d be d
e
liv
e
r
ed w
i
th accept
ed lev
e
l of user
satisfaction.
Ke
y
w
ords
:
qua
lity of s
e
rv
ice,
multi
m
edi
a
co
mmu
n
ic
ati
on, a
udi
o a
n
d
vide
o, n
e
tw
ork charact
e
ri
z
a
t
i
on,
li
mited ban
dw
idth
netw
o
rk
Copy
right
©
2016 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
Multimedia
communi
catio
n
over th
e In
ternet h
a
s
b
een the
subj
ect of ma
ny
resea
r
ch
works ove
r
th
e last fe
w ye
a
r
s [1]
-[3] . M
u
lt
imedia co
mmuni
cation techn
o
logy
h
a
s been
avail
able
to the tran
sm
issi
on of a v
a
riety types
of medi
a
su
ch as text, graphi
cs, a
n
im
ation, audi
o, an
d
video ove
r
th
e Intern
et tha
t
are
ca
rri
ed
out conti
nuo
u
s
ly into d
a
ta
strea
m
. Basi
cally, multimedia
strea
m
ing
ap
plicatio
ns
are
divided i
n
to
3 catego
ri
e
s
[
4
] : (i)
stream
ing of
stored audio/video, (ii)
strea
m
ing
of live audio/vid
eo, and
(iii)
conversation
al
of voice/
vide
o-ove
r-IP. Th
ese
cate
gori
e
s
are also
known
as (i
) video on demand, (ii)
live streaming, and (iii)
real
time streaming.
Fi
rst,
video on de
mand allo
ws use
r
s to p
e
r
form pl
ayba
ck
cont
rol su
ch a
s
pa
use, rewi
nd, or f
a
st
forwa
r
d. Se
cond, live stre
aming all
o
ws use
r
s
to
re
ceive live broad
ca
sts
such as b
r
o
adcast
television o
r
radio. Third, real time strea
m
ing a
llows u
s
ers to com
m
unicate with a
udio or video
in
real
-time.
One ap
plication of multim
edia commu
nicati
o
n
tech
nology for e
ducation is
dista
n
c
e
learni
ng thro
ugh virtual
classro
o
m te
chnolo
g
y.
This technolo
g
y allows a le
arnin
g
proce
s
s
facilitated by
a teacher in a real
classroom
c
an
be virtually shared in
to
other classroom
s at
different lo
cat
i
ons [5] . With
virtual cla
s
sroom
technolo
g
y, learning
p
r
ocess b
e
twe
en two o
r
more
cla
s
sroo
ms can b
e
con
d
u
c
ted through
a
re
al-time
mu
ltimedia
com
m
unication
o
v
er the
Intern
et.
Learning
pro
c
e
ss in
ea
ch
cla
s
sroo
m is
recorded into
audio a
nd video an
d tran
smitted bet
wee
n
those
classro
o
ms. Moreov
er, data also can be
comm
unicated bet
wee
n
cla
s
sro
o
ms for exam
ple
in remote p
r
e
s
entatio
n or d
e
sktop shari
n
g.
Quality of audio and vide
o durin
g the
learnin
g
pro
c
e
ss th
rou
g
h
a virtual cla
s
sroo
m
techn
o
logy b
e
com
e
s a
sig
n
ificant fa
ctor for the
su
cce
ssful delivery of
lear
ning m
a
terial
s. Thi
s
is
becau
se stu
d
ents in virtu
a
l cla
s
sroom
p
a
rticip
ate vi
rtually in a lea
r
ning p
r
o
c
e
ss
with a tea
c
he
r in
the real
cla
ssroom. T
hus, t
he comfort of
learni
ng m
a
terial
s delive
r
y depen
ds
o
n
the qu
ality of
audio
-
video
transmitted
through th
e Inte
rnet. Lo
w
qu
ality of audio
-
video can m
a
ke th
e lea
r
ni
ng
pro
c
cess
bei
ng inte
rru
pte
d
or even
cannot b
e
h
e
l
d
. Our stu
d
y reveal
ed th
at the qu
ality of
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
A Methodolo
g
y for
Cha
r
a
c
terizin
g
Re
al-Tim
e Multim
e
d
ia Qualit
y of… (Yoan
es B
andu
ng)
1535
multimedia service
s
whi
c
h
produ
ce a
larg
e
am
ou
nt of data [6
] can
be aff
e
cted
by several
netwo
rk pa
ra
meters [7]
su
ch
as ban
dwi
d
th, end
-t
o-e
nd d
e
lay o
r
l
a
tency, jitter,
and
pa
cket l
o
ss
rate
(PLR).
Whe
n
the
av
ailable
ban
d
w
idth
be
com
e
na
rrow, th
e
audi
o a
nd vi
deo t
r
an
smi
s
sion
may be di
stu
r
bed [5] . Mo
reove
r
, the q
uality of audi
o and vid
eo requires
end
-t
o-en
d del
ay and
packet lo
ss rate as lo
w as
possibl
e.
The co
ncept of quality of service
(QoS
) is
define
d
in ITU-T Re
com
m
endatio
n E.800 [8]
as the
coll
e
c
tive effect
of se
rvice p
e
rform
a
n
c
e,
whi
c
h d
e
termines th
e d
egre
e
of u
s
ers
sat
i
sf
a
c
t
i
on [
9
]
. QoS aims to provide a
better qu
ality of
servi
c
e for
variou
s ne
ed
s of the existi
ng
netwo
rk infra
s
tru
c
ture so t
hat users g
e
t satisf
a
c
tion
i
n
usi
ng n
e
twork-b
a
sed a
p
plicatio
ns. Q
o
S
can
arran
ge the provisio
n
of diffe
rent
service
s
q
ualit
y
for
the
dive
rse nee
ds of
servi
c
e su
ch as
providin
g sp
ecific
ban
dwi
d
th, decre
asi
ng pa
cket
lo
ss, d
e
cre
a
si
n
g
delay time
and jitter.
Q
o
S
function
s can
be descri
be
as follo
ws [1
0]
: (1) gradi
ng pa
ckets to
prov
ide diffe
rent se
rvices
for
each cl
ass o
f
packets;
(2
) con
g
e
s
tion
handli
ng to h
andle th
e ne
eds
of differe
nt se
rvice
s
; (3
)
controling
of packet traffi
c to re
strict a
nd contro
lli
ng
delivery of d
a
ta pa
ckets;
(4)
sig
naling
for
control the de
vice functio
n
s that s
uppo
rts commu
nication in the Internet.
There are
se
veral ob
stacl
e
s whi
c
h
can
affect
the quality of audio, video, or other data
transmitted throu
gh the
Internet. O
n
e
of the ob
st
a
c
le
s is
availa
ble ba
nd
widt
h of the net
work
whi
c
h is typically very limited parti
cula
rl
y in ru
ral are
a
s like in Indone
sia [5] . Increa
sing a
c
ce
ss
links of the network is ve
ry expensive
and often ta
kes several ye
ars fo
r their d
eployment [1
1] so
that the Internet acce
ss is
varied bet
we
en regi
on
s.
Another o
b
sta
c
le is asym
etri
c pro
perty of the
netwo
rk
whe
r
e the band
wi
dth capa
ci
ty of the upload
link is differe
nt with those
of the downlo
ad
link. Typically, the upload l
i
nk of the n
e
twork i
s
great
er than th
e d
o
wnl
oad lin
k
of the netwo
rk.
Tech
nically, real time
mult
imedia
comm
unication
req
u
ire
s
symmet
r
ical
b
and
wid
t
h ca
pa
city i
n
orde
r to tran
smit the mult
imedia which
con
s
ist
s
of audio, video,
or othe
r dat
a in full-du
pl
ex
comm
uni
cati
on scen
ario.
In this pap
er,
we aim to
study the effects
of netwo
rk cha
r
a
c
ters to the audi
o a
nd video
quality in a re
al-time multi
m
edia d
e
livery espe
cia
lly i
n
limited ba
n
d
width
netwo
rk. Mo
reove
r
,
we
prop
ose a methodol
ogy
for cha
r
a
c
terizi
ng mult
i
m
edia QoS
whi
c
h is aff
e
cted by ne
twork
con
d
ition
s
. There h
a
ve be
en many stud
ies on t
he me
asu
r
em
ent of network an
d multimedia, b
u
t
only a fe
w
studie
s
on
the
developm
ent
of a meth
odol
ogy
for ch
ara
c
teri
zation
of multimedia Q
o
S
in
limited ban
dwidth
netwo
rk.
In previou
s
studie
s
of
[5] [11] [12] ,
netwo
rk
charac
teriz
a
tion
was
done by mea
s
uri
ng the bro
adba
nd networks. Mea
s
u
r
ement ho
sts
were co
nne
ct
ed to high-sp
eed
aca
demi
c
net
work. Study in [12] propo
sed met
hod
olo
g
y for charact
e
rizi
ng re
sid
e
ntial broa
dba
nd
netwo
rks but
did
not yet
inclu
de th
e
chara
c
te
rizati
o
n
of m
u
ltime
d
ia Q
o
S. In
this te
chni
qu
e,
netwo
rk me
a
s
ureme
n
t wa
s
con
d
u
c
ted
by usi
ng
a p
r
obe
trai
n
se
nt from
ho
sts located i
n
fo
ur
aca
demi
c
n
e
t
works - three in
No
rth
Ameri
c
a
an
d on
e in Eu
rope. In
the
study
of [11]
,
cha
r
a
c
teri
zin
g
real
-time
video traffic
in re
sidential
broad
ban
d netwo
rks
wa
s co
ndu
cted
by
analyzi
ng n
e
tworks pe
rforma
nce for transmi
tting
real-tim
e video. Measurem
ents
were
perfo
rmed b
e
twee
n DSL
end-h
o
st
s in apartm
ent compl
e
xes a
nd sin
g
le ho
use
s
which
we
re
con
n
e
c
ted to
the university network. In [5]
, network mea
s
u
r
em
ent and
simu
lation wa
s d
o
n
e
from DSL ho
st located in ru
ral area conn
ected to a hig
h
-spee
d acad
emic net
wo
rk.
2. Methodol
og
y
for Multimedia QoS Char
acteriza
t
ion
This se
ction
descri
b
e
s
o
u
r
pro
p
o
s
ed
me
thodolo
g
y
to cha
r
a
c
teri
ze
multimedia Q
o
S
wh
en
variou
s medi
a su
ch a
s
au
dio, video, an
d data
are transmitted
over limited b
a
ndwi
d
th network.
Basically, the methodolo
g
y
prese
n
ts two impo
rtant
procedu
re
s:
i) network chara
c
te
rizatio
n
to
obtain net
wo
rk
con
d
itions base
d
on
several net
wo
rk p
a
ra
meters mea
s
u
r
em
ent in testbe
d
netwo
rk
and
ii) netwo
rk simulatio
n
to
dra
w
the
q
uality of multimedia
conte
n
ts if tran
smi
tted
across the ne
twork. The p
r
opo
sed meth
odolo
g
y c
an
be de
scribe
d in more d
e
tail
as follows.
2.1. Choosin
g Net
w
o
r
k P
a
rameters fo
r Measur
e
ment
Several
studi
es
reveal that network paramete
rs
such as bandwidt
h availability, end-to-
end del
ay, an
d packet lo
ss rate have g
r
eat impa
ct
on
the quality of audio, video,
or data in th
e
multimedia delivery services [5] [11
]
[12] . Based o
n
those stu
d
ies,
we sugge
st to ch
oo
se
th
e
three
pa
ram
e
ters to b
e
m
e
asu
r
ed
o
n
th
e multime
d
ia
QoS
cha
r
a
c
terization.
We
present
seve
ral
rea
s
on
s be
hi
nd the sel
e
cti
on of
the parameters a
s
follows.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 4, Dece
mb
er 201
6 : 1534 – 154
4
1536
1.
In data t
r
an
smissi
on, b
a
n
d
width
capa
city is
related
to
the width of
the comm
unication pip
e
and ho
w qui
ckly bits ca
n b
e
sent.
2.
Multimedia
QoS will decline wi
th increasing
delay time. Moreov
er, if the multimedia
contents
spe
nd m
u
ch
time towa
rd
s
the de
stinatio
n ho
st,
it can cause failures in
real-time multimedi
a
comm
uni
cati
on.
3.
Packet loss
can b
e
o
c
cu
rre
d wh
en th
e packet
s
a
r
e disca
r
de
d durin
g its tra
n
smittion to
destin
a
tion h
o
st. Thi
s
i
s
b
e
ca
use of
co
nne
ction
failu
re, sudd
en
ro
ute chang
e, traffic overl
oad
,
or co
nge
stion
in the netwo
rk. In gene
ral
,
packets loss is the majo
r cau
s
e of de
grad
ation o
n
audio a
nd vid
eo quality.
2.2. Measuri
ng and Inv
e
s
t
igating
Rur
a
l Net
w
o
r
k
Network mea
s
ureme
n
t aims to captu
r
e variation
s
of network p
r
opertie
s
.
Me
asu
r
em
ent
method used in
this study can be de
scri
bed
a
s
foll
ow
s. 1) Band
wid
t
h cap
a
city of
netwo
rk lin
k
is
measured b
y
using T
C
P (Tran
s
fe
r Control Prot
oc
ol
) traffic.
In our exp
e
rime
nts, se
veral
measurement
s we
re done
repe
atedly
u
s
ing certai
n
ti
me inte
rvals
and t
r
affic l
o
ad. Acco
rdin
g to
the stu
d
y of [
5
] [11] [12] ,
minimum
me
asu
r
em
ents should
be
co
n
ducte
d d
u
rin
g
a
day. Here,
each iteration
of mea
s
u
r
e
m
ents fo
r
up
to 10
se
con
d
s
wa
s
co
nd
ucted
with
int
e
rval time
of
30
minutes. Ba
ndwi
d
th capa
city wa
s me
a
s
ured by fl
oo
ding the li
nk
with data
pa
ckets
se
nt fro
m
a
sou
r
ce ho
st to a de
stinatio
n ho
st. This
scen
ario
wa
s
aimed to
satu
rate ba
nd
widt
h
.
2) End
-
to-e
nd
delay was
m
easure
d
from
host
-
to-h
ost
usin
g IC
MP (Internet Cont
rol
Me
ssa
ge P
r
otocol)
pa
ckets
with diffe
rent
si
ze
s. Fi
rst,
so
urce
no
d
e
sent
the
same si
ze
I
C
MP
echo re
q
uest pa
ckets
and
destin
a
tion n
ode
re
spo
n
d
ed with
the
sam
e
size
ICMP e
c
ho
re
spo
n
se p
a
ckets. Se
co
nd,
measurement
was cond
ucted us
ing the
MTU (Maxi
m
um Tra
n
sm
issi
on Unit
) value on ADS
L
(
A
s
y
mme
tr
ic
D
i
g
i
ta
l Su
bscr
ib
er
L
i
ne
)
ne
tw
o
r
k
.
Me
asurem
ents
we
re p
e
rfo
r
med
rep
eatedly i
n
a
day by usin
g a sp
ecifi
c
interval to chara
c
te
ri
ze the end
-to-en
d delay. 3)
Packet loss
rate
para
m
eters
were mea
s
u
r
ed usi
ng
UDP (User
Data
gram Protocol) by se
ndin
g
UDP d
a
tag
r
a
m
packet
s
from
source to d
e
stinatio
n by using
different traffic load. We u
s
ed
UDP datag
ram
packet
s
b
e
ca
use
it
could
repre
s
e
n
t mult
imedia
delive
r
y. UDP d
a
ta
gram
pa
ckets with
si
ze
14
70
bytes per d
a
tagra
m
we
re sent repe
atedl
y with different data rate.
Cha
r
a
c
teri
zat
i
on of li
mited
ban
dwi
d
th n
e
twor
k wa
s d
one
by inve
stigating
ea
ch
netwo
rk
para
m
eters
measured in
the previo
u
s
se
ction. 1
)
Bandwidth
cap
a
city wa
s investigated
and
cha
r
a
c
teri
ze
d
by using th
e re
sult of b
and
width me
asu
r
em
ent. The data
wa
s retrieve
d fro
m
sou
r
ce ho
st and de
stinatio
n host du
ring
the m
easu
r
e
m
ent pro
c
e
ss. We investig
ated band
wid
t
h
prop
ertie
s
, i.
e. allo
cated
band
width,
u
p
stre
am
and
do
wn
stre
am
ca
pa
city, an
d ratio b
e
tween
upstream a
n
d
downst
r
ea
m cap
a
city. 2) End
-
to-
e
n
d
delay characteri
zatio
n
wa
s co
ndu
ct
ed by
usin
g RTT
(Rou
nd T
r
ip
Time) va
riati
on of ICMP
packet
s
deliv
ery. End-to
-e
nd del
ay bet
wee
n
sou
r
ce ho
st to destinatio
n host
wa
s obtaine
d by
cal
c
ulatin
g h
a
lf of the RT
T (equ
ation (1)).
Investigation
of end-to
-en
d
delay refers
to IT
U-T re
co
mmend
ation
G.114 [13] . The delay th
at
occurre
d
in e
a
ch
se
co
nd d
u
ring t
he me
asu
r
em
ent proce
s
se
s was also i
n
vestig
ated. 3) Pa
cke
t
loss
rate i
s
th
e num
be
r of
packet
s
whi
c
h a
r
e
not
received by
de
stination
ho
st. In ou
r
ca
se
st
udy,
PLR was
cha
r
acte
ri
zed by
usin
g mea
s
u
r
ement re
sult
s from two different lo
cation
s at the testb
e
d
netwo
rk. PL
R was
cal
c
ulat
ed usi
ng equ
ation (2
).
Furt
herm
o
re, we investigate
d
wheneve
r
pa
cket
loss ha
s
occured,
wh
ethe
r it ha
s o
ccurred d
u
ri
ng
the me
asurement p
r
o
c
e
s
s or ha
s o
n
ly
occurre
d
at certain time
s.
We al
so inve
stigated
ho
w much th
e pa
ckets l
o
ss at
certain time
s a
n
d
how ma
ny times the em
ergen
ce of pa
ckets
lo
ss duri
ng the mea
s
u
r
eme
n
t pro
c
e
ss.
-
-
(
1
)
100%
(
2
)
2.3. Simulating and Ch
ar
act
eriz
ing Multimedia QoS
We characte
rized multime
d
ia QoS thro
ugh network
simulatio
n
wi
th a simulati
on model
that appro
a
ches real net
work
con
d
ition
s
. The si
m
u
la
tion wa
s set
up acco
rdin
g
to the result
s of
measurement
param
eters: band
width, e
nd-to
-en
d
del
ay, and packet loss
rate.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
A Methodolo
g
y for
Cha
r
a
c
terizin
g
Re
al-Tim
e Multim
e
d
ia Qualit
y of… (Yoan
es B
andu
ng)
1537
2.3.1. Simula
ting Multimedia Deliv
er
y
In the s
i
mulation, we were s
e
nding a
udio and video pac
k
et along with back
ground
traffic. Param
e
ters
co
nfigu
r
ed for the
deli
v
ery of
audio
were data
rat
e
and
size of
the tran
smitte
d
audio. In
this re
sea
r
ch,
we u
s
ed
data
rate which rep
r
esenting
Vorbis te
ch
nolog
y. Same a
s
t
h
e
delivery of a
udio, data
ra
te for video
delivery c
an
use th
e si
ze
req
u
ire
d
for the delive
r
y o
f
comp
re
ssed
video such a
s
VP8 [14] .
Based
on thi
s
data rate, we dete
r
mine t
he len
g
th of
each
packet
and t
he inte
rval b
e
twee
n pa
ckets. The
si
ze
of audi
o a
n
d video
pa
ckets u
s
e
d
in t
h
e
simulatio
n
d
e
pend
s
on th
e
data
rate
an
d du
ration
of
sim
u
lation
a
nd
wa
s
cal
c
u
l
ated u
s
in
g t
he
equatio
n (3
).
∗
∗
(
3
)
For me
cha
n
ism of audio and video deli
v
ery, we
installed a VoIP
appli
c
ation in
the first
host
as
a
se
nder an
d VoI
P
appli
c
ation
in the
se
c
o
n
d
ho
st a
s
a
receive
r
. Simil
a
rly, the vide
o
delivery u
s
ed
UDP ap
plica
t
ions in
stalled
together
with
VoIP applica
t
ions which the first ho
st
as
serve
r
and
the second
ho
st a
s
a
clie
nt. Aud
i
o an
d vide
o pa
ckets
were tran
sm
itted
simultan
eou
sl
y from the first ho
st to the
se
con
d
ho
st
. We in
clud
e
d
ba
ckgro
u
n
d
traffic
so th
at
approa
che
d
real
con
d
ition
s
in
the fiel
d. Backg
r
ou
nd
traffic wa
s set
up usi
ng UDP
traffic
f
r
om
sou
r
ce to de
stination. In ad
dition, the del
ivery
of the p
a
ckets
we
re l
i
mited to maximum delive
r
y
time of 100 seco
nd
s.
Simulation of
multimedia
delivery
wa
s rep
eated
u
s
ing
seve
ral
different
da
ta rate
s.
Multimedia
p
a
ckets
whi
c
h
were con
s
ist
ed of au
dio a
nd video
we
re se
nt over t
he net
work
with
the lowest b
a
ndwi
d
th allo
cation up to th
e high
es
t d
u
ri
ng the me
asu
r
eme
n
ts. Parameters of e
nd-
to-end d
e
lay and pa
cket loss we
re al
so config
ure
d
according to
netwo
rk
cha
r
acteri
stics in
the
field.
2.3.2. Chara
c
teri
zing Multimedia QoS
In cha
r
a
c
teri
zing multime
d
i
a
QoS, we m
easure
d
an
d
analyzed the
netwo
rk pa
ra
meters
usin
g obj
ecti
ve testing. A
fter mea
s
u
r
in
g and
an
alyzing th
e mul
t
imedia p
a
cket delivery,
we
investigated the characteri
stics of multimedia QoS usin
g PESQ [15] and PSNR. PESQ
was
use
d
to estim
a
te the audio
quality and PSNR was u
s
ed to estim
a
te the video quality. Then we
applie
d use
r
'
s
perce
ption expre
s
sed in Mean Opi
n
io
n Score (MO
S
) to charact
e
rize multime
d
ia
QoS. The
correlation
between PSNR and MOS
c
an
be seen in T
able 1
while
PESQ using
MOS
scale. MOS v
a
lue
wa
s calculated u
s
in
g the E-m
odel
f
o
rmul
ation [1
6] . The first
step in
E-mo
del
formulatio
n is calculating th
e rating facto
r
.
9
3
.
2
(4)
0
.
024
0
.11
177
.3
177
.3
(5)
30ln
1
15
(
6
)
The equ
ation
for
aims to describ
e disturban
ce fa
ct
ors o
n
the n
e
twork that affect
multimedia Q
o
S. Rating fa
ctor
is cal
c
ul
ated u
s
ing e
q
uation (4). Th
e variabl
e of
is
a
fa
c
t
or
of
quality deg
ra
dation
cau
s
e
d
by delay. T
he varia
b
le
of
is a fa
cto
r
of
quality deg
ra
dation
cau
s
e
d
by com
p
ressi
on te
chniq
u
e
and
pa
cket l
o
ss
.
The
con
s
tants in
equ
ation (4),
(5)
and
(6
) a
r
e t
he
recomme
nde
d numb
e
rs b
y
[16] which
are n
o
t ch
an
ged be
ca
use
of the netwo
rk
con
d
ition
s
. To
cal
c
ulate MO
S (ITU-T P.800) ba
sed on
the estimated
-value, there are provi
s
io
ns pre
s
e
n
ted in
equatio
n (7
).
ITU-T
re
co
mmend
ation
G.107 [16]
has
esta
bli
s
he
d
a cla
s
sificatio
n
of use
r
satisfa
c
tion b
a
se
d
on ratin
g
facto
r
of the E-mod
e
l formula an
d the
estimate
d M
O
S as
sh
own
in Table 2.
1
1
0,035
R
R
R6
0
100
R
71
0
4,5
0
0
1
0
0
100
(7)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 4, Dece
mb
er 201
6 : 1534 – 154
4
1538
Table 1. Co
rrel
a
tion bet
ween
PSNR an
d MOS
Table 2. Rel
a
tion betwe
en
MOS and
User Satis
f
action
PSNR
MOS
(up
p
er limi
t
)
R-Value
(lo
w
er limit)
MOS
(lo
w
er limit)
User satis
fac
t
io
n
> 37
4,5
90 4,34
Ver
y
satisfied
32
−
37
3,5
80 4,03
Satisfied
26 – 31
2,5
70
3,60
Some users dissatisfied
20 – 25
1,5
60
3,10
Man
y
users dissatisfied
< 20
0,5
50
2,58
Nearl
y
all users
dissatisf
ied
2.4. Validating Important I
ssues o
f
th
e Metho
dolog
y
I
n
t
h
is se
ct
io
n we c
o
n
s
ide
r
t
o
dis
c
u
ss t
he
validation
of the metho
dology propo
sed in thi
s
r
e
sear
ch w
o
rk
.
1.
Doe
s
ou
r me
asu
r
em
ent already refle
c
t
the ch
ara
c
te
ri
stics of rural netwo
rks?
We
pe
rform
measurement
thro
ugh
VPN (Virtu
al
Priva
t
e Net
w
o
r
k)
chann
el to
en
sure
that
sent
probe
p
a
cket i
s
an
swered
by
co
rre
s
po
ndin
g
d
e
stination h
o
st,
not by
any
ro
uter i
n
mi
ddle
of
the net
work li
nk. Me
asure
m
ent ho
st lo
cated in
the
te
stbed
net
work i
s
config
ure
d
a
s
a
VPN
cl
ient
and a VPN server i
s
located in a high
-spe
ed a
c
ad
e
m
ic net
work. In this mea
s
u
r
eme
n
t, we u
s
e
the same IP
add
re
ss seg
m
ents
betwe
en me
asure
m
ent ho
sts a
s
if
without
any intermed
iate
route
r
. Each
measurement
is mo
nitore
d
ca
refully
so that ou
r mea
s
urem
ent can
be ju
stified. T
o
con
d
u
c
t mea
s
ureme
n
t in real net
wo
rk
condition,
we
desi
gne
d me
asu
r
em
ent scenari
o
a
s
d
e
tai
l
as
poss
ible.
The mea
s
u
r
e
m
ents a
r
e co
ndu
cted repe
atedly to
ensure that the
measurement
results
will have eno
ugh informati
on for ch
ara
c
terizin
g
t
he n
e
twork. In the meas
urem
ent scena
rio
that
we have d
e
s
ign
ed, ban
d
w
idth me
asu
r
eme
n
ts a
r
e
performed
200 time
s, end-to
-en
d
de
lay
measurement
are perfo
rm
ed 2000 time
s, and pa
ck
e
t
loss mea
s
u
r
eme
n
ts are perfo
rmed 2
0
0
time
s
.
H
o
w
e
ve
r
,
w
e
limit ea
c
h
me
as
ur
eme
n
t
o
n
l
y +
10 second
s to ke
ep the in
terest
s of oth
e
r
use
r
s.
We
se
t the inte
rval
time b
e
twe
e
n
me
asure
m
ents
as a
n
a
v
erage
of
30
minute
s
. In
the
measurement
s of b
and
widt
h capa
city, we blo
c
ked
oth
e
r u
s
e
r
s from
usi
ng the
sa
me net
wo
rk
so
that the maximum ban
dwi
d
th cap
a
city can be mea
s
u
r
ed.
We ve
rify th
e a
c
cura
cy
o
f
the net
wo
rk me
as
ur
e
m
en
ts
as
fo
llo
w
s
.
F
i
rs
t,
w
e
c
o
mpa
r
e
band
width
ca
pacity from
measurement
with band
wit
dh ca
pa
city advertise
d by Internet Servi
c
e
Provider
(ISP). The re
sult
s sh
owed that the
bandwidth cap
a
city
from the measure
m
ent
wa
s
simila
r with
the ban
dwi
d
th ca
pa
city adverti
sed
by the ISP. Secon
d
, we com
p
a
r
e
the
measurement
result
s on th
e first locatio
n
of te
stbed netwo
rk
with the se
con
d
location. It sho
w
n
simila
rity in the mea
s
u
r
e
m
ent re
sult
s, ther
efo
r
e it
can
be
con
c
l
uded th
at the mea
s
u
r
em
ent
s
have bee
n ab
le to capture the network chara
c
te
risti
c
s.
2.
Why do we chara
c
te
rize multimedia Qo
S through a
simulation?
There a
r
e two re
asons wh
y we cha
r
a
c
terize multime
d
ia QoS th
ro
ugh
simulatio
n
. First
,
netwo
rk capa
city is limite
d
, parti
cula
rly i
n
ru
ral
area li
ke i
n
Indo
ne
sia. Testin
g
wi
th audio
-
vide
o
strea
m
ing
directly on the
netwo
rk
re
qu
ires th
e ma
j
o
rity of netwo
rk re
so
urce
al
locatio
n
an
d
it
usu
a
lly take
s
a long
time.
While
ou
r sce
nario
re
qui
re
s
strea
m
ing
whi
c
h i
s
repe
ated ma
ny times,
multimedia QoS characteri
zation
performed di
rectly
on the net
work will
overl
o
ad the net
work
itself. On the other ha
nd, the user ha
s h
i
gh
inten
s
ity to use the n
e
twork resources.
Secon
d
ly, ou
r metho
dolo
g
y
requi
re
s two ho
st
s fo
r n
e
twork m
e
a
s
urem
ent at th
e sa
me
time. Howev
e
r, we a
r
e no
t able to perf
o
rm an
d co
ntrol the expe
ri
ments in two
different pla
c
e
s
simultan
eou
sl
y. With simu
lation, efforts fo
r ch
aract
e
rizating m
u
ltimedia QoS
become mo
re
flexible. We can perfo
rm m
easure
m
ent
s repe
atedl
y wi
thout burd
eni
ng the network in the field.
3. Case Stud
y
of Net
w
o
r
k
Measurem
e
nts
This stu
d
y uses testb
ed ne
twork in Cia
n
j
u
r
Re
gen
cy, We
st Java Province of Ind
one
sia.
The n
e
two
r
k con
s
i
s
ts of
real
-time mul
t
imedia
deliv
ery sy
stem t
hat lies
on a
netwo
rk of two
edu
cation
al e
n
tities ba
se
d
on Asymm
e
tric Di
gital
Sub
s
cribe
r
Li
ne (ADSL)
acce
ss net
wo
rk. In
its
architectu
re, we define two Local Area Networks
(LA
N
s) in Cianj
ur and one Ca
mpus
Network i
n
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
A Methodolo
g
y for
Cha
r
a
c
terizin
g
Re
al-Tim
e Multim
e
d
ia Qualit
y of… (Yoan
es B
andu
ng)
1539
Bandun
g: i)
LAN A i
s
a
l
o
cal
network of edu
ca
tio
nal entity in
locatio
n
A
where
some
e
nd-
system
s con
s
ist of data en
d-sy
stem, au
dio end
-sy
s
te
m, and video
end sy
stem
s are d
eployed,
ii)
LAN B is a l
o
cal
netwo
rk
of edu
cationa
l entity in location B with t
he same e
n
d
-
sy
stems
as i
n
LAN A,
and iii)
Campus network
where
a multim
edia proxy
server
is inst
alled to facilitate
multimedia d
e
livery betwe
en two LANs.
The two edu
cation
al entities may be lo
cated in seve
ral
km
until som
e
ten
s
km fa
r a
w
ay from
each ot
h
e
r.
On the
othe
r han
d, the
di
stan
ce
betwe
en
camp
us n
e
twork in Ba
ndu
ng and the e
d
u
catio
nal enti
t
ies in Cia
n
jur regen
cy is a
r
ound 6
0
km.
In many ca
ses, broad
ban
d Internet
co
nne
ction
s
in
edu
cation
al e
n
tities in ma
ny rural
area
s in Indo
nesi
a
are p
r
e
dominate
d
by ADSL acce
ss network as
their last mile
conn
ection.
As
sho
w
n in
Fig
u
re 1, two ed
ucatio
nal enti
t
ies ar
e conn
ected to
a telepho
ne comp
any throu
gh fi
xed
line tele
phon
e cable.
The
data
sent f
r
o
m
a
Digita
l
Subscri
b
e
r
Li
n
e
Acce
ss Mul
t
iplexer
(DSL
AM)
is splitted
by a
splitter and
sent
to
DSL modem
u
s
ing
a dedi
cate
d point-to
-
poi
nt con
n
e
c
tion.
T
h
e
data
conn
ecti
on to the
glo
b
a
l net
work i
s
establi
s
h
ed u
s
ing
dial
-up
conne
ction
wit
h
Point-to
-Poi
nt
Protocol (PP
P
). In several
edu
cation
al
entities i
n
Ci
anjur,
ban
dwi
d
th capa
citie
s
of th
e la
st
mile
con
n
e
c
tion
s are typically rangin
g
from 5
12 kb
ps u
n
til 2 Mbps.
Figure 1. A topology of test
bed net
work
3.1. Measure
ment Scena
r
io
Measurement
scena
rio i
s
d
e
sig
ned to
de
termi
ne ch
aracteri
stics
of
the testbe
d n
e
twork.
In the scen
ari
o
, several net
work p
a
ramet
e
rs i
n
cl
udin
g
band
width, e
nd-to
-en
d
del
ay, and pa
cket
loss are m
e
a
s
ured a
s
de
scrib
ed bel
ow:
1. Bandwi
d
th
In
gen
eral,
I
n
ternet ban
d
w
idth co
nsi
s
ts
of
d
o
wnstream ban
d
w
idth and u
p
stre
am
band
width. In
our expe
rim
ents, the do
wn
st
rea
m
ba
ndwi
d
th wa
s
measured by
flooding the
link
band
widt
h u
s
ing Ip
erf
appli
c
ation
.
Ea
c
h
me
as
ur
eme
n
t
wa
s ru
n fo
r
up
to
1
0
s
e
c
o
nd
s w
i
th
interval between mea
s
u
r
e
m
ents was a
pproxim
ately 30 minutes.
On the other han
d, the
measurement
of the up
stream ba
nd
wid
t
h wa
s do
ne
by sen
d
ing fil
e
s of a
ce
rtai
n si
ze from
client to se
rver.
2. End-to-end
Delay
End-to-end
d
e
lay wa
s me
asu
r
ed
usin
g
ICMP
ech
o
requ
est a
nd ICMP echo re
spo
n
se.
Source ho
st (clie
nt) sent IC
MP echo re
que
st (1,000
packet
s
) an
d
destinatio
n host (se
r
ver)
respon
ded
wi
th ICMP ech
o
respon
se p
a
ckets of
the
same si
ze.
The pa
ckets
have differen
t
sizes rangin
g
from
32
bytes
up
to 1
4
7
2
bytes.
The
amount
of 3
2
byte refers t
o
the
default
value of PING comm
and,
while the si
ze of 147
2 bytes refe
rs
to MTU valu
e of ADSL
netwo
rks. We
repeate
d
the
experime
n
t 10 times with 3
0
minutes tim
e
intervals.
3.
Packet Loss
Rate (PL
R
)
PLR p
a
ra
met
e
r
wa
s me
asured
u
s
ing v
a
riou
s
si
ze
s
of UDP
data
g
ram
s
with 1
470 byte
s
per
datag
ram
on Ip
erf a
p
p
licatio
n. We
deliver
ed
UDP data
g
ram
s
fro
m
5
64
Kbps
until 1
0
Mbps
. The s
i
z
e
of 564 Kbps
was
c
a
lc
ulated fr
om the data rate of audio-video pack
et when
applying Vo
rb
is and VP8 compressio
n schem
es.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 4, Dece
mb
er 201
6 : 1534 – 154
4
1540
3.2. Results of Ne
t
w
o
r
k
Measur
e
men
t
This sectio
n
presents th
e me
asure
m
ent re
sults i
n
a
c
cordan
ce with
the
scena
rio
pre
s
ente
d
in the pre
c
edi
n
g
se
ction. Fu
rtherm
o
re
, we investigate
each parame
t
er to obtain the
cha
r
a
c
teri
stics of the rural netwo
rk.
1. Bandwi
d
th
It was o
b
se
rved that b
and
width all
o
cati
o
n
of the testb
ed
netwo
rk
du
ring th
e
measurement
process
ha
s
fluctuated.
From ten
time
s
of mea
s
u
r
em
ent, it was kn
own
that th
e
band
width ra
nged from 0.
13 Mbp
s
up to 12.5 Mbp
s
with avera
ge 8.20 Mbp
s
in the first
locatio
n
(LAN A). In
the
se
con
d
lo
catio
n
(LA
N
B
)
, the
ban
dwi
d
th
ra
nged
from
0.
546
Mbp
s
t
o
12.1 M
b
p
s
wi
th average
7.
83 M
b
p
s
. We
found
t
hat th
e lo
we
st ba
n
d
width
on
LA
N A
wa
s
0.13
Mbps o
r
1.3
%
from sub
s
criptio
n
ban
d
w
idth up to
10 Mbp
s
. In some m
e
a
s
u
r
eme
n
ts, the
maximum ba
ndwi
d
th allo
cated exce
ed
ed the ca
pa
city subscri
b
e
d
band
width
of up to 12.5
Mbps o
r
ap
proximately 25% from sub
scribed b
and
wi
dth.
Figure 2. Bandwidth all
o
cat
i
on of testbed
network
Up
strea
m
ba
ndwi
d
th ha
s fluctuated in
line
with the
band
width all
o
catio
n
altho
ugh th
e
ratio
wa
s
no
t alway
s
di
rectly p
r
op
orti
onal
. T
he
ra
tio between
upstream
an
d ba
nd
width
allocation in
the two
LANs wa
s
rangi
ng
from 4
to 10
. In ord
e
r to
cha
r
a
c
teri
ze
downstrea
m
,
we redu
ce t
he ban
dwi
d
th allocation
by the
upst
r
eam ba
nd
wi
dth. We al
so
comp
ared the
downstream
and up
strea
m
. Figure 3
sho
w
s th
at down
s
tream
was
always large
r
tha
n
upstream
with ratio of 3 to 9.
We fou
nd tha
t
decrea
s
e
of band
width
do
not di
rectly d
e
crea
se u
p
st
ream ba
nd
wid
t
h with
the same
rati
o. When the
band
width all
o
catio
n
is de
cre
a
sed, we f
ound the ratio of upstre
a
m
to do
wn
stre
a
m
sometime
s in
crea
se
d. The
results of me
asure
m
ents on
th
e first lo
catio
n
sho
w
e
d
whe
n
the band
width allo
cat
i
on wa
s de
cre
a
sed to 4.71 Mbp
s
, upstre
a
m to
d
o
w
n
s
tre
a
m
r
a
tio
w
a
s
inc
r
ea
se
d
to
1
:
3
(
1
.18
Mb
p
s
)
.
O
n
the
s
e
c
o
n
d
loc
a
tio
n
,
w
hen
band
width
all
o
catio
n
wa
s
only 3.95
Mb
ps, the
ratio
of up
stre
am t
o
do
wn
stre
a
m
was 1 :
3
(0,97 M
b
p
s
).
To a
s
sure t
h
is
result, we
tried o
n
oth
e
r net
wo
rk
a
nd obtai
ned
simila
r re
sult
s
with testbe
d netwo
rk.
2. End-to-end
d
e
lay
We
cal
c
ulate
d
end
-to-end
delay u
s
ing
equatio
n (1
)
and ig
nored
no re
sp
on
se
packet
s
.
End-to-end d
e
lay on the first and
se
con
d
locatio
n
wa
s ra
nging fro
m
7 ms to 55
6 ms an
d fro
m
8 m
s
to
19
18
ms,
re
sp
ecti
vely. RTT
wa
s hi
gh
only i
n
a fe
w
mome
nts, an
d thi
s
woul
d
cau
s
e
interferen
ce
on multime
d
i
a
streamin
g. The
ave
r
a
g
e
of en
d-to
-e
nd del
ay on
the first a
n
d
se
con
d
location wa
s 25.6
8
ms and 88.7
9
ms.
3. Packet
Loss
Rate
PLR
wa
s cal
c
ulate
d
u
s
ing
equatio
n (2).
From
th
e m
easure
m
ent
s we
obtaine
d
that the
PLR in LAN
A and LAN B were 5.8% an
d 8.6%, resp
ectively.
100%
5
.8%
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
A Methodolo
g
y for
Cha
r
a
c
terizin
g
Re
al-Tim
e Multim
e
d
ia Qualit
y of… (Yoan
es B
andu
ng)
1541
100%
8
.
6
%
Figure 3. Do
wn
strea
m
to Up
strea
m
Ra
tio
Figure 4. Packet Lo
ss
Rate
of the Testbe
d Network
In this study
we al
so in
vestigate the
oc
curren
ce
of packets loss du
ring
netwo
rk
measurement
s. The exp
e
ri
ments
rev
eal
that packet l
o
ss only o
c
curs
at ce
rtain
times. Figu
re 4
sho
w
s the occuren
c
e of pa
cket loss rate
for
the LAN A and LAN B during the m
easure
m
ent
s. It
wa
s sh
own that the packet loss rate vari
ed from 8% to 90%.
4. Simulation Stud
y
of Characteri
z
i
ng Multimedia QoS
4.1. Validating Important I
ssues o
f
th
e Metho
dolog
y
We buil
d
a n
e
twork
simul
a
tion mod
e
l that rep
r
e
s
ent
s multimedi
a
delivery in th
e testbed
netwo
rk
whi
c
h applie
s
We
bRT
C
techno
logy. From
th
e netwo
rk topology on Fi
gure
1, we d
r
aw
the simul
a
tio
n
model
ba
sed on
OM
NeT++
Netwo
r
k Simulato
r
with INET F
r
ame
w
o
r
k. T
he
simulatioi
n m
odel o
n
Fi
gu
re 5
con
s
ist
s
of seve
ral
comp
one
nts,
i.e. inte
rn
et clo
ud, route
r
s,
acce
ss p
o
ints, and host
s
. In gene
ral, we
arra
nge the
comp
one
nts i
n
to three g
r
o
ups: Lo
catio
n
1
or
LAN A to
repre
s
e
n
t e
d
u
c
ation
a
l e
n
tity 1, Lo
ca
tio
n
2 o
r
LA
N B
to represent
e
ducational
en
tity
2, and intern
et cloud to
re
pre
s
ent the I
n
ternet. In LA
N A dan LA
N B, each ho
st
is co
nne
cted
to
an a
c
cess
p
o
int that works on f
r
equ
e
n
cy of 2.4
G
H
z with
a m
a
ximum data
rate
cha
nnel
54
Mbps
.
T
he a
c
ce
ss
point i
s
con
n
e
c
ted to
a route
r
which act
s
a
s
a g
a
teway fro
m
each LAN to t
h
e
Internet
clou
d
via ethernet l
i
nk
with data
rate of
1
00 M
bps. T
he d
a
ta rate
of con
nectio
n
bet
ween
both route
r
s to the internet
clou
d is set to the
values th
at appro
a
ch real co
ndition i
n
the field.
Data rate b
e
t
ween route
r
and intern
et cl
oud is
co
nfigure
d
ba
sed on the u
p
stre
am
band
width
ca
pacity in e
a
ch location. Simulation
st
ud
y is built u
s
in
g symmet
r
ic
band
width,
while
testbed
network ha
s
asy
mmetric ba
n
d
width
whi
c
h
is in
ge
nera
l
, the up
stre
am ba
nd
widt
h is
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 4, Dece
mb
er 201
6 : 1534 – 154
4
1542
alway
s
small
e
r than the downs
tream
band
width. We dete
r
min
e
the data rate used in
the
s
i
mu
la
tion
wa
s
tw
ic
e
o
f
u
p
s
tre
a
m
b
a
n
d
w
id
th
. T
h
e in
te
r
n
e
t
c
l
o
u
d
is
ass
u
med
a
s
h
i
g
h
s
peed
netwo
rk a
nd i
s
co
nfigured to 100 Mbp
s
.
Figure 5. Net
w
ork Simulati
on Model
We u
s
e mea
s
ureme
n
t results on ea
ch
netwo
rk
te
stb
ed to configu
r
e end
-to-en
d delay
para
m
eter on
the
simulatio
n
. From
the
measurem
ent
re
sults,
end
-t
o-en
d d
e
lay p
a
ram
e
ter i
s
set
to 25.68 m
s
at LAN A a
n
d
88.79
ms
a
t
LAN B.
Del
a
y paramete
r
in internet cl
oud i
s
a
s
sum
ed
and
config
ure
d
to 5 m
s
. In
addition, fo
r
config
ur
in
g Bi
t Erro
r Rate
(BER) p
a
ra
me
ter on
the lin
k
betwe
en ro
uter and inte
rn
et cloud, it is assu
me
d that the value for the para
m
ete
r
is 10
-5
.
4.2. Multimedia Content
Deliv
er
y
Scenario
In this re
sea
r
ch,
simulatio
n
of multimedia
delivery that con
s
i
s
ts
of audio, video, an
d
backg
rou
nd t
r
affic i
s
confi
gure
d
u
s
ing
o
ne-way
com
m
unication. T
he u
s
e of
ba
ckgroun
d traffi
c is
aimed to ap
p
r
oximate the
real n
e
two
r
k
con
d
iti
on. Th
e simul
a
tion
model i
s
run
on a laptop f
o
r
100 second
s.
The delivery
of audio, vide
o, and
ba
ckground traffic
was configu
r
ed
as follows.
1. Audio
To simulate
audio pa
cket
s delivery, we used
sim
p
leVoIPSende
r
applicatio
n that wa
s
installe
d in
sende
r
ho
st. In the
re
ceiv
er
ho
st, it was i
n
stall
ed
sim
p
leV
o
I
P
R
e
cei
v
e
r
. In t
h
is
simulatio
n
, audio pa
ckets wa
s
tran
smitted usin
g
talk
P
a
ck
etSiz
e
=
160
b
y
te
s
an
d
packeti
zationI
nterval=
0.02
se
con
d
s. T
h
i
s
configu
r
atio
n wa
s ta
ke
n
to app
roa
c
h
audio
delive
r
y
usin
g Vorbi
s
with data rate
of 64 Kbps.
2. Video
We used
UDPVideoStrea
m
S
vr
that
wa
s in
stalle
d in
se
nde
r
ho
st. In the
receiver
ho
st,
we installed
UDPVide
oS
tream
Cli
. In this sim
u
latio
n
, video pa
ckets
wa
s tra
n
smitted u
s
i
ng
packetL
en
=1
024 byte
s an
d
se
ndInterval
=0.0
16 se
cond
s.
This confi
guration
wa
s u
s
ed in
orde
r
to appro
a
ch video delive
r
y usin
g VP8 with data rate of
500 Kbps.
3. Backgroun
d
Traffic
We used
UD
P
B
a
sicA
p
p
t
hat was i
n
st
alled in
send
er h
o
st.
Whil
e in the
rece
iver ho
st
UDPSink
was in
stalled.
In this simulation, video pa
cket
s wa
s tra
n
s
mitted u
s
in
g
messag
eLe
n
g
th=5
12 byt
e
s a
nd
se
ndInterval
=0.
032
se
con
d
s
. Backg
r
ou
nd traffic
wa
s
transmitted d
u
ring a
udio a
nd video deliv
ery pro
c
e
s
s.
4.3. Chara
c
teriza
tion of
Multimedia QoS
In this se
ctio
n we d
e
scrib
e cha
r
a
c
teri
zation of multi
m
edia Q
o
S both audi
o a
nd video
delivery. We i
n
vestigate e
n
d
-to-end d
e
la
y and packe
t
loss to obtai
n these
cha
r
acteri
stics. End-
to-end
delay
of audio
wa
s varie
d
ap
p
r
oximatel
y 0.
14 second
s to 1.2 se
co
n
d
s. Wherea
s its
averag
e
wa
s
0.16
se
con
d
s. The
high
est
delay
wa
s o
c
curred
in th
e first second
, whi
c
h
wa
s
1.2
se
con
d
s a
n
d
next second
was va
ried
betwe
en 0.1
4
se
con
d
s to
0.19 se
con
d
s
. This
con
d
i
t
ion
wa
s simila
r a
s
the network characte
ri
stics
whe
r
e
in first second, b
and
width allo
cation ten
ded
to
be
small. Accordin
g to IT
U-T G.1
14, e
n
d
-to-end
dela
y
of audio
tra
n
smi
ssi
on
ha
s b
ad u
n
til go
od
quality. Howe
ver, overall e
nd-to
-en
d
del
ay can be a
c
cepte
d
with g
ood qu
ality.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
A Methodolo
g
y for
Cha
r
a
c
terizin
g
Re
al-Tim
e Multim
e
d
ia Qualit
y of… (Yoan
es B
andu
ng)
1543
Audio pa
ckets that was d
e
livered for 1
00
se
con
d
s
experie
nced
42 pa
cket
s
loss wit
h
varied
rate
b
e
twee
n 0%
to 25%. If the
pa
cket lo
ss
wa
s o
b
serve
d
eve
r
y seco
nd, the
high
e
s
t
packet lo
ss
(25%) o
c
curre
d
whe
n
simul
a
tion run
n
ing
until 2 secon
d
s, wh
ere
a
s t
he next packet
loss rate
wa
s varie
d
bet
wee
n
0% to 5%. PLR wa
s co
nfigu
r
ed
at 1.3%. Accordin
g to ITU-T
G.1010
[17] ,
PLR for
audi
o conversatio
n
s
sh
ould
be
less tha
n
3%.
Overall, audi
o qu
ality in th
is
simulation can be well accepted be
cause the value of PLR is still
less than 3%, whi
c
h is 1.3%
.
Figure 6. End-to-e
nd del
ay of audio
Figure 7. Packet loss rate of audio
End-to-end
d
e
lay of audi
o
wa
s varie
d
appr
oximatel
y 0.121 seco
nds to
0.172
se
con
d
s
with average
value of 0.1
35 se
co
nd
s. Bas
ed on
IT
U-T G.
114 recom
m
en
dati
on,
end
-to-e
n
d
delay video d
e
livery that can be
well a
c
cepte
d
is
le
ss than 0.4
se
con
d
s. In the
simulatio
n
, e
nd-
to-end
delay
of video delivery wa
s 0.13
5 se
con
d
s.
T
h
is sho
w
ed t
hat the video
can b
e
a
cce
p
t
ed
with goo
d qu
ality.
Figure 8 End-to-end d
e
lay of video
Figure 9. Packet loss rate of video
Within
100
seco
nd
s a
m
ou
nt of 6,2
50,0
00 by
te
s vid
e
o
pa
ckets
we
re
delivered f
r
om th
e
sou
r
ce and
6
,
088,704 byt
e
s data
we
re
received at
the de
stinatio
n. From this
experim
ent, we
cal
c
ulate
d
th
e PLR at ap
proximately 2
.
6% t
hat am
ount of 1
61,
296 byte
s p
a
c
kets
we
re l
o
st.
Based
on o
u
r observation,
it was
sh
own that a numb
e
r
of pa
cket lo
ss
rate
we
re
not the same
at
every time, b
u
t it only o
c
curred fo
r
cert
ain second.
Acco
rdi
ng to
ITU-T
G. 10
1
0
, PLR fo
r vi
deo
delivery sh
ou
ld be less th
an 1%. Whe
n
the pac
ket
loss wa
s ob
served at every seco
nd, video
delivery at ce
rtain times
was suffered lo
ss that can n
o
t be tolerate
d.
Furthe
rmo
r
e
we
cal
c
ulate
d
R-val
u
e
s
ba
sed
on
equ
ation (4).
R-valu
e of au
dio a
n
d
vide
o
delivery obtai
ned are 84.1
1
and 80.0
8
. MOS value wa
s calculated
using e
quati
on (7
). MOS of
audio
delivery was 4.17
a
nd MOS
of video d
e
livery
wa
s 4.03. Ba
sed
on T
able
2, the qu
ality of
multimedia d
e
livery over t
e
stbe
d netwo
rk a
c
hi
eved u
s
er
sati
sfacti
on level which wa
s satisfie
d. It
can
be
co
ncl
uded
that the
testbe
d net
work was fea
s
i
b
le for re
al-ti
m
e multime
d
i
a
delive
r
y su
ch
as di
stan
ce le
arnin
g
activities thro
ugh virtual cla
s
sroo
m technol
ogy
.
Evaluation Warning : The document was created with Spire.PDF for Python.