Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
9
, No
.
2
,
Febr
ua
ry
201
8
,
pp.
37
3
~
379
IS
S
N:
25
02
-
4752
, DO
I: 10
.11
591/
ijeecs
.
v9.i
2
.
pp
373
-
379
373
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
Perform
ance An
alys
i
s of Up
lin
k Schedulin
g A
l
gorith
ms in LTE
Network
s
Shafinaz
Bt
Is
mail
1
, Darm
aw
at
y B
t M
oh
d
A
li
2
, Nors
uz
ila
Ya’acob
3
Facul
t
y
of Electr
ic
a
l
Eng
ineeri
ng
,
Univer
si
ti
Te
kn
ologi
MA
RA,
M
al
a
y
si
a
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Oct
16
, 201
7
Re
vised
Dec
20
, 2
01
7
Accepte
d
Ja
n 2
, 201
8
Schedul
ing
is
r
efe
rring
to
the
proc
ess
of
al
l
oca
t
ing
resourc
es
to
Us
er
Equi
pm
ent
base
d
on
sche
duli
ng
al
gorit
hm
s
tha
t
is
loc
at
ed
a
t
the
LT
E
bas
e
stat
ion
.
Vari
ous
al
gor
it
hm
s
ha
ve
be
en
propo
sed
as
th
e
ex
ec
ut
ion
of
sche
duli
ng
al
go
rit
hm
,
which
r
epr
ese
nts
an
o
pen
issue
in
Long
Te
rm
Evol
uti
on
(LTE
)
standa
rd
.
This
pape
r
m
ake
s
an
attempt
to
stud
y
and
compare
the
p
e
rform
anc
e
of
th
ree
wel
l
-
known
upli
nk
sche
dul
ers
namel
y
,
Maximum
Thro
ughput
(MT),
First
Maximum
E
xpansion
(FM
E),
and
Round
Robin
(RR).
Th
e
eva
lu
at
ion
is
conside
red
for
a
single
cell
with
int
erf
ere
n
ce
for
thre
e
flows
such
as
Best
eff
ort,
Video
and
VoIP
in
a
pede
stria
n
envi
ron
m
ent
us
ing
the
L
TE
-
S
IM
net
work
si
m
ula
tor.
The
p
erf
orm
anc
e
eva
lu
at
ion
is
co
nduct
ed
in
t
erms
of
sy
st
em
throughput
,
fai
rn
ess
inde
x,
de
l
a
y
and
pac
k
et
loss
rat
io
(PLR)
.
Th
e
sim
ula
ti
ons
res
ult
s
show
tha
t
R
R
al
gori
thm
al
wa
y
s
re
ac
h
es
t
he
lowest
PLR,
del
iv
eri
ng
h
ig
he
st
throughput
fo
r
vide
o
and
VoIP
flows
am
ong
al
l
those
s
tra
t
egi
es.
Thus,
RR
is
the
m
o
st
suita
bl
e
sche
duli
ng
a
lgo
rit
hm
for
VoIP
and
vide
o
flo
ws
while
MT
and
FM
E
is
appr
opriate
for BE f
lows i
n
L
T
E
ne
tworks.
Ke
yw
or
d
s
:
LTE
Sche
du
li
ng al
gorithm
s
Up
li
nk
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
:
Sh
a
finaz Bt
Is
m
ai
l
,
Faculty
of Elec
tric
al
Engineer
ing
U
niv
e
rsiti
Tekno
l
og
i M
A
RA,
40450 S
hah A
l
a
m
, S
el
ango
r,
Ma
la
ysi
a.
Em
a
il
:
sh
afina
z.ism
a
il
@g
m
a
i
l.com
1.
INTROD
U
CTION
The
3rd
Ge
ne
r
at
ion
Partne
rs
hip
Pro
j
ect
(
3GPP
)
has
im
pl
e
m
ented
the
st
and
a
r
dizat
ion
of
the
Lo
ng
Term
Evo
luti
on
(LT
E),
wh
e
r
e
the
syst
e
m
is
reco
m
m
end
ed
to
deliver
pea
k
data
rates
of
50
Mbit
/s
in
up
li
nk
and
100
Mbit
/
s
in
dow
nlink
with
a
20
M
Hz
sp
ect
ru
m
pro
vision
(1).
Sing
le
car
rier
fr
e
qu
e
ncy
di
visio
n
m
ul
ti
ple
acce
ss
(S
C
-
F
DMA
)
has
bee
n
c
ho
sen
as
the
upli
nk
acce
ss
sc
hem
e
wh
il
e,
ort
hogonal
fr
e
quency
div
isi
on
m
ultip
le
acce
ss
(
OFDMA
)
is
s
el
ect
ed
as
t
he
dow
nlink
acce
ss
s
chem
e
fo
r
LT
E
(
2)(3).
SC
-
F
DM
A
offer
s
lo
wer
pe
ak
-
to
-
a
ver
a
ge
power
rati
o
(
PA
PR
)
as
com
par
e
d
to
OFD
MA,
thus,
m
a
king
SC
-
F
DM
A
m
or
e
su
it
able
f
or
upli
nk
tra
ns
m
is
sion,
as
t
he
User
Eq
uip
m
ent
(
UE)
ha
s
t
he
a
dv
a
ntage
of
t
ran
sm
it
te
d
powe
r
eff
ic
ie
ncy
al
on
g
with
i
ncr
eas
ed
dat
a
rates
,
in
orde
r
to
im
pr
ove
the
batte
r
y
li
fe
of
the
U
E
(4)(5).
SC
-
F
DMA
syst
e
m
ensu
res
to
deliver
hi
gher
th
rou
ghput,
lowe
r
PAPR
,
higher
s
pectral
eff
ic
ie
ncy,
an
d
lowe
r
bit
error
rat
e
than
the
co
nve
ntion
al
O
FD
M
A
te
chn
i
qu
e
(
6).
In
s
pite
of
the
adv
a
ntage
s
of
S
C
-
F
DMA
,
it
req
uires
th
at
all
su
bc
ar
riers all
oc
at
ed
to
a sin
gl
e U
E
m
us
t be a
dj
ace
nt t
o
eac
h othe
r o
n
the
fr
equ
e
ncy
do
m
ai
n (4).
The
Fr
e
quency
Divisi
on
D
uple
xing
(FDD
)
a
nd
Tim
e
Division
D
up
le
xing
(TDD)
are
tw
o
duple
xi
ng
schem
es
us
ed
in
the
LTE
upli
nk
tra
ns
m
issi
on
.
I
n
F
DD,
di
f
fer
e
nt
fr
e
quen
cy
bands
are
ut
il
iz
ed
fo
r
the
up
li
nk
transm
issi
on
,
wh
il
e
in
TD
D
t
he
up
li
nk
s
ha
r
e
the
sam
e
fr
eq
u
ency
ba
nd
bu
t
are
se
par
at
e
d
in
tim
e.
The
ne
twork
arch
it
ect
ure
f
or
LTE
co
ns
ist
s
of
E
vo
l
ved
N
od
e
B
(e
N
od
e
B),
Ev
olv
e
d
Pa
cket
Syst
e
m
(EPS
)
a
nd
the
U
Es.
The
LTE
transm
issi
on
is
div
i
ded
into
fr
am
es,
w
hich
co
ns
ist
of
10
s
ubfr
am
es.
A
subf
ram
e
d
ur
at
io
n
is
1
m
s
in
l
eng
th
.
Eac
h
s
ubfr
am
e
is
div
i
ded
into
t
wo
slots
w
he
re
each
slot
is
0.5
m
s
in
le
ngth
.
A
sub
fr
am
e
is
al
so
know
n
as
the
Tra
ns
m
issi
on
Tim
e
In
te
rv
al
(TTI).
The
physi
cal
la
ye
r
inter
face
is
a
tra
ns
po
rt
blo
c
k
with
c
om
m
on
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vol
.
9
,
No.
2
,
Fe
br
uary
201
8
:
37
3
–
37
9
374
Modula
ti
on
an
d
C
od
i
ng
Sc
he
m
e
(MCS).
Ea
ch
TT
I
c
onta
in
s
at
m
os
t
on
e
trans
port
blo
c
k
per
UE
(
7).
Ea
ch
sl
ot
pr
ese
nt
in
f
requen
cy
do
m
ai
n
is
div
ide
d
int
o
a
num
ber
of
resour
ce
blo
c
ks
.
T
he
str
uctur
e
of
a
ti
m
e
slot
in
fr
e
qu
e
ncy
do
m
ai
n
is
div
ide
d
i
nto
re
gions
of
180
kHz
that
c
on
ta
in
a
c
onti
guous
set
of
12
su
b
car
riers.
He
nce
the
total
nu
m
ber
of
upli
nk
physi
cal
resource
blo
c
ks
(P
R
Bs)
ra
ng
es
bet
ween
6
PRB
s
for
the
sm
al
lest
(1
.
4
MHz) a
nd
100 PRB
s for
t
he
la
rg
est
ba
ndwidt
h (20MHz
) res
pecti
vely
.
In
OFDM
-
bas
ed
m
ulti
-
us
er
fr
am
ewo
r
k,
sche
du
le
r
s
pla
ys
an
i
m
po
rta
nt
ro
le
in
optim
izing
the
netw
ork
pe
rform
ance
and
pr
ov
i
de
Q
ualit
y
of
Se
rv
ic
e
(
Q
oS
)
requirem
ents
in
the
Me
diu
m
Access
Con
tr
ol
(MAC)
la
ye
r
.
The
MAC
sch
edu
le
r
in
t
he
e
NodeB
is
dep
l
oyed
i
n
the
up
li
nk
a
nd
dow
nl
ink
tra
ns
m
issio
n,
are
m
ai
nly
resp
ons
ible
f
or
al
lo
cat
ing
re
s
ource
bl
ock
(RBs
)
am
on
g
UE
s
to
sup
port
the
di
ver
s
e
Q
oS
require
m
ents.
The
ta
sk
of
th
e
scheduler
de
pends
on
the
s
pecific
al
gorithm
us
ed
and
t
he
Cha
nn
el
Q
ualit
y
In
dicat
or
(CQ
I),
wh
ic
h
pro
vid
e
s
fee
db
ac
k
fro
m
UEs
on
w
he
ther
t
he
c
hannel
co
ndit
ion
is
go
od
or
poo
r,
a
nd
al
locat
e
RB
s
accor
dingly
(
8)
(
9).
T
he
co
nt
igu
it
y
co
ns
trai
nt
is
on
e
of
t
he
m
ajo
r
c
onstr
ai
nts
in
t
he
up
li
nk
sc
he
du
li
ng.
This
const
raint
re
fe
r
s
to
hav
i
ng
al
l
PRB
s,
al
locat
e
d
to
a
sin
gle
U
E
to
be
ad
j
ace
nt
al
ong
t
he
f
re
qu
e
ncy
do
m
ain
.
T
he
con
ti
guit
y
const
raint
can
re
duce
the
sp
ect
ra
l
eff
ic
ie
ncy
of
the
up
li
nk
tra
nsm
issi
on
,
since
UE
bein
g
al
lo
cat
ed
a
PRB
desp
it
e
the
existe
nce
of
oth
e
r
UEs
wit
h
bette
r
cha
nnel
qu
al
it
y
ov
er
the
sa
m
e
PRB
(1
0).
Cha
nnel
awar
e
sche
du
li
ng
al
gorithm
s
is
reco
gniz
e
by
de
ve
lop
in
g
the
m
u
lti
-
us
e
r
dive
r
sit
y
gain
by
assigni
ng
t
he
re
so
urces
a
m
on
g
se
ver
al
us
e
rs
de
pendi
ng
on
their
ch
ann
el
co
ndit
ion
s.
I
n
t
his
paper,
t
he
perf
orm
ance
of
t
hr
e
e
well
-
known
upli
nk
schedulers
is
evaluated
na
m
el
y,
Ma
xi
m
u
m
Thr
ou
ghput
(MT),
Fi
rst
Ma
xim
u
m
Exp
an
si
on
(F
ME)
,
a
nd Ro
und
R
obin
(RR).
1.1
M
axim
u
m Thro
ughpu
t
(
MT)
Sche
d
uler
The
MT
is
us
e
d
to
m
axi
m
iz
e
the
ove
rall
thr
oughput
by
c
onti
nu
al
ly
assi
gnin
g
eac
h
RB
t
o
UE
t
hat
is
capab
le
of
m
a
xim
iz
ing
the
overall
thr
ough
pu
t
in
the
c
urr
ent
TTI
inter
va
l.
In
MT
sch
edu
le
r,
U
E
wi
th
the
highest
value
of
CQ
I
will
be
ser
ved
fi
rst
w
it
h
the
re
qu
ire
d
RB
s.
T
hu
s
,
UEs
with
po
or
CQI
values
(s
uch
as
cel
l
-
edg
e
us
e
rs
)
are
no
t
assig
ned
with
s
uffici
ent
resou
rces
.
Su
c
h
U
Es
wi
ll
su
ff
e
r
f
ro
m
low
th
rou
ghpu
t,
an
d
even sta
r
vatio
n m
a
y occu
r
(
11
).
T
he
m
et
ric cal
culat
ion
of
MT i
s expr
e
sse
d
as:
,
=
(
)
(1)
wh
e
re,
m
_(
i,k)^
MT pr
es
ents t
he
m
et
ric o
f
t
he
i
-
th
us
e
r on t
he k
-
t
h
RB
a
nd d_k
^i
(t)
is t
he
expect
ed data
-
rate
for
the
i
-
th
u
se
r
at
ti
m
e t on
th
e k
-
th RB
.
1.2
Fir
st M
axim
um Ex
pa
n
s
ion
(FM
E
)
Sc
heduler
The
FME
sc
he
du
le
r
f
oc
us
es
on
m
axi
m
iz
ing
the
perform
ance
of
th
rou
ghput
an
d
fair
ness
.
Th
e
m
ai
n
pr
i
nciple
in
F
ME
is
to
assi
gn
RB
re
source
s
sta
rting
f
ro
m
the
RB
wit
h
the
highest
m
etr
ic
in
m
at
rix
M,
an
d
exp
a
nd
on
it
in
both
directi
ons
of
the
RB
as
show
n
in
F
igure
1,
as
lo
ng
as
the
c
ha
nnel
m
ai
ntains
its
bes
t
conditi
on
am
on
g
oth
e
r
us
ers
.
As
the
al
gorithm
traver
se
th
rou
gh
eac
h
RB
,
it
check
s
it
s
m
axi
m
u
m
m
e
t
ric
an
d
determ
ines
whet
her
the
m
axi
m
u
m
m
et
ric
s
ti
ll
belongs
to
the
UE
in
w
hich
res
ources
are
cu
r
ren
tl
y
bei
ng
assigne
d,
or
w
hethe
r
the
m
axi
m
u
m
m
et
ric
belongs
to
a
no
t
her
UE.
If
t
he
conditi
ons
are
fu
lfil
le
d,
t
he
R
B
is
assigne
d
t
o
t
he
sel
ect
ed
UE;
oth
e
rw
ise
,
the
UE
is
co
ns
ide
r
ed
se
rved
,
a
nd
the
c
urren
t
sel
ect
ed
RB
is
as
sign
e
d
to
a
new
U
E.
T
he
sche
dule
r
th
en
reit
erates
th
e
exp
a
ns
io
n
pr
ocedu
re.
As
sig
ning
the
RB
to
the
oth
er U
E w
ould
br
ea
k
t
he
c
on
ti
nu
it
y co
ns
trai
nt
s.
Figure
1.
The
UEs
c
ha
nn
el
qual
it
y fo
r
eac
h R
B (12)
1.3
R
ou
n
d
Robin (
RR)
Sche
duler
RR
scheduler
is
chan
nel
unawa
re,
sim
ple
and
easy
to
i
m
ple
m
ent
s
cheduli
ng
sch
e
m
e.
In
this
sche
du
li
ng
st
ra
te
gy,
the
UEs
are
al
locat
ed
with
e
qu
al
nu
m
ber
of
RB
s.
The
sc
he
du
li
ng
is
on
ly
base
d
on
the
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Perf
orma
nce A
na
ly
sis
of
Up
li
nk Sche
duli
ng
Algo
rit
hms
in LTE
Ne
tw
or
k
s
(
Shafina
z
Bt I
s
ma
il
)
375
avail
able
RB
s,
and
the
RB
is
g
r
oupe
d
i
nto
num
ber
of
RB
s f
or
ea
c
h
UE
du
rin
g
the
sche
du
li
ng
p
r
ocess
T
he
UE
is
serv
e
d
base
d
on
the
first
c
om
e
first
ser
ve
d
strat
egy.
RR
m
ay
cause
re
duct
ion
in
t
he
e
f
fici
ency
of
the
syst
e
m
since e
ver
y
UE
does
not ha
ve t
he
sam
e Q
oS r
equ
i
rem
ents and e
xp
e
rience
d diff
e
re
nt ch
a
nn
el
co
ndit
io
n.
Seve
ral LTE upli
nk
sc
he
du
li
ng
sc
hem
es h
ave b
ee
n
disc
usse
d
by m
any re
searche
rs.
T
he per
form
ance
evaluati
on
of
the
sche
duli
ng
schem
es
hav
e
been
discusse
d
in
(13)(14)(
15)
,
w
her
e
thes
e
pap
e
rs
f
ocus
ed
on
m
axi
m
iz
ing
th
e
basic
obj
e
ct
ives
s
uc
h
as
th
rou
ghput
an
d
fa
irne
ss.
The
re
fore,
t
he
pro
pose
d
sc
hem
e
did
not
consi
der
the
Q
oS
pro
vision
i
ng.
T
he
papers
of
(4)(1
6)(17
)(11)
f
oc
us
e
d
on
m
ulti
ple
traff
i
c
su
c
h
as
vid
e
o,
V
oIP
and
be
st
ef
fort (BE)
i
n
the
upl
ink
t
ran
sm
issi
on
a
nd
to
ok
the Q
oS
int
o
the
c
on
si
der
at
io
n.
T
he
pap
e
r
o
f
(
16
)
ha
s
evaluate
d
dif
fe
ren
t
LTE
up
li
nk
sc
he
du
le
r
s
with
f
ocu
s
on
sing
le
-
bea
rer
a
nd
m
ulti
-
bear
e
r
scena
rio
an
d
Qo
S
.
The
pa
pe
r
of
(
17)
pr
ov
i
ded
a
ver
y
com
pr
e
he
ns
ive
st
ud
y
on
LTE
a
nd
L
T
E
-
A
dvance
d.
T
he
ai
m
of
the
stud
y
is
to
com
par
e
an
d
evalu
at
e
sev
eral
up
li
n
k
sc
hedulers
for
di
ff
e
ren
t
traf
fic
scenari
os
su
c
h
as
vid
eo
strea
m
ing
.
VoIP
a
nd
FT
P
within
a
sing
le
-
cel
l
env
i
r
on
m
ents.
Fina
ll
y,
the
pap
e
r
of
(11)
eval
uated
the
sch
edu
li
ng
perform
ance
of
the
upli
nk
sche
du
le
r
s,
w
hich
f
ocused
on
th
rou
ghput
an
d
fair
ness.
Ne
ver
t
heless,
othe
r
perform
ance
m
et
rics
su
ch
a
s
delay
and
pa
cket
loss
rati
on
are
not
ta
ke
n
into
co
ns
ide
ra
ti
on
.
V
ery
fe
w
pap
e
rs
hav
e
f
oc
us
e
d
on
th
e
in
vestiga
ti
on
of
m
ulti
pl
e
traf
fics
in
th
e
pe
destria
n
e
nvir
on
m
ent.
I
n
t
his
pap
e
r,
we
a
i
m
to
evaluate
the
pe
rfor
m
ance
of
sever
al
sc
he
du
li
ng
al
gorit
hm
s
fo
r
V
oIP
,
vid
e
o
an
d
BE
app
li
cat
io
ns
.
T
he
perform
ance
evaluati
on
is
cond
ucted
in
te
rm
s
of
throu
ghput,
fair
nes
s,
delay
and
PLR
in
a
pedest
rian
env
i
ronm
ent.
The
sim
ulati
on
res
ults
we
re gener
at
e
d
us
in
g
the
ope
n
s
ourc
e
LTE
syst
em
si
m
ulator
cal
le
d
L
on
g
Term
Ev
olu
ti
on
-
Sim
ulator (L
TE
-
S
IM) (
7)
.
2.
R
ESE
A
RCH M
ET
H
OD
In
this
pa
per
,
vi
deo
a
nd
V
oI
P
flo
ws
are
us
ed
for
real
tim
e
serv
ic
es
wh
il
e
in
finite
-
bu
ff
e
r
as
known
as
Be
st
effor
t
(B
E)
flo
ws
repre
sented
the
non
-
real
ti
m
e.
Vo
I
P
fl
ow
s
ha
ve
m
uch
stric
te
r
de
la
y
requirem
e
nt
th
a
n
that
of
vid
e
o
a
nd
BE
flo
ws.
Packets
tra
ns
porte
d
by
a
de
di
cat
ed
rad
i
o
be
arer
are
ge
ner
a
te
d
at
the
app
li
cat
ion
la
ye
r
by
th
ree
diff
e
re
nt
traf
fi
c
ge
ner
at
or
s;
tr
ace
-
base
d,
V
oI
P
an
d
i
nf
i
nite
-
buff
e
r.
The
tra
ce
-
base
d
a
ppli
ca
ti
on
delivers
pac
ke
ts
base
d
on
vi
deo
trace
file
s,
w
hich
a
re
obta
ined
f
ro
m
(7
)
.
T
he
voic
e
flo
ws
of
G.7
29
are
gen
e
rati
ng VoIP app
li
cat
io
n.
An
ON
/
OF
F
Ma
rkov
c
hain
is
m
od
el
ed
f
or
t
he
voic
e flow,
wh
e
re th
e m
ea
n
value
of
3
s
is
distr
ibu
te
d
ex
pone
ntial
ly
with
th
e
O
N
per
i
od
and
the
OFF
per
i
od
has
a
t
runcated
ex
po
nen
ti
al
pro
bab
il
it
y
dis
tribu
ti
on
f
un
ct
i
on
wit
h
a
n
a
ve
rag
e
val
ue
of
3
s
a
nd
a
n
up
per
lim
it
of
6.9
s
(
18).
T
he
so
urce
delivers
20
byte
s
siz
ed
packet
s
eve
ry
20
m
s
durin
g
the
O
N
per
i
od,
as
the
s
ta
nd
a
rd
source
d
at
a
rate
is
8
kbps,
wh
il
e the d
at
a
rate i
s zero
duri
ng
the
OF
F
pe
rio
d
beca
us
e of
the V
oice Acti
vity
D
et
ect
or
. Fi
nally
, th
e Inf
init
e
-
Buffer a
ppli
cat
ion
m
od
el
dem
on
st
rates a
gr
e
edy s
ource that
constantl
y p
osse
sses
pack
et
s
to b
e
d
el
iv
ere
d (
7)
.
The
pe
rfo
rm
anc
e
of
MT,
FM
E
and
RR
al
go
rithm
s
is
evaluated
base
d
on
t
hro
ughput,
fair
ness
in
de
x,
pack
et
delay
and
PLR.
T
he
Fairnes
s
ind
e
x
is
cal
culat
ed
us
in
g
Jai
n’s
fa
irness
in
de
x
m
et
hod
(
19)
an
d
it
is
expresse
d
as:
=
(
∑
1
)
2
∑
2
1
(2)
wh
e
re
xi is the
thr
oug
hput assi
gn
e
d
t
o user i
a
m
on
g
N
co
m
peting flo
ws
The
perform
a
nce
m
et
ric
of
thr
oughput
(i
n
Mbp
s
)
r
ep
res
ents
the
rate
of
s
uccess
f
ul
pack
et
bei
ng
delivere
d
over
ph
ysi
cal
c
hann
el
.
The
pa
ram
e
te
r
is
cal
culat
ed
by
div
i
ding
t
he
nu
m
ber
of
s
uccess
fu
ll
y
rec
ei
ved
bits
with
the
durati
on
of the
flow a
nd can
be m
a
them
a
ti
cally
ex
presse
d
as
:
Through
put
=
1
T
∑
∑
ptransmit
i
(
t
)
T
i
=
1
K
i
=
1
(3)
wh
e
re
ptra
ns
m
it
_i
(t)
is
the
s
iz
e
of
tra
ns
m
itted
pack
et
s
of
us
er
i
at
ti
m
e
t
,
K
is
t
he
total
num
ber
of
us
ers
an
d
T
is
the
total
sim
ul
at
ion
ti
m
e.
A
sing
le
cel
l
of
1
km
of
ra
diu
s
w
it
h
eN
od
eB
l
oc
at
ed
at
the
ce
nt
er
of
the
cel
l
is
m
odel
ed.
T
he
num
ber
of
UE
is
v
aried
from
10
to
50.
Eac
h
U
E
is
handlin
g
t
hr
ee
fl
ow
s
w
hi
ch
a
re
VoIP,
vid
e
o
an
d
BE
as
show
n
in
Figure
2.
T
he
m
ov
em
ent
of
UE
in
the
c
el
l
is
ado
ptin
g
the
rand
om
directi
on
m
od
el
.
The
spe
ed
is
set
to
3
km
/h,
wh
ic
h
re
sem
bles
the
ped
est
rian
sce
nari
o.
The
sim
ulati
on
pa
ram
et
ers
us
e
d
in LTE
-
S
IM ar
e su
m
m
arized in
Ta
ble
1.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vol
.
9
,
No.
2
,
Fe
br
uary
201
8
:
37
3
–
37
9
376
Fig
ure
2.
LTE
si
m
ulate
d
scen
ario
Table
1.
L
TE
Up
li
nk
Sim
ulatio
n Param
et
ers
Para
m
eter
Valu
e
Si
m
u
latio
n
Durati
o
n
4
0
seco
n
d
s
Tr
an
s
m
iss
io
n
Pow
er
4
3
dB
m
Cell radiu
s
1
k
m
Ch
an
n
el M
o
d
el
Macr
o
-
cell U
rban
Macr
o
c
ell P
rop
ag
atio
n
M
o
d
el
L
= 12
8
:1
+
3
7
:6
log
1
0
d @
2GHz.
(
2
0
)
Nu
m
b
e
r
o
f
Users
1
0
,20
,30
,4
0
,50
Us
ers
Tr
af
f
ic f
lo
ws
1
BE,
1 Vo
I
P,
1
V
i
d
eo
Mob
ility Mo
d
el
Ran
d
o
m
Di
re
ctio
n
Tr
an
sp
o
rt
Proto
co
l
UDP
Sy
ste
m
Ban
d
wid
th
1
0
M
Hz
Frequ
en
cy
Ca
rr
ie
r
1
.92
GHz
Nu
m
b
e
r
o
f
RBs
50
RB
Ban
d
wid
th
1
8
0
KHz
Tr
an
s
m
iss
io
n
T
i
m
e I
n
terval
1
m
s (
TT
I)
Maxi
m
u
m
Del
ay
0
.1 s
Sp
eed
3
k
m
/h
r
Vo
IP
Bit Rat
e
8
kb
p
s
VIDEO B
it Ra
te
1
2
8
kbps
3.
RES
ULTS
AND DIS
C
USSION
S
Fig
ure
3
s
ho
ws
the
a
ver
a
ge
throu
ghput
for
BE
flo
ws
’
as
the
nu
m
be
r
of
us
e
r
inc
reases.
T
he
thr
oughput
f
or
RR
al
go
rithm
decr
ease
s
as
th
e
nu
m
ber
of
U
E
increases
,
w
hile
the
thr
oughput
pe
rfor
m
ance
of
FME
and
MT
al
go
rithm
s
is
kep
t
betwee
n
20
Mb
ps
to
40
Mb
ps.
The
RR
al
g
or
it
hm
delivers
the
l
ow
est
thr
oughput
sin
ce
the
RR
pri
or
it
iz
ed
t
he
re
al
-
tim
e
flow
s
ov
e
r
BE
flo
w.
The
t
hroug
hput
of
vid
e
o
an
d
V
oI
P
flo
ws
are
sho
wn
in
Fig
ure
4
and
Fig
ure
5
r
especti
vely
.
It
is
ob
ser
ve
d
tha
t
the
thro
ug
hput
of
vi
deo
a
nd
V
oI
P
for
RR
al
go
rithm
i
ncr
eases
as
the
nu
m
ber
of
us
e
rs
incre
ases.
MT
and
FME
al
go
rith
m
s
giv
es
un
sa
ti
sfie
d
serv
ic
e
t
o
m
ult
i
m
edia
flow
s
le
avin
g
a
high
quota
f
or
BE
f
low.
Mo
reover
,
MT
an
d
FME
trie
d
t
o
m
axi
m
iz
e
thr
oughput
f
or
BE
flow
a
nd
le
aving
few
of
RB
s
to
VoIP
an
d
vid
e
o
flo
ws
th
at
le
ad
to
lo
wer
val
ues
of
thr
oughputs
for
th
os
e
flo
ws
a
s co
m
par
ed
to
t
he
RR
.
Fig
ure
3.
Th
r
ough
pu
t
f
or
BE
Flow
s
Figure
4.
Th
r
ough
pu
t
f
or
Vide
o
Fl
ow
s
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Perf
orma
nce A
na
ly
sis
of
Up
li
nk Sche
duli
ng
Algo
rit
hms
in LTE
Ne
tw
or
k
s
(
Shafina
z
Bt I
s
ma
il
)
377
Fig
ure
5.
Th
r
ough
pu
t
f
or
V
oIP Flows
Fig
ure
6.
Fair
ne
ss Index
for B
E Flo
ws
Fig
ure
7.
Fair
ne
ss Index
for V
ideo Fl
ows
Fig
ure
8.
Fair
ne
ss Index
for V
oI
P
Flo
ws
The
Jai
n
fair
ness
i
nd
e
x
for
BE,
vid
e
o
and
V
oIP
fl
ows
ar
e
prese
nt
ed
in
Fi
g
ure
6
,
7,
an
d
8
resp
ect
ively
.
F
ro
m
the
figu
r
e
s,
it
is
sh
own
that
MT
and
F
ME
al
go
rithm
s
are
hav
i
ng
si
m
il
a
r
values
of
fairn
es
s
ind
e
x
f
or
al
l
the
traf
fic
fl
ows.
The
fair
ness
ind
e
x
f
or
MT
and
FME
al
gorithm
s
m
ai
nta
ins
as
the
num
ber
of
us
ers
i
ncr
ea
ses
.
O
n
the
oth
e
r
hand,
t
he
fair
ne
ss
ind
e
x
for
RR
al
go
rithm
sta
rts
to
re
du
c
e
wh
e
n
the
nu
m
ber
of
us
ers
i
ncr
ease
s
wh
ic
h
is
sho
wing
the
ri
gh
t
trend
as
t
he
RR
al
go
rithm
is
deliverin
g
hi
gh
e
r
thr
ough
put
for
vid
e
o
a
nd Vo
I
P f
l
ow
s
. I
t
can
be
c
on
cl
ud
e
d
t
hat MT a
nd F
ME al
gorithm
s
are fai
rer t
ha
n t
he
RR
alg
or
it
hm
.
Fig
ure
9.
Dela
y for BE
Flo
ws
Fig
ure
10.
Del
ay
f
or
Vide
o
Fl
ow
s
The
delay
ex
pe
rience
d
f
or
B
E
flo
w
is
pr
es
ented
i
n
Fig
ure
9.
The
delay
will
al
ways
be
a
co
ns
ta
nt
value
of
0.0
01
seco
nd
s
for
th
e
best
effor
t
fl
ow
s
beca
us
e
it
is
m
od
el
ed
usi
ng
the
in
finite
buff
e
r
m
od
el
,
for
al
l
sche
du
li
ng
stra
te
gies.
Fig
ure
10
s
hows
the
delay
exp
e
rien
ced
by
vi
deo
f
lows
.
T
he
FM
E
al
gorithm
d
el
ivers
the
lowest
del
ay
.
The
MT
and
FME
al
gori
thm
s
sh
ow
a
sta
ble
delay
whil
e
the
RR
al
g
or
it
hm
is
hav
ing
the
highest
delay
,
wh
ic
h
in
acc
orda
nce
to
the
throu
ghput
be
ing
delivere
d.
Acc
ordin
gl
y,
there
is
a
tradeof
f
betwee
n
t
hroughp
ut
a
nd
de
la
y;
wh
e
n
th
r
oughput
is
m
axim
iz
ed,
the
delay
will
increase
.
T
he
delay
exp
e
rience
d
by
Vo
I
P
is
il
lustrate
d
in
Fi
g
ure
11,
w
hich
i
s
sign
i
ficantl
y
le
ss
than
the
delay
exp
e
rien
ced
by
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vol
.
9
,
No.
2
,
Fe
br
uary
201
8
:
37
3
–
37
9
378
vid
e
o
flo
ws.
T
his
is
m
os
tl
y
du
e
to
th
e
pac
kets
corres
pondin
g
with
voic
e
traff
ic
m
us
t
be
giv
e
n
ve
ry
hig
h
pr
i
or
it
y
and
al
l
ocated
to
a
guaran
te
e
d
ba
ndwidth
c
ha
nn
el
to
m
ake
su
re
that
the
pac
ket
sen
ding
is
within
a
n
al
lowab
le
dela
y
lim
i
t.
The
RR
al
go
rithm
reaches
the
lowest
delay
as
com
par
ed
to
the
MT
and
FME
al
gorithm
.
Fig
ure
11.
Del
ay
f
or
VoIP
Fl
ow
s
Fig
ure
12.
PL
R for BE
Flo
w
s
Fig
ure
13.
PL
R for
Vide
o Fl
ow
s
Fig
ure
14.
PL
R for
VoIP
Flo
ws
The
pac
ket
los
s
rati
os
(P
LR
)
exp
e
rience
d
by
al
l
flow
s
are
dem
on
strat
ed
i
n
Fig
ure
12
,
13,
and
14.
I
n
the
fr
am
ewo
r
k
of
QoS
pro
vis
ion
in
g,
PLR
is
a
sign
ific
a
nt
pa
ram
et
er
of
re
al
tim
e
flow
s.
PLR
increases
wh
e
n
the sche
dule
r
i
s unab
le
t
o
ti
m
el
y deli
ver
t
he real
tim
e p
acke
ts. It ca
n be
de
scribe i
n
Fi
g
ure
13 a
nd 14 t
ha
t PLR
increases
with
the
nu
m
ber
of
us
e
rs
due
to
higher
netw
ork
loa
d.
T
he
V
oI
P
flo
ws
e
xperience
d
sig
nif
ic
antly
lowe
r
P
LR
tha
n
vid
e
o
flo
ws
of
RR
al
gorith
m
becau
se
the
VoIP
traf
fic
is
delive
rin
g
l
ower
sou
rce
bit
rate
as
com
par
ed
t
o
vi
deo
flo
ws
as
il
lustrate
d
in
13
a
nd
14.
The
PLR
ac
hie
ve
d
for
t
he
real
ti
m
e
(RT)
fl
ows
ha
ve
vio
la
te
d
the
Q
oS
valu
e
f
or
P
LR.
T
he
acce
pt
able
ra
ng
e
f
or
VoIP
an
d
vid
e
o
flo
ws
s
houl
d
not
e
xceed
1x
10
-
2
t
o
1x10
-
3
res
pect
ively
.
RR
al
go
rithm
has
achieve
d
the
lo
west
PLR
for
vid
eo
a
nd
V
oI
P
by
sacrifi
ci
ng
the
avail
able
re
s
ou
rces
f
or
BE
fl
ow
s
.
Mo
reove
r,
lo
wer
valu
e
of
t
he
ta
r
get
de
la
y
ind
ic
at
es
higher
value
of
PLR
du
e
to
a
lar
ger
quota
of p
ac
ke
ts violat
in
g
the
d
ea
dline, w
hic
h
is a
dhere
d for the
v
i
deo fl
ows.
4.
CON
CLU
S
ION
This
pap
e
r
has
stud
ie
d
the
pe
rfor
m
ance
of
three
differe
nt
sche
du
li
ng
al
gorithm
s
fo
r
r
eal
-
tim
e
and
best
effo
rt
ser
vi
ces
us
in
g
the
LTE
-
S
IM.
T
he
stud
y
com
pares
the
pe
rfor
m
ance
of
th
ree
s
cheduli
ng
al
go
rithm
s
,
nam
ely
the
M
T,
FME
and
R
R
fo
r
the
pe
rfor
m
ance
m
et
ri
cs
of
thr
ough
put,
pac
ket
delay
,
PLR,
and
fa
irness
.
The
be
st
effort
,
vid
e
o
an
d
VoIP
tra
ff
ic
a
re
delive
red
by
each
UE
in
the
ped
e
stria
n
en
vir
on
m
ents
that
is
m
ov
ing
at
3
km
/h.
Fo
r
RT
Traffic
,
MT
and
FME
hav
e
the
highest
pa
cket
loss
rati
o
value
an
d
the
lowest
thr
oughput.
T
he
refor
e
,
these
al
gorithm
s
m
a
y
be
a
good
sol
ution
f
or
no
n
-
real
-
ti
m
e
flows
bu
t
is
uns
uitable
to
handle
t
he
RT
m
ul
tim
edia
serv
ic
es.
RR
al
gorithm
reaches
t
he
lo
west
PLR
am
on
g
al
l
th
ose
strat
egies
a
nd
is
the
m
os
t
su
it
able
for
V
oIP
f
lows
a
nd
vid
e
o
flo
ws.
T
his
stud
y
sho
ws
the
i
m
po
rta
nce
of
a
good
sc
he
du
li
ng
strat
egy
in
a
ne
twork
base
st
at
ion
.
F
uture
work
will
fo
c
us
on
t
he
de
vel
op
m
ent
of
new
al
go
rit
hm
,
ta
kin
g
ca
re
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Perf
orma
nce A
na
ly
sis
of
Up
li
nk Sche
duli
ng
Algo
rit
hms
in LTE
Ne
tw
or
k
s
(
Shafina
z
Bt I
s
ma
il
)
379
of
diff
e
re
nt
le
vel
of
fair
ness
a
m
on
g
us
er
s
as
well
as
qu
al
it
y
of
ser
vices
po
li
ci
es
that
is
su
it
able
fo
r
real
tim
e
and no
n real
ti
m
e traff
ic
.
ACKN
OWLE
DGE
MENTS
We a
re
gr
at
ef
ul
to
the Mi
nistr
y of Hi
gh Educ
at
ion
(MO
HE)
and Unive
rsity
Tech
nolo
gy
Ma
ra
(U
iTM
)
for t
he
r
esea
rch g
ran
t
of FRG
S
gr
a
nt
(60
0
-
RM
I/FR
GS
5/3 (
23/2
015))
as t
he
fi
na
ncial
sup
por
t d
ur
i
ng
the cou
rse of
this r
e
searc
h.
REFERE
NCE
S
[1]
ET
SI
TS
136
2
11
V10.0.
0
(2011
-
01)
Te
chnica
l
Speci
ficat
ion.
LT
E;
Evo
lve
d
Univer
s
Te
rr
Radi
o
Acc
ess
(E
-
UTRA);
Ph
y
s
ch
anne
ls Modul
(3
GP
P T
S 36211
ver
sion 1
000
R
ele
ase
10)
.
2014;0
:
0
–
121.
[2]
LT
E;
Evol
ved
U
nive
rs
al
T
err
estr
ia
l
Radio
Acc
ess
(E
-
UTRA);
Ph
y
sica
l
ch
anne
ls
a
nd
m
odula
ti
on
(3GP
P
TS
36.
211
ver
sion 8
.
3
.
0
R
e
le
ase
8). 2014
p
.
0
–
121.
[3]
Hass
an
RS
,
Rah
m
an
TA,
Abdul
rah
m
an
AY
.
LT
E
Covera
ge
Net
work
Planni
ng
and
Com
par
ison
with
Diffe
ren
t
Propaga
ti
on
Models.
TEL
KOMN
IK
A
(
Tele
communic
ati
on
Comput
Elec
tron
Control
)
.
vo
l
12
(1)
,
pp.
153.
2014
[4]
Sala
h
M.
Com
pa
rat
iv
e
pe
rform
anc
e
stud
y
of LT
E upli
nk
sch
edul
er
s.
The
sis
.
2011
.
[5]
Hida
y
at
D.
SC
-
FD
MA
LT
E
Perform
anc
e
throu
gh
High
Alti
tud
e
Plat
form
s
Com
m
unic
at
ions
(
HA
PS
)
Channe
l.
TEL
KOMNIKA
(Tele
communic
a
t
ion
Comput
El
e
c
tron Cont
rol)
.
v
ol
14(2)
,
pp.
515
–
23.
2016
.
[6]
Yada
v
SP
,
Ber
a
SC
.
PA
PR
ana
l
ysis
of
single
ca
r
r
ie
r
FD
MA
s
y
ste
m
for
upli
nk
wir
el
ess
tr
ansm
ission.
2015
10
th
Int
Conf
Informatio
n,
Comm
un
Sign
al
Proc
ess
ICICS 2015.
2016;
[7]
Piro
G,
Griec
o
LA,
Boggia
G,
Capozzi
F,
C
amarda
P.
Sim
ula
ti
ng
LT
E
C
e
ll
ula
r
S
y
stems
:
an
Open
Source
Fram
ework.
IEEE
Tr
ans V
eh
Tec
hnol
.
2010
,
pp
.
1
–
16.
[8]
Univer
sal E, Co M
A.
Et
si
Ts
1
1
36
213.
Et
si
.
201
5;V12.
7.
0
.
[9]
Li
u
Z
,
Shen
Y,
Yu
Z,
Qin
F,
Chen
Q.
Adaptive
Resourc
e
Al
l
oca
t
ion
Algorit
h
m
in
W
ire
le
ss
Acc
ess
Network.
TEL
KOMNIKA
(Tele
communic
a
t
ion
Comput
El
e
c
tron Cont
rol)
.
vol
14(3), pp.
887
.
2016.
[10]
Sala
h
M,
Ali
N, Ta
ha
A.
D
esigning Sta
ndar
d
-
Co
m
pli
ant
LT
E
Sc
hedul
ers. 2011
[11]
Ahm
ed
RE,
Al
m
uhal
la
bi
HM
.
Throughput
-
fa
ir
ne
ss
tra
deof
f
in
LT
E
upli
nk
sch
edul
ing
al
gor
it
h
m
s.
2016
Int
Co
nf
Ind
Informatic
s
Comput
Syst
CII
CS
2016.
2016
;
[12]
De
Te
m
iño
LA
MR,
Bera
rdin
el
l
i
G,
Frattasi
S,
Mogensen
P.
Channe
l
-
awa
re
sch
edul
ing
al
gori
th
m
s
for
SC
-
FD
M
A
in
L
TE
up
li
nk
.
2
008
IEEE
In
t
S
y
mp P
ers Ind
oor
Mob
Radi
o
Comm
un
PIM
RC
.
[13]
El
ga
zz
ar
K,
Sal
ah
M,
Ta
ha
A
-
EM,
Hass
ane
in
H.
Com
par
ing
upli
nk
sche
dul
ers
for
LT
E.
Proc
6th
Int
W
ire
l
Com
m
un
Mob C
om
put
Conf
ZZZ
-
IW
CMC
’10
2010
,
pp
.
189.
[14]
Safa
H,
Tohme
K.
LTE
uplink
sche
duli
ng
al
gorit
hm
s:
Per
form
anc
e
and
cha
l
le
nges.
201
2
19th
Int
Co
nf
Tele
commun IC
T
2012.
2012
.
[15]
Alqaht
an
i
S
a.
,
Alhassan
y
M.
Perform
anc
e
Mo
del
ing
and
Ev
aluati
on
of
Novel
Schedul
ing
Al
gorit
hm
for
LT
E
Networks.
2013
IEE
E
12th
Int
S
y
mp Ne
tw
Compu
t
App
l.
101
–
5.
[16]
Marwat
SN
K,
W
ee
rawa
rda
n
e
T,
Z
aki
Y,
Goer
g
C,
Ti
m
m
-
Giel
A.
Perform
anc
e
eva
lu
at
ion
of
b
andwidt
h
and
Q
oS
awa
re
LTE
upl
in
k
sche
dul
er.
201
2;7277
LNCS:2
98
–
306.
[17]
Abu
-
Ali
N,
Ta
h
a
AEM,
Sala
h
M,
Hass
ane
in
H.
Uplink
sche
duling
in
LT
E
and
L
TE
-
adv
anced:
T
utori
al,
surve
y
a
nd
eva
lu
at
ion
fra
m
e
work.
IE
EE
Co
mm
un
S
urv
Tutor
ial
s
.
vol
16(3)
,
pp.
1239
–
65.
201
4.
[18]
Chuah
CN,
Ka
t
z
RH.
Ch
aract
er
iz
ing
pac
k
et
aud
io
stre
ams
from
Inte
rn
et
m
ultim
edi
a
applic
at
ion
s.
IEEE
In
t
Con
f
Comm
un
.
2002;
2:1199
–
203.
[19]
Jain
R,
Chiu
D
-
M,
Hawe
W
R.
A
quant
itati
v
e
m
ea
sure
of
fai
r
ness
and
discrim
ina
t
ion
for
r
esourc
e
a
ll
o
ca
t
ion
in
share
d
comput
er
s
y
st
em.
Vol.
cs.
NI/9809,
DEC
technical
r
epor
t
T
R301.
1984.
p.
1
–
38.
[20]
3Gpp.
Evol
v
ed Universal
T
err
es
tri
al Ra
dio
Acc
e
ss
(
E
-
UTRA );
Radi
o
Freque
n
c
y
(
RF
)
req
uir
e
m
ent
s for
LT
E
Pico
Node
B,
TR
36
.
931.
2011;0
:0
–
28
.
Evaluation Warning : The document was created with Spire.PDF for Python.