TELKOM
NIKA
, Vol.13, No
.3, Septembe
r 2015, pp. 8
44~850
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v13i3.1799
844
Re
cei
v
ed Ma
rch 2
3
, 2015;
Re
vised J
une
5, 2015; Accepted June 2
0
, 2015
A Congestion Control Algorithm Based on Queuing
Game for Space Information Network
Chao G
uo*, Haita
o
Xu, Guocai Jia, Zh
iy
ong Yao
Schoo
l of Com
puter an
d Com
m
unic
a
tion En
gin
eeri
ng,
Un
iv
ersit
y
of Sci
enc
e and T
e
chno
l
o
g
y
Bei
jin
g,
Beiji
ng, 10
00
8
3
, P. R. China
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: guo9
9ch
ao@
163.com
A
b
st
r
a
ct
In space infor
m
ati
on n
e
tw
ork, the long de
lay an
d
hig
h
l
i
nk error rate
are the most differen
t
character
i
stics
from th
e traditi
ona
l gro
u
n
d
n
e
t
w
o
rk.
Based o
n
the
que
ui
ng
ga
me th
eory, a
nove
l
co
ngesti
o
n
control is pro
p
o
sed i
n
this pa
per. It assume
s that
the users pay an ad
mi
ssion fee for e
n
terin
g
the que
ue,
and
then
cost
duri
ng th
e w
a
it
ing ti
me, at l
a
s
t
they o
b
tain
a
be
nefit fro
m
t
he
nod
e w
h
e
n
they ar
e servic
ed
compl
e
tely. T
h
i
s
pap
er not o
n
l
y desi
gns
a v
a
ria
b
le
ad
miss
i
on fee
in or
der
that
the pro
p
o
s
ed a
l
gor
ith
m
i
s
mor
e
su
itab
le
for the sp
ace
envir
on
me
nt, but als
o
c
onsi
ders th
e e
ndi
n
g
profits
of th
e ga
me w
h
ic
h
is
descri
bed
by a
disco
unt rate.
T
he si
mul
a
tio
n
resu
lt
show
s the pro
pos
ed
alg
o
rith
m i
m
p
r
ove the
netw
o
rk
perfor
m
a
n
ce.
Ke
y
w
ords
:
co
ngesti
on co
ntrol; que
ui
ng ga
me; sp
ace
info
rmati
on n
e
tw
ork; endin
g
profit
Copy
right
©
2015 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
The
sp
ace in
formation
net
work is devel
oped
an
d fo
rmed
on th
e
basi
s
of the
grou
nd
netwo
rk. It co
nsi
s
ts of the
deep
spa
c
e
netwo
rk,
the
satellite net
work, the
aviation network a
n
d
the traditio
nal
gro
und
network [1]. Und
e
r the current
si
tuation that e
a
ch
co
unt
ry b
egin
s
to
snat
ch
the limited communi
catio
n
s resou
r
ces over spa
c
e,
how to utili
ze
spa
c
e inf
o
rmatio
n net
work
resou
r
ces effi
ciently ha
s be
come a fo
cu
s [2]. Ther
efore, whethe
r a con
g
e
s
tion control sch
e
m
e
is
good
o
r
n
o
t i
s
extremely
im
portant.
Ho
wever, d
ue to
t
he lo
ng
delay
and
the
hig
h
link
erro
r
rate
of
the high l
a
ye
r net
work in
cluding th
e d
eep
spa
c
e
n
e
twork
and t
he satellite n
e
twork [3], the
traditional
co
nge
stion cont
rol meth
od of
the gro
und
netwo
rk i
s
u
n
suitabl
e for t
he whole
spa
c
e
informatio
n n
e
twork.
In order to solve the
congestion probl
em
in the satellite network
, there
were some
resea
r
chers
who
propo
se
d rea
s
ona
ble
re
so
urce
all
o
catio
n
sche
mes a
s
in
[4]. The
auth
o
rs
adju
s
ted the traffic spe
ed
by changi
ng the con
g
e
s
tio
n
windo
w si
ze of the node in the network.
These sche
m
e
s were all b
a
se
d on the feedb
ack
information of the node
s [5, 6]. Although the
s
e
scheme
s
we
re able to
ref
l
ect the
current netwo
rk
state, the to
o long d
e
lay
cau
s
e
d
by th
e
prop
agatio
n
of the inqui
ry and feed
b
a
ck info
rmati
on led th
e
netwo
rk
re
source
s
wa
ste.
Con
s
id
erin
g t
he glo
bal
situ
ation of the
n
e
twork,
a
u
tho
r
s i
n
[7] built
a ba
rgai
ning
model
ba
sed
on
game the
o
ry. They ob
serv
ed and
analy
z
ed the
ch
aracte
ri
stic
of the network reso
urce fro
m
the
perspe
c
tive o
f
game theory, which offered a new
ide
a
for dealing
with the network
con
g
e
s
tio
n
.
What’
s
mo
re,
autho
rs i
n
[8
] formulated
a co
mp
lex
m
a
thematical model based on
the sto
c
ha
stic
differential
ga
me theo
ry. It de
scribe
d d
y
namic
ch
an
ges of n
e
two
r
k
re
so
urce
s
at any time,
and
obtaine
d an
optimal
strategy for
re
sou
r
ce
all
o
cation throug
h solving th
e feedb
ack
Na
sh
equilib
rium
of the mo
del. It wa
s
built a
c
cording
to th
e re
al n
e
two
r
k e
n
viron
m
en
t and
co
uld b
e
solving, but some oth
e
r
model
s whi
c
h were
built in the same way might not exist the
corre
s
p
ondin
g
feedba
ck Nash e
quilib
riu
m
solution.
In this p
ape
r, a cong
esti
on control al
gorithm
ba
se
d on q
ueui
n
g
gam
e for
spa
c
e
informatio
n n
e
twork (CCQ
GS). Queui
ng
game theo
ry
was firstly prese
n
ted by Naor a
s
sho
w
n
in
[9]. It modeled a que
uin
g
system
wh
ich was b
e
lo
nged to the
eco
nomi
c
field by combi
n
ing
queui
ng a
n
d
game th
eory. A consta
n
t
admissio
n
f
ee wa
s
defined to
re
stri
ct the a
c
tion
of
cu
stome
r
s
whether to j
o
in
the que
ue. In
contra
st, we
desi
gn a dyn
a
mic a
d
mi
ssi
on fee in
stea
d of
the con
s
tant
admissio
n
fe
e to
ada
pt th
e real-t
im
e
chang
es in
sp
ace
info
rmati
on n
e
two
r
k.
The
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
A Conge
stion
Control Alg
o
rithm
Based on Queui
ng G
a
m
e
for Space Inform
ation… (Ch
ao Gu
o)
845
appli
c
ation
of
CCQGS
ma
ke
s the
n
e
twork
so
cial
pro
f
it rea
c
he
s th
e maximu
m.
Then
we g
e
t
the
optimal que
u
e
length to achieve the co
n
gesti
o
n
co
ntrol in spa
c
e inf
o
rmatio
n net
work.
The rest of this
pa
pe
r
is
org
ani
zed a
s
fo
llo
ws. In
se
ction
2, the sy
stem
model i
s
descri
bed. Th
e desi
gn of conge
stion
co
ntrol
algo
rith
m unde
r the queui
ng gam
e model is gi
ven
in Section
3. The
simulat
i
on an
d com
pari
s
on
s
a
r
e
done i
n
Se
ction 5. At l
a
st, we
drew a
con
c
lu
sio
n
in S
e
ct
ion 6.
2. Sy
stem
Model
CCQGS is d
e
sig
ned to b
a
lan
c
e the space
informa
t
ion netwo
rk
load and
red
u
ce the
con
g
e
s
tion. It dete
r
mine
s the
rule
s fo
r
u
s
ers to
e
n
ter the
qu
eue
s o
f
the system
node
s.
Sin
c
e
a
newly arrived
user n
eed
s
to wait for the servi
c
e
if there a
r
e oth
e
r users
in front of him, this
waiting time
cau
s
e
s
an
extra income
when the
serv
i
c
e en
ds.
We
descri
be it as a discou
nt rate
[10] in the followin
g
system
model.
i
l
l
1
2
R
i
Figure 1. The
system mod
e
l of cong
esti
on co
ntrol for
spa
c
e info
rm
ation network
In our p
r
o
p
o
s
ed mod
e
l a
s
sho
w
n i
n
Fig
u
re 1, a
user
ii
N
who
wa
nts
to enter th
e
space i
n
form
ation net
work will obtain t
he user’
s
benefit
R
from th
e system if t
he service is
compl
e
ted. Howeve
r, the u
s
er
i
need
s to
pay an a
d
mi
ssi
on fee
i
bef
ore joi
n
ing th
e que
ue of
the
servi
c
e n
ode. Users a
rrive as
Poi
s
son
di
stributi
on
with the
a
rrival
rate
and the
net
wo
rk
node
s
se
rvice time follo
w expon
ential
di
strib
u
tion with
the servi
c
e
rate
[11]. Whe
n
a
user
enters the
qu
eue a
nd b
e
fo
re bei
ng
se
rviced, h
e
mu
st
wait for
a pe
riod of time
whi
c
h forms t
he
waiting cost p
e
r unit time
C
.
The cu
rrent numbe
r of use
r
s in the que
u
e
called the q
ueue len
g
th
are de
noted
as
l
.
The gai
n
i
Gl
of
use
r
i
is the criteria fo
r de
ciding whethe
r to join the q
ueue o
r
n
o
t. It
is comp
uted
by the u
s
e
r
’s
ben
efit subtra
cti
ng th
e admi
s
sion
fee an
d th
e waitin
g
co
st. If
0
i
Gl
, the use
r
will
join the que
ue. On the
contra
ry, if
0
i
Gl
, the user will
balk because
there i
s
n
o
g
a
i
ns eve
n
lo
sses fo
r him.
Notice that th
e
servi
c
e
nod
e
anno
un
ce
s th
e gain
i
Gl
to
the use
r
i
, it
utilizes the user’
s
benefit
R
which
sho
u
l
d
be obtaine
d after servi
c
e compl
e
ted.
Therefore, a
discou
nt rate
is introdu
ce
d
in our propo
sed syste
m
model.
In traditio
nal
queui
ng
gam
e mo
del, the
admissio
n
fe
e is
set a
s
a
con
s
tant. T
h
i
s
d
e
fine
is
not suita
b
le u
nder th
e spa
c
e inform
ation
netwo
rk
environment. We redefine th
e a
d
mission fe
e i
s
a variabl
e asso
ciated
with
the queu
e length
l
. As a result, the admission fe
e
of our sy
ste
m
model is d
e
n
o
ted by
i
l
. Once the no
de
can
not se
rvice the
use
r
s in the queu
e timely, it will
increa
se th
e
admissio
n
fe
e in o
r
de
r to
preve
n
t too
many u
s
ers t
o
join th
e q
u
eue, an
d the
n
reduce the possi
bilit
y of network
congesti
on. The rel
e
vant
parameters in thi
s
paper are shown
in
Table 1.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 3, September 20
15 : 844 – 850
846
Table 1. Spe
c
ificatio
ns of
the paramete
r
s of CCQGS
Parameters
Meaning
The user’s arrival rate
The node’s service rate
Utilization factor
=
l
The queu
e lengt
h of service node
R
The user’s benefi
t
after service completed
C
The
w
a
iting cost of the user in the
sy
stem p
e
r unit t
i
me
The discount rat
e
3.
A Con
g
es
tio
n
Control Al
gorithm Ba
s
e
d on Queui
ng Game
The system
welfare con
s
i
s
ts of
the
u
s
er’s
g
a
in
and
the n
ode’
s
profit mai
n
ly
from the
admissio
n
fe
e. Due to the
determin
ed
use
r
’s
ben
efit and the in
creasi
ng ad
mission fee
with
the
increa
sing q
u
eue length, the user’
s
gai
n and the no
de’s p
r
ofit are confli
cting. Whe
n
the qu
eue
length g
r
ows,
the use
r
’s g
a
in de
cre
a
se
s. On
ce
it is less than
zero, the use
r
b
a
lks the que
ue.
Mean
while, t
he n
ode’
s
profit increa
se
s. We
a
s
su
me
s that
a
user
joins the
que
ue at
time
0
and
c
o
mplete the s
e
rvic
e at time
1
t
. In our
system m
ode
l, t
he system
welfare of u
s
er
i
ca
n be
cal
c
ulate
d
as:
1
1
0
i
i
t
t
t
i
SE
R
e
E
C
e
d
t
(1)
Whe
r
e
t
e
denot
es en
ding b
e
nefits.
After getting the each use
r
’s
system
welfare,
the av
erag
e sy
stem
welfare co
ntribution
can b
e
com
p
uted as:
1
1
0
0
i
i
t
t
t
l
l
Sq
E
R
e
E
C
e
d
t
(2)
Whe
r
e
l
q
denot
es the p
r
oba
b
ility of observing
l
use
r
s in t
he que
ue.
Acco
rdi
ng to
the
re
stri
ctio
n conditio
n
0
i
Gl
, CCQGS
jud
ges a
user can joi
n
the
queu
e only if
he satisfie
s th
is conditio
n
. Our
co
nge
st
i
on control alg
o
rithm i
s
an
i
n
itiative sche
me,
so that users are able to
choi
ce the fa
vorable b
eha
vior. We turn
this rest
rictio
n to a thresh
old
queu
e le
ngth
of no
de. It is
ea
sy for user’
s
di
scrimination. T
h
e u
s
er’
s
gai
n
i
Gl
can
be
cal
c
ulate
d
as:
1
1
0
i
i
t
t
t
ii
Gl
E
R
e
C
e
d
t
l
(3)
Becau
s
e
that
the service
time of the n
ode
in
spa
c
e
informatio
n
netwo
rk follo
ws t
he
expone
ntial d
i
stributio
n wit
h
a pa
ram
e
te
r
. When a
ne
w u
s
er arrive
s to the n
ode
and o
b
serve
s
that there h
a
v
e bee
n
l
use
r
s already i
n
the qu
eue, th
e servi
c
e
co
mpleted
time
1
t
follows the
Erlang dist
rib
u
tion
with pa
rameters
and
1
l
. Acco
rdin
g t
o
these a
s
su
mption
we o
b
t
ains the
formulatio
n a
s
:
1
1
i
t
l
Ee
(4)
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
A Conge
stion
Control Alg
o
rithm
Based on Queui
ng G
a
m
e
for Space Inform
ation… (Ch
ao Gu
o)
847
Whe
r
e
.
Comp
uting th
e expre
ssi
on
in (3), we ca
n
obtains:
1
l
ii
CC
Gl
R
E
l
(5)
Whe
n
the u
s
er’s
gain
sati
sfies
0
i
Gl
, the re
stri
ctionexp
r
e
ssi
on can be
rewrite a
s
follow:
1
0
l
i
CC
RE
l
(6)
If the user’
s
g
a
in not only satisfies
0
i
Gl
,
but
also
sat
i
sf
ie
s
10
i
Gl
, this queu
e
length is the t
h
re
shol
d den
oted as
th
l
. The con
d
ition
s
are given as:
1
2
0
10
l
i
l
i
CC
RE
l
CC
RE
l
(7)
Solving the a
bove the i
n
e
quality equ
ations, th
e thre
shol
d qu
eue
l
ength
th
l
is
cal
c
ulated
as:
1
lo
g
2
l
o
g
1
it
h
i
t
h
th
CC
El
El
l
CC
RR
(8)
Whe
r
e
th
lN
.
The u
s
e
r
firstly come
s to
the
spa
c
e
informatio
n n
e
twork, a
nd t
hen th
e
servi
c
e
nod
e
anno
un
ce
s a
n
admi
s
sion
fee to th
e u
s
e
r
with th
e
con
s
ideratio
n of th
e current
nu
mber of u
s
e
r
s in
the que
ue. T
he u
s
er
cal
c
u
l
ates hi
s g
a
in
accord
ing to
the re
ceived
admissio
n
fe
e and th
e wai
t
ing
co
st. If the user’s
gain i
s
n
o
less than zero, he
joi
n
s
the queu
e. Otherwi
se h
e
b
a
lks. Du
ring t
he
discu
ssi
on of the user’
s
g
a
in, a thresh
old queu
e le
ngth is ded
u
c
ted for the conge
stion co
ntrol
scheme fo
r space inform
ation network. The flow
cha
r
t of CCQGS i
s
dra
w
a
s
Fig
u
re 2.
After the threshol
d queu
e length ha
s b
een ca
l
c
ul
ate
d
, we get the upper bo
und
of the
queu
e length.
This re
stri
cti
on is u
s
ed fo
r maximi
zin
g
the system
welfare an
d ba
lanci
ng the lo
ad
of the netwo
rk. To m
a
ke
our al
gorith
m
suit fo
r
spa
c
e info
rmatio
n network, we introd
uce the
endin
g
profit to the gam
e model. It descri
b
e
s
th
e inform
ation
transmissio
n
behavio
rs
more
suitabl
e with
the actu
al
situation. Fo
r
si
ngle n
ode, th
e user
com
e
s an
d de
cid
e
s
whethe
r to
join
the queue
or
not whic
h only deponds
on
his
gain. On the
other word
, if the length of the
queue
beyond
s
th
l
, the user
will bal
k. Ho
wever,
due to the
ga
ins of the join
ed use
r
s are
non-neg
ative
,
the system
welfare will stay in a opti
m
al leve
l. At last, the conge
stion co
n
t
rol improve
the
perfo
rman
ce
of the network.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 3, September 20
15 : 844 – 850
848
1
1
0
i
i
t
t
t
ii
Gl
ER
e
C
e
d
t
l
i
1
lo
g
2
l
o
g
1
it
h
i
t
h
th
CC
El
El
l
CC
RR
0
i
Gl
1
1
0
0
i
i
t
t
t
l
l
Sq
E
R
e
E
C
e
d
t
Figure 2. The
flow cha
r
t of CCQGS
4.
Simulation and Comparis
ons
In this se
ctio
n, we give so
me simul
a
tio
n
s an
d analy
s
is to prove
the co
rrectn
ess and the
effectivene
ss of CCQGS. Firstly,
the influence of the discount ra
te
to the thr
e
sh
old que
u
e
length
th
l
is experime
n
ted. Under t
he rest
riction conditio
n
of
th
l
, Matlab
is used to cal
c
ulate the
interrelated
p
a
ram
e
ters a
n
d
sh
ow the n
u
meri
cal
re
s
u
lt. It is
helpful for us
to
realiz
e the extra
benefit of system from end
ing profit of the game.
The
n
we ob
se
rve
the relation
ship between t
h
e
queu
e le
ngth
and
the
arri
val rate.
Use
r
s in
spa
c
e
i
n
formatio
n n
e
twork
co
me
with
differe
nt
probabilities, for
understanding the impact
to the sy
stem, we do t
he second ex
perim
ent. Since
our algo
rithm
is ba
sed
o
n
queui
ng
gam
e theo
ry
whi
c
h
is rarely u
s
ed in
cong
est
i
on
cont
rol, it
is
hard
to
com
p
are
with
othe
r
similar al
gorithm. We
ju
st do
so
me
nu
meri
cal
simul
a
tion to
analy
s
i
s
and compa
r
e
the results wit
h
different pa
ramete
rs.
On ba
sis of the queuin
g
game theo
ry, we
built the system mod
e
l and simul
a
te the
netwo
rk e
n
vironment. The
simulatio
n
pa
ramete
rs a
r
e
sho
w
n in Ta
b
l
e 2.
Table 2. The
value of the simulation pa
rameters
R
C
40 60
100
10
In this
simula
tion, we
defi
ne the
admi
s
sion fe
e
ll
to sho
w
n th
e
chang
es
of th
e
system. Th
e
discou
nt rate
is a
s
sumed
as
0
.
0
1
,
0
.02
,
0.03
,
0
.04
,
0.0
5
,
0
.06
,
0.07
,
0
.08
,
0
.
08
,
0
.1
.
As sho
w
n in
Figure 3, with the
incre
a
se of the discount rate
,
the threshold q
ueue len
g
th
th
l
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
A Conge
stion
Control Alg
o
rithm
Based on Queui
ng G
a
m
e
for Space Inform
ation… (Ch
ao Gu
o)
849
decrea
s
e
s
. T
he
rea
s
on
i
s
that
expre
s
s the
en
ding
profits. Th
e
more
is
, the
s
m
aller the
user’s gain becom
e
s. Once
the
user’
s
gain
drops,
the new arrival user
will
bal
k the queue
with
the restri
ctio
n
0
i
Gl
. Due to
th
lN
, we
roun
d t
he valu
e in
the
simulatio
n
and
get
the
threshold q
u
e
ue length.
Figure 3. Thresh
old qu
eue
length for different
As sho
w
n in
Figure 4, we
can ob
serve
the threshol
d queue le
ng
th
th
l
grows with the
increa
se of the se
rvice ra
te
. When the servi
c
e rat
e
s are the same, the thresh
old que
u
e
length is lo
wer with the la
rge
r
discou
nt rate. It means that after use
r
s joi
n
in the queue of
the
node i
n
spa
c
e inform
ation
netwo
rk, if the nod
e
servi
c
es the
mo
re
use
r
s pe
r u
n
i
t
time, the m
o
re
netwo
rk t
r
affic can b
e
p
r
o
c
e
s
sed in
ou
r propo
se
d model.Th
e
th
reshold que
u
e
length den
otes
the capa
city
of nod
e. Th
e
high
er
it is,
the m
o
re in
formation
ca
n be
h
andle
d
in
the
syst
em.
Acco
rdi
ng to
the qu
euin
g
game
mod
e
l, we
can
get the
optim
al system welfare
an
d avoid
t
h
e
netwo
rk con
g
e
stion.
Figure 4. Thresh
old qu
eue
length for different
0.
01
0.
0
2
0.
03
0.
04
0.
05
0.
06
0.
07
0.
08
0.
09
0.
1
160
180
200
220
240
260
280
Thr
e
sh
ol
d l
engt
h of
t
he queu
e l
th
Be
f
o
r
e
<f
l
o
o
r
>
A
f
t
e
r
<
f
l
oor
>
40
45
50
55
60
65
70
75
80
10
0
15
0
20
0
25
0
30
0
35
0
40
0
Thr
e
shol
d
l
eng
t
h
of
t
he queue l
th
=0
.0
1
=0
.0
3
=0
.0
5
=0
.0
7
=0
.0
9
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 3, September 20
15 : 844 – 850
850
5. Conclu
sion
Con
s
id
erin
g
the
sp
eci
a
l sp
ace inf
o
rmat
io
n
n
e
twork
enviro
n
ment, we desi
gn a
con
g
e
s
tion
control al
go
rithm CCQ
G
S to achi
eve
the re
ason
able di
strib
u
tion of net
work
resou
r
ces. O
n
the ba
sis
of the queui
ng theo
ry an
d the gam
e
theory, a system mod
e
l
is
formulate
d
to
redu
ce th
e p
r
obability of th
e co
nge
stion
occurrin
g. We define
a variable a
d
missi
o
n
fee in
stead
o
f
a
con
s
tant
one
so
that t
he m
odel
ca
n de
scri
be th
e net
wo
rk sy
stem m
o
re
real.
Beside
s, a di
scount rate is introdu
ce
d to the pr
opo
sed mod
e
l to
pre
s
ent the
e
nding
profit o
f
the
game. At last, better perfo
rmance of the netwo
rk i
s
ob
tained in the
simulatio
n
s.
In
the
future work, we will do some
rese
aches on the
performance
compari
s
ion between
our alg
o
rithm
and othe
rs a
n
d
solve the ro
uti
ng pro
b
lem
in spa
c
e info
rmation n
e
twork.
Ackn
o
w
l
e
dg
ements
We
gratefully
ackno
w
le
dg
e ano
nymou
s
re
viewers
who
re
ad d
r
afts and
ma
de ma
ny
helpful su
gge
stion
s
.
Thi
s
work
i
s
sup
port
ed
by
the
Nat
i
onal S
c
ien
c
e
Fou
ndatio
n
Proje
c
t of P.
R.
Chin
a (612
72
507),
China
Postdo
ctoral Scien
c
e
Fou
ndation
(201
3M530
526
), t
he F
und
ame
n
tal
Re
sea
r
ch Fu
nds for th
e Central Universities (FRF-T
P-09-015A
).
Referen
ces
[1]
F
i
aschetti A, F
i
aschetti
M, Pietrab
i
ssa A,
Petrone M.
Con
gestio
n
pri
c
ing for dy
na
mic
ban
dw
idth
alloc
a
tio
n
in s
a
tellit
e netw
o
rks: A game-th
eoretic a
ppro
a
c
h
. Satellite Telec
o
mmun
i
cat
i
ons (EST
EL),
201
2 IEEE First AESS Europe
an Co
nfere
n
ce
. 2012: 1-6.
[2]
Hon
g
y
in
g L, Z
h
imin
g Y, Z
h
i
g
ang
C. Ma
x-Mi
n Rate
Co
ntrol
on T
r
affic in Br
oad
ba
nd Mu
lti
beam S
a
tel
lite
Communications S
y
stems.
C
o
mmunic
a
tio
n
s
Letters
,
IEEE.
201
3; 17(7): 13
96-1
399.
[3]
Ganga
dh
ar S, Ngu
y
e
n
T
A
N,
Umap
athi G, Sterbenz JPG.
T
C
P W
e
stw
ood(+
)
protocol i
m
p
l
e
m
e
n
tatio
n
in ns-3
. Proce
edi
ngs of the 6
t
h Internatio
nal
ICST
Confere
n
ce on Sim
u
lat
i
on T
ools and
T
e
chniques.
Can
nes, F
r
anc
e: ICST
(Institute for
Comp
ut
er Sc
i
ences, S
o
cial-I
nformatic
s
an
d T
e
lecom
m
unic
a
tion
s
Engi
neer
in
g). 2013: 16
7-1
75.
[4]
T
a
leb T
,
Kato
N, Nemot
o
Y
.
REF
W
A: An Effi
cient a
nd
F
a
ir Co
ngesti
o
n
Co
ntrol Sc
h
e
me for
LEO
Satell
ite Net
w
o
r
ks.
Networking, IEEE/
ACM Transactions
. 20
06; 14(5): 1
031
-104
4.
[5]
T
o
rres R, Bord
er J, C
hoq
uett
e
G, Jun
X, Je
-Hon
g J.
C
o
n
g
e
stion
contr
o
l
usin
g RE
D a
n
d T
C
P w
i
n
dow
adj
ustment.
MILIT
A
RY COMMUNICAT
ION
S CONF
ER
ENCE, 2012 - MIL
C
OM 2012. 20
12.
[6]
Utsumi S, Sa
li
m Z
abir SM, S
h
iratori
N. T
C
P-C
herr
y
: A
ne
w
appr
oac
h for
T
C
P congestio
n
contro
l ov
er
satellit
e IP netw
o
rks.
Co
mp
uter Co
mmu
n
icat
ions.
20
08; 31(
10): 254
1-2
561
.
[7]
Petraki DK, Anastasopoulos
MP, Hsiao-H
w
a C, Co
ttis PG. Distributed Resour
ce Allocation for Delay
-
Sensitiv
e Servi
c
es in Satell
ite
Net
w
o
r
ks Usi
ng Game T
heor
y
.
Co
mputati
ona
l Intelli
ge
n
c
e and AI in
Gam
e
s, IEEE Transactions
. 200
9; 1(2): 134
-144.
[8]
Lon
g Z
,
Xia
n
w
e
i Z
.
Joi
n
t cro
ss-la
yer
optimi
s
ed ro
utin
g a
n
d
d
y
n
a
mic
po
w
e
r a
lloc
a
tio
n
in d
e
e
p
sp
ace
informati
on net
w
o
rks un
der pr
edicta
b
le c
onta
c
ts.
Commun
ic
ations, IET
.
2013; 7(5): 41
7-4
29.
[9]
Naor P. T
he Regu
latio
n
of Queue Siz
e
b
y
Le
v
y
i
ng T
o
lls.
Econo
metric.
196
9; 37(1): 15-2
4
.
[10]
Che
n
H, F
r
ank M. State D
epe
nde
nt Pricin
g w
i
t
h
a Queu
e.
IIE T
r
ansaction
s.
2001; 33(
10)
: 847-86
0.
[11]
Yi G, Hua L, Krishn
amach
a
ri
B.
A packet dropp
ing-
bas
ed i
n
centiv
e mech
anis
m
for M/M/1 que
ues w
i
th
selfish us
ers
. INFOCOM, 2011 Procee
di
ngs
IEEE. 2011: 26
87-2
695.
Evaluation Warning : The document was created with Spire.PDF for Python.