Int. J. Communications, Network and System Sciences, 2010, 3, 835-842
doi:10.4236/ijcns.2010.311113 Published Online November 2010 (http://www.SciRP.org/journal/ijcns)
A Routing Strategy with Link Disruption Tolerance for
Multilayered Satellite Networks
Gang Zheng1, Yanxin Guo2
1Institute of Software, Chinese Academy of Sciences, Beijing, China
2National Astronomical Observatories, Chinese Academy of Sciences, Beijing, China
Received August 19, 2010; revised September 20, 2010; accepted October 20, 2010
Abstract
Link disruption has a considerable impact on routing in multilayered satellite networks, which includes pre-
dictable disruption from the periodic satellite motion and unpredictable disruption from communication
faults. Based on the analysis on the predictability of satellite links, a link disruption routing strategy is pro-
posed for multilayered satellite networks, where a topology period is divided into non-uniform slots, and a
routing table in each slot is calculated by the topology predictability of satellite networks, and a congestion
control mechanism is proposed to ensure the reliable transmission of packets, and a flooding mechanism is
given to deal with the routes selection in the case of unpredictable link disruption. This routing strategy is
implemented on a satellite network simulation platform, the simulation results show that the strategy has
lower delay and higher link utilization, and can meet the routing requirements of multilayered satellite net-
works.
Keywords: Satellite Networks, Routing, Link Disruption Tolerance
1. Introduction
The main components of satellite networks consist of the
space segment: satellites, and the ground segment: earth
stations. Satellites are situated on the different levels,
namely, Low Earth Orbit (LEO), Medium Earth Orbit
(MEO) and Geostationary Orbit (GEO) satellite layers. A
multilayered satellite network is a combination of dif-
ferent layers of satellites, which can provide a more effi-
cient network with better performance than these layers
individually, and show great promise for the future [1].
In the multilayer satellite network, satellites in the same
layer are connected to each other via Inter-Satellite Links
(ISLs) while the communication between different layers
is accomplished over interlayer ISLs. It is so important to
design a routing scheme for delivering, forwarding and
routing packets in the multilayer satellite network. Typi-
cal multi-layer satellite network routing schemes include
Hierarchical QoS Routing Protocol (HQRP) [2],
Multi-Layered Satellite Routing algorithm (MLSR) [3],
Satellite Grouping and Routing Protocol (SGRP) [4].
HQRP for connection-oriented multi-layer satellite net-
works ensures the quality of service to long-distance
dependency (LDD) business. MLSR employs higher
level satellites to calculate shortest delay paths efficiently
between the satellites in the satellite network and the
gateways on the Earth, where the routing tables are up-
dated regularly to cope with the satellite mobility and the
changes in the network load. SGRP divides LEO satel-
lites into groups according to the footprint area of the
MEO satellites in each snapshot period, and makes MEO
satellite managers compute the minimum-delay paths for
their LEO members based on the delay reports sent by
LEO satellites. In the mentioned routing algorithms
above, the processes of collecting delay information and
calculating routing tables are finished dynamically. As
the number of satellites increase, the topology of the
multilayered satellite network gets more complex, the
transmission delay becomes longer, and the packets-loss
rates of packets are larger, the overhead of computing the
routes by the on-board computer in a satellite gets in-
creased, and the performance of the multilayered satellite
network is reduced clearly. The objective of routing al-
gorithms for multiple satellite networks is to compute
paths with low communication and computational over-
head, and adapt the routing decisions to the dynamic

836 G. ZHENG ET AL.
satellite network topology in real time. As we have
known, compared with terrestrial computer networks,
satellite networks suffer more significant transmission
delay and links disruption from the orbital heights, satel-
lites motion, the robustness of links design, space rays,
satellites power supply, etc. Therefore, multilayered sat-
ellite networks can be seen as delay and disruption tol-
erant networks (DTNs) [5]. Recently there has been
much research activity in routing problem for DTNs [6].
The delay-tolerant networking routing problem is for-
mulated in [7], which amounts to a constrained optimiza-
tion problem where edges may be unavailable for ex-
tended periods of time and a storage constraint exists at
each node. The routing problems in intermittently con-
nected ad hoc networks and delay/disruption tolerant
networks are discussed in [8], which is categorized as the
deterministic case and the stochastic case. A contact-
duration-based probabilistic routing scheme is proposed
in [9] based on the probabilistic routing scheme. An
end-to-end path in DTNs may be unavailable at all times,
and routing is performed over time to achieve eventual
delivery by employing long-term storage at the interme-
diate nodes.
As a multilayered satellite network can be thought as a
typical DTN case, it is important to investigate the feasi-
bility of routing schemes in satellite networks by refer-
ring to the routing mechanisms in DTNs. In a multilay-
ered satellite network, link disruption mainly consists of
two cases: one is predictable because of regular satellites
motion around the earth, the other is unpredictable
caused by space environment or faults in on-board
communication equipments. For the predictable links
disruption, routing is computed by predicting the satellite
network topology. For the unpredictable case, a routing
strategy is required to provide a redundant mechanism to
forward packets correctly during links disruption.
In this paper, a disruption tolerant routing strategy
(DTRS) for a multilayered satellite network is proposed,
which combines the statically routing calculation for
predictable link state and dynamical link disruption tol-
erant mechanism for unpredictable link disruption. In
DTRS routing strategy, route tables are calculated by
considering the periodical topology of satellite networks
caused by the regular motion of satellite and stored in
satellites, also, a dynamical congestion control mecha-
nism and a flooding-based mechanism are proposed to be
tolerant links disruption. This routing strategy is imple-
mented in the satellite network simulation platform [10],
and the simulation results show that this strategy has
lower computation overhead and better performance for
links disruption tolerance.
2. Links Prediction Model
2.1. Multilayered Satellite Networks
The multilayered satellite network discussed in this paper
consists of three layers of satellites, namely, LEO, MEO
and GEO satellite layers, as shown in Figure 1. The
GEO layer is composed of NG geostationary satellites.
The MEO layer is composed of the MEO satellite con-
stellation, and the number of satellites is NM. The LEO
layer is composed of the LEO satellite constellation, and
the number of LEO satellites is NL. In the multilayered
satellite network, satellites communicate with the terres-
trial gateways over User Data Links (UDLs). A terres-
trial gateway can be directly connected to multiple satel-
lites in different layers. The type of links includes ISL,
IOL and UDL; the ISLs in the same layer are divided
into inter-orbit links and inner-orbit links.
2.2. Links Prediction
In the multilayered satellite network, the location of a
satellite is determined by the Kepler’s laws, in addition,
orbit perturbations are considered. A set of six orbital
parameters is used to fully describe the position of a sat-
ellite in a point in space at any given time: semi-major
axis a, eccentricity e, inclination of the orbit plane i, right
ascension of the node Ω, the argument of perigee ω, and
true anomaly

. The links between two satellites are
determined by the visibility analysis.
Let the position vector of each satellite in the geocen-
tric coordinate system be ),,( zyxr 

, the satellite loca-
tion in the geocentric coordinate system can be calcu-
lated by the following formula:
cos sinsincoscos
sin sincoscos cos
sin sin
xR R
yR R
zR



 







 (1)
MEO layer
LEO layer
GEO layer
UDL
GEO innerlayer ISL
MEO innerlayer ISL
LEO innerlayer ISL
Interlayer ISL
Interlayer ISL
ground station
Figure 1. Multilayered satellite networks.

G. ZHENG ET AL.
837
where R orbital
enote the distance from the earth center O to
th
a ) 0h:
is a satellite orbit radius; α is a satellite
inclination; β= iπ/n is the angle between the positive axle
of y and the intersection line of satellite orbit plane and
equatorial plane, n is a track number, i is a track sequen-
tial number. Let γ denote the satellite-phase at the time
instant t, γ = ωt + γ0, where ω is the angular velocity, γ0 is
the initial phase. The above parameters are shown in the
Figure2.
Let h d
e line connecting satellite1 and satellite 2. In order to
compute the visibility between two satellites in the mul-
tilayered satellite network, define the visibility function
Δh = h − Re, where Re is the redius of the earth, the visi-
bility condition is given as follows:
12
21 sin
rr
rr
h
 




,
where 










21
21 )(
arccos rr
rr 


is the angle between the two satellite geocentric position
how
vector. Let H denote the minimum visible height be-
tween two satellites. If hH , two satellites are visi-
ble. This visible relation is sn in Figure 3(a).
b) 0h: Let

be the angle between
r

and
21 rr


, if 11 2
11 2
()
arccos 90
rrr
rrr




If the position relation between two satellites satisfies







 , the two s
are visible. This visible relation is shown in Figure 3(b).
th
. Disruption Tolerant Routing Strategy
ompared with the terrestrial networks, particularly op-
satellite
e visibility conditions, two satellites can communicate
with each other over interstellar links.
3
C
tical networks, the multilayered satellite network suffers
long propagation delay, relative high bit error rate and
limited bandwidth, in addition, relatively frequent links
disruption. The disruption of links in the satellite net-
work is caused by the motion of satellites, faults in
communication subsystems, power supply, space envi-
ronments, etc. The routing strategy for the satellite net-
work must be tolerant of such constraints as long delay,
links disruption, available power. In order to save the
power supply and computation ability, a routing strategy
is proposed based on the idea combining the static rout-
ing and dynamic routing. In terms of cyclic motion of
satellites, the topology of the satellite network changes
regularly, thus the routing table is calculated on the basis
of the regularity and predictability of satellite motion and
stored in a satellite before satellites are launched. When
calculating the routing table, the real time requirements
of different kinds of traffic in the satellite network,
transmission delay caused long distance and the more
number of hops in the lower satellite network, are con-
sidered. The propagation delay and queuing delay are
defined as two weights of a link, in addition, a mecha-
nism to limit the number of hops is employed to make
comprise between the number of hops and the transmis-
sion delay. For the unpredictable link disruption caused
communication faults or other reason, a flooding mecha-
nism is proposed to guarantee the routing service.
Figure 2. Satellite location in the geocentric coordinate sys-
tem.
satellite
2
satellit e1
1
r

Re
Center of earthO
earth
Re
2
r

12 rr



h
φ
(a)
satellit
e
2
satellite1
e
Rhh 
Re Center of earth O
earth
Re
2
r

21 rr  
h
φ
(b)
Figure 3. (a) Satellites visibisis: Δh > 0; (b). Satel-
lites visibility analysis: Δh < 0.
lity analy

G. ZHENG ET AL.
838
etwork shows
3.1. Topology Periods Discretization
T
d
he topology of the multilayered satellite n
ynamic and periodic changes. If the topology period is
divided into some slots, where each slot is an interval of
a stable topology, i.e., a snapshot, it is feasible to design
the routing strategy in a slot. The partition in one topol-
ogy period is so important, and must not only reflect the
dynamic topology timely, but also set appropriate length
of each interval. The topological dynamics characteriza-
tion for layered satellite network is investigated in [11].
In this paper, the partition of one topology period is
computed by the satellite motion and visibility condition
in Subsection 2.2. In every snapshot, a routing table is
computed and stored in the onboard computer of a satel-
lite. Note that the length of each slot is not equal. Let T
denote one topology period of the satellite network, if the
period is divided into n slots, [t0, t1], [t1, t2], , [tn−1, tn],
a discrete time sequence is got, which is denoted as
},,,,,{ 1210 nnT tttttS  . The topology change happens at
the time instants 12
,, ,
n
tt t, which means that, at the ith
time instant ti in oneriod, there must one or more links
are connected or disrupted. However, during the ith in-
terval 1 [, ]
ii
tt
, the satellite network topology is mapped
to a static topology, which is the base of computing the
optimal routes. Let graph G(t) = (V, E(t)) denote the sat-
ellite network topology with system period T at time t, 0
≤ t ≤ T , where V is the set of satellite nodes and E(t) is
the set of satellite links at t. In order to execute the dis-
cretization process, a topology period T is divided into a
group of equal intervals.
The discretization process is presented as follows:
1) Initialize the discrete
pe
time sequence set T S


an
e time instants set T of links connecting or disruptin
d
th
Path Selection
obtained
portant to ensure the low delay
of
g;
2) Based on the visibility computation in Subsection
2.2, compute the time instants at which the topology
changes and link connections or disruptions happen, and
add the time instants into the set T.
Arrange the elements in T by time and obtain the dis-
crete sequence {,,, }Sttt.
01Tn
3.2. Optimal, Suboptimal
For each time instant in the discrete sequence ST
in Subsection 3.1, the state of a link in the multilayered
satellite network can be computed and predicted. Thus,
the entire network topology in each time instant can be
obtained, moreover, the on/off status and the length of
each link are known. In the DTRS, all satellites in the
network are treated equally, and the optimal path at each
time instant of ST is calculated based on the instantane-
ous network topology.
Since most of satellite network services have real time
requirements, it is so im
data forwarding. When computing the route in the
DTRS, a link propagation delay is a main weight so that
the delay on the optimal path is the shortest. In addition,
because of the motion of satellites, long data forwarding
delay may cause link disruption, therefore, a link main-
taining time as another weight must be considered. Note
that the processing delay and the queuing delay of satel-
lites will increase as the number of hops in the route is
increased, a compromise must be made between the
number of hops and transmission delay. If the number of
hops in the route is too much via the lower layer satellite
links, it will be more fast and reliable for the data trans-
mission to choose another routing path via interstellar
links on the higher layer. In the DTRS, an appropriate
queuing delay and the limit to the number of hops are
given to be as weights of a link to compute the optimal
path in the multilayer satellite network.
Suppose in a period of the satellite network topology,
a link
s
d
I
SL between the satellite s an

ppens N on/off switches. At the time instant t, the
maintaie of
d the satellite d
ha
ning tim
s
d
I
SL is defined as
off onoff
() ,()()
() sd sd
sd
tkttkttk
Tt 



0,others
sd

 (2)
where , denote the link betwe
satellite d establishes, disconnects at
on ()
sd
tk

s
tim
off ()
sd
tk

and the satellite
e respectively,
en the
the k-th 0,1,,kN.
Consider a transmission path in the multilayer satellite
network 12 (, ,, )
n
pss s

, where s denotes the i-th
sa
i
tellite node in the path p, i = 1,,n. The path p con-
sists of s, 12
inter-satellite link
s
s
ISL , 23
s
s
ISL , …,
1nn
s
s
ISL , where 1ii
s
s
ISL

denote link between the
satellite i and the satellite i + ,. The
the path ped as
s a
, i =1 1,n − 1
weight of is defin
1
11
11
12
()
()
()
nn
iDISL
Ds
Wp ww
TT

 

11
ii
ii ii
ss
ii
ss ss





(3)
where D(si) denotes the queue processing dela
satellite node s, and denotes a propaga-
y on the
i1ii
ss


tion delay of the link
()DISL
1ii
s
s
ISL 
, w1 and w2 are coeffi-
cients, 12 1ww

.
Given the definition of the th weight, for each time
instant ibsection 3.
pa
n ST of Su1, the optimal path with the
shortest delay can be selected based on the Dijkstra algo-
rithm from the path set 12
() { (),(),,()}
n
Ptp tptpt
between the source satellite s and the sink satellite d,
where pi(t) denotes the ith
and the satellite d, i = 1, , n.
The optimal path between s and d is
path between the satellite s

G. ZHENG ET AL.
839

(4)
where
*
*
(),s.t.(())min{((),()()}(( ()))
(),(())min{((),()()}(( ()))
ii i
ii i
W ptWptp tPtH pth
ptWptWptptPtHpth




(),if
()
sd
pt pt
pt




(())
i
H
pt
* is the li
is the number of hops for the i-th path
p(t), hmit to the number of hops, in this paper,
t is set to
When computing the sub-optimal path, all links othe
optimal path are removed from the satellite networto-
3.3. Conism
packets may
xperience a relatively long time before they arrive at the
i
the limi 4.
Assume the optimal path from the source satellite s to
the sink satellite d is
112 23344
(){(,),(,),( ,),(,),( ,)}
sd
pt sssssssssd
.
n
k
pology, thus a new network topology is obtained, and the
sub-optimal path is the optimal path in the new topology
graph. If the new path set in the new network topology is
12
() { (),(),,()}
m
Ptptp tpt
 
, the sub-optimal path
()
sd
pt

 is shown in Equation (5).
ngestion Control Mecha
In the multilayered satellite network, data
e
sink satellite. In the DTRS, the next hop is obtained
based on the packets arrival time and the sink satellite,
which can ensure the continuity and correctness of data
transmission. If the traffic load on a satellite link in-
creases too fast, the congestion in the link may occur [5].
In this paper, in order to determine whether the conges-
tion has happened or not, the queue occupancy in the
output port of a satellite is monitored.
Define the queue idle ratio of a satellite output port as
1U RT  (6)
where U denotes the length o
the total queue length.
queue. Note that U can be set
ba
f the occupied queue, T is
Let
 be the congestion threshold and U0 be the mini-
mum length of the idle0
sed on the network traffic throughput. If there exists a
link in the satellite network subjects to

R, it is
shown that the traffic in this link increases so quickly.
Note that a link disruption can cause the queue idle ratio
R of the link is large, therefore a link-state reporting
mechanism must be used to determine whether the link is
disrupted, moreover it can be determined whether the
congestion or the disruption occurs in the link. If a link
subjects to

R and the link is not disrupted by the
reporting mechanism, the congestion occurs in this link.
Then the fong-up packets will be routed to the
gested, the queue idle ratio of this link will be selected
dynamically. The selection process is given as follows:
1) Via the link-state report mechanism, obtain all
neighboring satellites set 12
(){ ,,,}
Am
Ss sss of sat-
llowi
sub-optimal path to ease the congestion
link to the next hop in the sub-optimal path is still con-
el
in the link. If the
lite s at the current time instant, where i
s
s
sISL


,
i
s
s
ISL  is the congested link in the sub-optim
ompute the maximum queue idle ratio Rnext of
ghboring satellites, and select the correspondi
al path.
2) Call
the neing
satellite snext, that is
next arg max{()}1,,
iA
ii
sS
s
Rs im


 (7)
3) If Rnext = 1, the data packets is routed
, else the data packets is hanged to the queue of
uptionolerant Mechanism
e queue idle
tio
to the link
next
ss
ink
ISL
the lnext
ss
ISL .
3.4. Disr T
As mentioned in the Subsection 3.3, if th


R ra in a link of the satellite network and the
g mec
acket arrives at the satel-
lit
link is determined to be disrupted via the link-state re-
portinhanism, then the link disruption is caused by
a fault in the satellite, which is not predictable. In the
DTRS, a disruption tolerant mechanism uses a technique
known as flooding to deal with the routing selection. The
process is presented as follows:
1) For a packet sent from the source satellite to the
sink satellite, suppose at tk, the p
e s. If the disruption of the link to the next hop satellite
1
(())
ss k
s
pt
in the optimal path happens and the disrup-
tion of the link to the next hop satellite 1
(())
ss k
s
pt


in
al path happens, the flooding is activated.
2) Based on the link state reporting mall
neighboring satellites at tk is obtained, denoted by the se
the sub-optim
echanism,
t
12
(){ ,,,}
Am
s sssS

, i
s
s
ISL  .
3) The satellite s sends the flooding message to all the
neighboring satellites
te table. If the routing is success-
fu

(5)
.
4) On receiving the flooding message, the neighboring
satellite si query its rou
l, a successful response message is sent back to the
satellite s, otherwise, the satellite si starts the flooding
process.
( ),( )t pt

 ( ),if( ),s.t.(())min{(ptpt WptW
 
*
*
( )}((( )))
() (),(())min{((),()()}((()))
ii i
sd
ii i
pPtHpt h
pt ptWptWptptPtHpth



 




G. ZHENG ET AL.
840
5) The flooding routing process is completed when a
satellite, which starting flooding, receives the first suc-
cessful response message, moreover, if it has the upper
ooding sponsor, it should send successful response
m
she above, the routing strategy must be
ng delay, and limited
ard computer in the
ultilayered satellite network. In order to evaluate the
delay
roposed
cheLSR scheme, and the simulation time
1440 minutes. The simulation results are presented as
Satellite Layers
In the simulation scenario, 300 pairs of users distrib-
uting uniformly in the globe send
multilayered satellite network. The traffic model subjects
to the Poisson distribution. The result of the end-to-end
a shown in Figure 4. In thh
satellite is treated equally; the propagation delay and the
ty play ain weights.
HoweLSR, thetest paors
on
ater than the one of the
Mhown in Figure 7, both the DTRS
an
fl
essage back to the upper sponsor, however, if all the
messages it received are failure response, a failure re-
sponse message will be sent back to the upper sponsor.
4. Routing Simulation and Performance
Evaluation
4.1. Simulation Configuration
mentioned in tA
adapted to the dynamic topology, lo
processing capability of the on-bo
m
performance of the routing strategy proposed in the pa-
per, the simulation is made on the satellite network
simulation platform. Note that the simulation platform is
constructed based on the HLA/RTI framework, and can
configure scenarios of satellite networks flexibly to test,
verify and validate the key technology of satellite net-
works. A multilayered satellite network scenario is con-
figured in this simulation platform, which consists of
LEO, MEO and GEO satellites, as shown in Table 1.
4.2. Simulation Results and Performance
Evaluation
In the simulation, we compared the end-to-end
difference and the cost difference between the p
me and the Ms
is
follows.
1) End-to-end average delay
Table 1. Satellite orbital parameters.
Parameters
LEO GEO MEO
Number of Satellites 3 10 48
Orbital 35786 10355 1400
ee)
1
ype Geostationary W
Height(Km)
Orbital Inclination(degr0 45 52
Orbital Cycle 24 h 6 h 14 min
Number of Orbits 1 2 8
Constellation Torbit
alker
delta
Walker
delta
and receive data by the
verage delay ise DTRS, eac
ransmission dela
ver, in the M
critical role the link
shorth algithm i
ly used within the satellite cluster in each layer; there-
fore, the end-to-end average delay in the DTRS is
smaller than that in the MLSR.
2) Link utilization
Link utilization is compared between the DTRS and
the MLSR, which includes three cases as follows.
a) Link utilization within the individual layer. Figures
5-7 give the link utilization within the LEO layer, MEO
layer, and GEO layer, respectively. Figures 5-6 show
that the link utilization within the LEO layer and MEO
layer of the DTRS is much gre
LSR. However, as s
d the MLSR have similar link utilization within the
GEO layer.
0
Figure 4. End-to-end average delay.
Figure 5. Link utilization within the LEO layer.

G. ZHENG ET AL.
841
Figure 6. Link utilization within the MEO layer.
Figure 7. Link utilization within the GEO layer.
b) Link utilization of inter layers. Figures 8-10 give
the interlayer link utilization of the LEO-MEO,
MEO-GEO, and LEO-GEO, respectively. It can be seen
that the interlayer link utilization of the DTRS is greater
than the one in the MLSR.
It can be concluded from the above simulation results
that the end-to-end average delay of can meet the maxi-
mum delay requirements of voice service (< 100 ms),
and the DTRS routing strategy can be used to forward
voice data packets in multi-layer satellite networks.
Meanwhile, because of the relative balanced link utiliza-
tion in the whole network, the DTRS strategy can reduc
the o
Note that the time and space complexity of the DTRS
s
e
ccurrence probability of bottleneck links.
i relatively low compared with other routing schemes.
In the DTRS, the time complexity mainly includes two
parts, one is the time of the topological prediction and
the other is the time of selecting route, both of which are
2
()ON, where N is the number of satellites, however,
for the onboard routing algorithm, e.g., Bellman algo-
hm, the average time complexity is (log)ONN , and
the maximum time complexity is 3
()ON under the ex-
treme case. The space complexity of the DTRS is
(( 1))
t
OK N
rit

, where Kt is the number of discrete time
segments, however, the other routing mechanisms need a
network connectivity metric of 2 ()ON and a route ta-
ble of ()ON .
Figure 8. Link utilization of the inte (LEO-MEO). rlayer
Figure 9. Link utilization of the inteMEO-GEO). rlayer (
Figure 10. Link utilization of the interlayer (LEO-GEO).

G. ZHENG ET AL.
842
. Conclusion
In this paper, a routing strategy with links disruption
tolerance is proposed for multilayered satellite networks,
which employs the predictability of topology to compute
the optimal route in each slot, and a dynamical conges-
tion control mechanism is designed to balance the traffic
load and ensure the reliability of packets transmission,
and a flooding mechanism to process the routing selec-
tion is presented for the unpredictable links disruption.
This routing strategy is verified on the satellite network
simulation platform, and the simulation shows that t
proposed strategy has better performance in terms of
delay and link utilization and can support routing in mul-
t
nd M. D. Bender, “MLS
Novel Routing Algorithm for Multilayered Satellite IP
ll and R. Patra, “Routing in a Delay Toler-
elay Tolerant
d M. T. Zhou,
5
[4]
pp.
he
Issue on Delay and Disruption Tolerant Wireless Com-
munication, Vol. 26, No.5, 2008, pp.820-827.
[6] S. Farrell and V. Cahill, “Delay- and Disruption-Tolerant
Networking,” Artech House, Boston, 2006.
[7] S. Jain, K. Fa
ilayered satellite networks efficiently.
. Acknowledgements 6
The work was supported by the Knowledge Innovative
Foundation of Chinese Academy of Sciences. The au-
thors wish to thank anonymous referees for their sugges-
tions for improving this paper.
7. References
[1] Z. L. Sun, “Satellite Setworking: Principles and Proto-
cols,” John Wiley & Sons Press, Hoboken, 2005.
[2] J. Lee and S. Kang, “Satellite over Satellite（SOS）
Network: A Novel Architecture for Satellite Network,”
Proceedings of the IEEE INFOCOM, Tel Aviv, Vol.1,
2000, pp. 315-321.
[3] I. F. Akyildiz, E. Ekici aR: A
Networks,” IEEE/ACM Transactions on Networking, Vol.
10, No. 3, 2002, pp. 411-424.
C. Chen, E. Ekici and I. F. Akyildiz, “Satellite Grouping
and Routing Protocol for LEO/MEO Satellite IP Net-
works,” Proceedings of the 5th ACM International
Workshop on Wireless Mobile Multimedia, Atlanta, 2002,
109-116.
[5] C. Caini, P. Cornice, R. Firrincieli and D. Lacamera, “A
DTN Approach to Satellite Communications,” IEEE
Journal on Selected Areas in Communications, Special
ant Network,” IEEE SIGCOMM, 2004, pp.145-148.
[8] Z. Zhang, “Routing in Intermittently Connected Mobile
Ad Hoc Networks and Delay Tolerant Networks: Over-
view and Challenge,” IEEE Communications Surveys &
Tutorials, Vol. 8, No. 1, 2006, pp. 24-37.
[9] C. S. Lin, W. S. Chang, L. J. Chen and C. F. Chou, “Per-
formance Study of Routing Schemes in D
Networks,” Proceedings of IEEE AINAW, Okinawa,
March 2008, pp. 1702-1707.
[10] Y. Xu and G. Zheng, “Modeling and Distributed Simula-
tion of Satellite Networks,” Computer Simulation, Vol.
25, No. 2, 2008, pp. 65-69.
[11] J. F. Wang, Y. M. Hu, H. X. Zhou an
“Topological Dynamics Characterization for Layered Sat-
ellite Networks,” Proceedings IEEE Conference Per-
formance, Computing and Communications, Phoenix,
April 2006, pp. 349-356.