The Egyptian Journal of Remote Sensing and Space Sciences 26 (2023) 557–561

Contents lists available at ScienceDirect

The Egyptian Journal of Remote Sensing and Space Sciences

journal homepage: www.sciencedirect.com

Research Paper

Evaluation of delay tolerant network routing protocols for data

streaming through LEO constellation
Hager M. Elhalafawy a, Abdelmgeid A. Ali b, Eman M.G. Younis c, Khalid elBahnasy d, M. Medhat Mokhtar e
a
Faculty of Computers and Information, Minia University, Egypt
b
Faculty of Computers and Information, Minia University, Egypt
c
Department of Information Systems, Faculty of Computers and Information, Minia University, Egypt
d
Artificial Intelligence-Information System Department, Faculty of Computers and Information, Ain Shams University, Egypt
e
Electronics and Communication Department, High Institute for Engineering and Technology at Obour, Egypt

a r t i c l e i n f o a b s t r a c t

Article history: In the Era of Mega Low Earth Orbit (LEO) satellite constellation, the efficient networking through Inter-
Received 27 March 2023 Satellite Links (ISLs) plays a vital role in its mission success. The deployed networking resources must be
Revised 13 June 2023 justified regarding the feasibility, reliability, and Quality of Service (QoS). The main challenge of such net-
Accepted 14 June 2023
works is the ISLs intermittence due to the dynamic nature of the constellation topology. In addition to the
space environment. From this perspective, Delay Tolerant Network (DTN) protocols were proposed for
alleviating the events of intermittence. This paper introduces a technique for using the discrete-event
Keywords:
network simulator (NS3) for the simulation of LEO constellation networks. This technique considers
DTN routing protocols
Leo Constellations
the dynamic nature of network’s nodes in data streaming through ISLs. The paper presents a performance
NS3 Simulation assessment of LEO constellation networks dedicated for forwarding the payload data through ISLs. The
Congestion control performance measures include both the delay and delivery ratio. The assessment considers different local
Epidemic resources such as transmitted power, Queue size, and transmission rate. The performance measures were
Binary Spray and Wait applied for some selected DTN protocols such as ‘‘Epidemic” and ‘‘Spray and Wait”. The results illustrate
the effects of different network resources and configurations that may affect the constellation mission.
Ó 2023 National Authority of Remote Sensing & Space Science. Published by Elsevier B.V. This is an open
access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction flexible communication requirements. A comprehensive survey

on DTN protocols is introduced in (Caini., C. , 2021) and (Madni
Mega satellite constellations developments are being deployed et al., 2020), where three groups of DTN routing protocols are pre-
for achieving high-quality services of imaging applications, and sented. The first is Flooding-Based (FB) (Yulianti et al., 2020),
communications. The need for more effective space-terrestrial where an infinite number of replicas for each generated bundle
communication methods and technologies that can successfully can be distributed throughout the network. The second is Quota-
transfer massive amounts of data between space and ground net- Based (QB) in which the replicas are limited to a fixed number.
works has been driven by these developments. Recently, a bunch The third is Forwarding-Based (FoB), where a single copy of mes-
of starlink satellites has been launched to join the constellation sages or bundles is forwarded onto contacts until it reaches its des-
of few thousands already existing in space. The satellites of this tination. The well-known FB ‘‘Epidemic (Ep)” routing spreads
bunch are equipped with optical Inter-Satellite Links (ISLs) for data bundles over the network is similar to how a viral Ep spread, which
streaming through the space network rather than continuous might result in a high likelihood of messages reaching their desti-
bouncing between space and ground. A significant improvement nations in the least amount of time. The QB ‘‘Spray and Wait
in the network performance is expected, specifically in terms of (SaW)” technique can be implemented by injecting a limited num-
network delay and capacity. In this regard, and due to the intermit- ber of copies into the network. Also, Binary-SaW (B-SaW) routing
ted nature of ISLs, Delay-Tolerant Networking (DTN) has been rec- technique aims to enhance the effectiveness of the SaW in the for-
ognized as a strategy that can alleviate the challenges of utilizing warding process. Generally, DTN architecture is supported by con-
nectionless protocols with an upper bound on the signal
propagation delay. Therefore, intermediary nodes may choose to
E-mail addresses: dr_abdelmgeid@minia.edu.eg (A.A. Ali), eman.younas@mu. temporarily retain packets to be forwarded later to an appropriate
edu.eg (E.M.G. Younis), Khaled.bahnasy@cis.asu.edu.eg (K. elBahnasy), medhat.
neighbor or drop them to avoid congestion. The decision is mainly
mokhtar@oi.edu.eg (M. Medhat Mokhtar)

https://doi.org/10.1016/j.ejrs.2023.06.005
1110-9823/Ó 2023 National Authority of Remote Sensing & Space Science. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
H.M. Elhalafawy, A.A. Ali, Eman M.G. Younis et al. Egypt. J. Remote Sensing Space Sci. 26 (2023) 557–561

governed by the available local resources, which must be justified availability of buffers, transmission time window, and packet ser-
regarding their feasibility, reliability, and QoS (Iranmanesh et al., vice time.
2012). This paper presents a performance assessment of a LEO net-
work of 66 satellites of starlink constellation, currently orbiting the
Availability of buffers: The buffer of the receiving node is avail-
space. These satellites were picked to fulfill a uniform distribution able if the occupancy does not exceed a fixed threshold
in space. The assessment task is performed using a DTN model on (Bohloulzadeh and Rajaei, 2020). If it does, a congestion is rec-
NS3 platform. This model was originally introduced by (Lakkakorpi ognized, and a notification of congestion will be sent.
and Ginzboorg, 2013) for a terrestrial ‘‘ad hoc” network and recon-
Transmission time window: A congestion is detected if the
figured in this work to the previously mentioned space network. transmission time of a data packet exceeds a predefined thresh-
This paper is organized as follows. Section II presents an overview old defined by the time window.
of DTN protocols and Congestion Control (CC) methods. In Sec-
The packet service time is the interval between the packet’s
tion III, the DTN space model is described. The performance mea- transmission time and MAC layer arrival time. Every time the
sures are described in Section IV, and simulation results and packet service time exceeds the threshold, congestion is
discussion are presented in Section V. Finally, the conclusion is recognized.
outlined in Section VI.
3.2. Congestion notification
2. DTN routing protocols
The sending nodes must be alerted when congestion is identi-
fied in order to take the necessary countermeasures. Explicit or
There are three classes of DTN routing protocols; FB, QB, and
implicit propagation of con– gestion information is possible. One
FoB. One of the main targets of this work is qualitatively comparing
objective of notification is preventing congestion in proactive net-
these protocols in a dynamic environment of LEO satellite constel-
works that function to prevent congestion-induced message drops
lations. This is different from previously published research. The
(Lakkakorpi, 2011). The technique notifies nearby nodes of a node’s
quality measures are the Delivery Ratio (DR), and Latency. The
buffer occupancy so that they can make local decisions and refrain
DR is defined as the ratio of successively delivered packets to its
from sending messages to nodes whose buffers are almost full.
sum with the lost packets. The Latency is the delay contribution
When resources are available, the suggested CC method permits
from all the network elements including the satellite’s Queue as
maximum resource use. However, the method delays message
a network node (Verma et al., 2021). When it comes to maximizing
transfers until enough resources are available to avoid congestion
the DR and minimizing delivery Latency,
when resources are limited. Any source will automatically deter-
FB protocols can perform better. However, because bundle
mine the level of network congestion based on the buffer status
replication and dissemination occur at such a high rate, FB over-
notifications it receives from its neighbors.
head is significant. QB protocols may perform more efficiently in
terms of cutting down delays and boosting throughput. FoB proto-
cols are thought to be resource-friendly, as stated in (Godha et al., 3.3. Congestion mitigation
2020). A relatively simple illustration of FB is the pandemic. When
a node comes into touch with other nodes, this protocol makes sev- During this phase, network traffic is managed in accordance
eral replicas of the message and transfers messages across them. A with available bandwidth using a variety of techniques, including
high probability of delivery can be reached using the naive Ep pro- adjusting traffic rates, increasing and changing the usage of
tocol, on the account of the nodes and network resources. Here, resources (channels and nodes), giving priority to certain nodes
message delivery can eventually occur with just pair-wise connec- and packets, and modifying queue lengths (Ghaffari, 2015).
tivity (Alaoui et al., 2015) and (Garg et al., 2021). This point must Another more convenient CC strategy in the dynamic space net-
be considered carefully due to large dynamic range of the number work is presented in (Lakkakorpi, 2011). Its aims is to use an adap-
of neighbors in such dynamic network, particularly near the earth tive CC decision at the nodes rather than basing these decisions on
poles. For avoiding the network congestion in some regions, in this the network conditions at the time of the message creation. The
dynamic range must be limited, by lowering the upper bound of nodes decide forwarding to neighbors that advertise having
the number of neighbors. An alternative approach is applying a enough available buffer capacity. In the suggested technique, Hello
CC method. The effectiveness of SaW protocol may be enhanced packets, which nodes frequently broadcast to one another, are used
by the B-SaW technique (Cao and Sun, 2012). During the spray to spread buffer occupancy information (nodes use the same pack-
phase, a source node will send half of each message’s copies to ets to discover each other). The advertise buffer availability is used
each unique node until that node has only received one copy. At by nodes forwarding messages to determine how much data the
that point, the network switches to the wait phase. The network next hop node allows for forwarding.
then transmits a copy of the message directly to the destination
during the waiting phase. 4. DTN space model

The model uses a dataset of Starlink LEO satellite constellation

3. Congestion control (CC)
operated by SpaceX. It contains 66 uniformly distributed satellites
of different altitudes and inclinations. The Simplified General Per-
CC is a method for regulating or managing the quantity of net-
turbations (SGP4) propagator is used to predict the progress in the
work traffic for efficient resources utilization. In general, it has
altitude of each satellite at intervals of length one second. The
three phases; detection, notification, and mitigation (Garg et al.,
propagator provides the altitude (state vector) of each satellite in
2021).
(x, y, and Z) coordinates, and its velocity components with their
time epoch (t). These data are the main entry to the trace file of
3.1. Congestion detection the simulator. The DTN routing protocol and CC are similar to the
work presented in (Cao and Sun, 2012). In addition to the traffic
Congestion detection: To detect congestion, three factors may and wireless channel model. The main difference in the presented
be considered (Bohloulzadeh and Rajaei, 2020). These factors are work is the mobility model where the nodes mobility is determin-
558
H.M. Elhalafawy, A.A. Ali, Eman M.G. Younis et al. Egypt. J. Remote Sensing Space Sci. 26 (2023) 557–561

istic and steady. While the network topology is dynamic. This is a Table 1
typical behavior of LEO Satellite in a constellation. In the CC Network Simulation Parameters.

applied method, forwarding messages of buffer availability are Parameter Value

advertised by nodes. The forwarding process is governed by receiv- Rx Noise Figure 7 and 2
ing an availability message from the intermediate node that adver- Tx Power Levels 1dBw
tises an available buffer capacity. In the suggested technique, Hello Tx Power End 12.5 and 11.5 and 10.5
packets, which are nodes frequently broadcasting to one another. Tx Power Start 12.5 and 11.5 and 10.5
EnergyDetectionThreshold 96dBw
They are used to spread buffer occupancy information (nodes use ClearChannelThreshold 99dBw
the same packets to discover each other). The advertised buffer size Tx Antenna Gain 1.0 dB
Ba is determined by subtracting the current Buffer occupancy (Bo) Rx Antenna Gain 1.0 dB
and the overall buffer size Bs multiplied by the Congestion Thresh-
old (TC).
The Hello and anti-packets messages, as an efficient mechanism cern in inspecting the trace file are the altitude periodicity and
for reducing the massage delivery delay, are applied in the model accuracy of each satellite in reference to the neighboring satellites
(Cao and Sun, 2012). Every 100 ms, DTN nodes send Hello mes- and the calculated distance between each other. The effect of
sages to one another in order to announce the buffer status. After changing the network parameters on the number of the available
750 s, the bundle’s lifetime is scheduled to expire, and all copies neighbors is carefully validated. For example, Noise Figure, Energy
of the bundle will be removed. If the sender doesn’t receive a Detection Threshold (receiver sensitivity), and Clear Channel
return receipt within retransmission timeout (1000 s), it will re- Threshold (accepted Carrier to Interference level). The intuitive
transmit the bundle. Return receipts serve as anti-packets. Their validation of the second phase is investigated through the pre-
lifetime is the minimum of retransmission timeout less bundle for- sented simulation results.
warding time and bundle lifetime (750 s). Message copies are This section presents the simulation results of different network
deleted once a copy arrives at its destination to prevent the nega- setup (Configuration and Initialization). The configuration includes
tive impacts of congestion. the number of satellites, buffer size, and applied DTN protocols.
While the initialization denotes different parameters that may
5. Performance metrics impact the network performance like power level, Noise Fig-
ure (NF), Energy detection threshold, and clear Channel threshold.
The metrics were chosen to be relevant for a network mission in The performance measures are DR, DP, Average and Maximum BO.
space. As mentioned earlier, it is devoted for data streaming The level of Energy detection threshold, and clear Channel thresh-
between the satellites as a network node. Therefore, the key Met- old are fixed and equal 96, and 99dBw; respectively. The setting
rics of the performance are the DR and the Delay Packets (DP). At of the network parameters is selected as shown in Table 1.
the same time, there are other metrics that reflect the network per- The definition of the parameters is as follows:
formance on the nodes and ISL resources like Average and Maxi- Noise Figure (NF): is the measure of the degradation of the Sig-
mum of Buffer Occupancy (BO), and Overhead Ratio (OH). These nal to Noise ratio (S/N) in a device that is caused by the compo-
metrics are described in detail below (Bohloulzadeh, 2014): nents of the system. As the NF values become lower, the
Delivery Ratio (DR): It is the ratio of the number of delivered performance is better.
messages to the number of created messages (the sum of delivered TxPowerLevels: Number of transmission power levels between
and dropped), as defined in Eqn (1). TxPowerStart and TxPowerEnd included.
TxPowerEnd: Maximum transmission level (dBm).
DR ¼ Ndeliv ered=Ncreated ð1Þ TxPowerStart: Minimum transmission level (dBm).
Delay Packets (DP): It is the sum of the number ‘‘N” of Packets EnergyDetectionThreshold: The minimum energy threshold of a
subjected to a delay (Dpn) in the source and intermediate nodes received signal in dBm at the receiver for successful signal
until arriving to the destination, as defined in Eqn (2). processing.
CcaMode1Threshold: The maximum energy threshold of sig-
X
N
nals in the used channel at the sender to consider the channel as
DP ¼ Dpn ð2Þ
n¼0 free.
TxGain: Transmission gain in dB of the transmitting antenna.
Average and Maximum Buffer Occupancy (BO): The average RxGain: Reception gain in dB of the receiving antenna.
BO denotes the average over a specified time period, while the The network is configured as a 66 satellites constellation in
Maximum BO is the maximum buffer length which is recorded in orbits specified from the Starlink network. The buffer size equals
a specified time period. 11Mbytes. While the applied protocols are Ep, Binary SaW (B-
Overhead Ratio (OR): It is the ratio of total messages relayed
(replicated) to total messages delivered, as defined by Eqn(3).
OH ¼ Nrelayed=Ndeliv ery ð3Þ

6. Simulation results

The validation process of the simulation model is performed in

two main phases:
The first phase includes both the accuracy of the satellites alti-
tude in the trace file and the effect of different network parameters
listed in Table 1. The second phase is an intuitive validation of the
impacts of different traffic parameters on network performance.
In the first phase, since the satellites (network nodes) mobility Fig. 1. The BO for Tx power levels 10.5, 11.5, and 12.5dBw at NF equals 7, and
is steady and deterministic in a dynamic network, the main con- different protocols.

559
H.M. Elhalafawy, A.A. Ali, Eman M.G. Younis et al. Egypt. J. Remote Sensing Space Sci. 26 (2023) 557–561

SaW), Ep with TC = 0.8, and 0.9, B-SaW with TC = 0.8, and Ep-
Adaptive. Fig. 1.a, b and c show the BO at NF equals 7 for three
Tx power levels; 12.5, 11.5, and 10.5 dBw; respectively. For the
case of 12.5 dBw in Fig. 1.a, it shows that the Ep requires larger size
of buffer by about 52% more than B-SaW. It also shows that apply-
ing the CC causes an increase in the buffer occupancy in both Ep
and B-SaW by around 62%. This result is expected since the net-
work CC is attained on the account of the buffer size. By reducing
the transmitted power to 11.5 and 10.5 dBw, the available number
of neighbors is decreased, and the nodes must store the packets for
longer time. In average, it causes an increase in the buffer occu-
pancy by about 6% and 15% relative to the case of 12.5dBw as Fig. 3. The Delay for Tx power levels 10.5, 11.5, and 12.5dBw at NF equals 7, and
shown in Fig. 1.b, and c; respectively. different protocols.
The network performance is illustrated in Figs. 2, and 3 in terms
of DR and DP; respectively. As shown in Fig. 2.a and 3.a, the case of
Tx power equals.
12.5 dBw, the DR and DP characteristics are directly related to
the BO for the applied protocols. Where the increase in size of
the BO causes an increase in both measures. This feature is also
noted in both Fig. 2.b, and c, and Fig. 3.b, and c for the other two
cases of Tx power; 11.5 and 10.5 dBw; respectively.
The simulation results of the OH are shown in Fig. 4.a, b, and c
for three the three different Tx power level 12.5, 11.5, and 10.5
dBw; respectively. For the case of 12.5dBw in Fig. 4.a, the OH in
Ep is greater than B-SaW by about 40%. When the CC is applied,
the OH is increased in Ep by 20%, and in B-SaW by 15%. This feature
is also expected since increasing the number of packets in the Fig. 4. The OH for Tx power levels 10.5, 11.5, and 12.5dBw at NF equals 7, and
replica process led to an increase in the DR and OH. In the two different protocols.
cases of Tx power equal 11.5 and 10.5dBw, the OH is decreased
as a result of reducing the available number of neighbors was
decreased.
In the following, the network performance is investigated at a
much better NF of values 2 rather than 7 of the above cases. For
this case, a much better network performance is expected for all
the applied protocols. In terms of the four-performance metrics
at NF = 2, Figs. 5–8 are the correspondence to the cases at NF = 7
in Figs. 1–4. The comparison shows that reducing the NF causes
a small change in the BO, much higher than utilization in the DR
accompanied with an increase in the DP and the OH. The lower
value of NF increases the number of available neighbors. Hence
increases the network traffic as well as the number of DP. It worth Fig. 5. The BO for Tx power levels 10.5, 11.5, and 12.5dBw at NF equals 2, and
notifying that the small change in the BO is due to an increase in different protocols.
the number of packets departure rate as well as the arrival rate.
Generally, the net difference between both is small. These results,
as expected, are associated with an increase in DR. From the per-
formance point of view, as can be noted in the Figures, the B-
SaW performance is better than others in this particular configura-
tion of satellite constellation.

Fig. 6. The DR for Tx power levels 10.5, 11.5, and 12.5dBw at NF equals 2, and
different protocols.

7. Conclusion

The simulation results presented a satisfactory performance of a

space network between satellites in a constellation in an intermit-
ted environment where a DTN routing protocol is deployed. The
intermission is due to limited resources of a satellite as a network
Fig. 2. The DR for Tx power levels 10.5, 11.5, and 12.5dBw at NF equals 7, and node; transmitted power, and data storage size. From a selected
different protocols.

560
H.M. Elhalafawy, A.A. Ali, Eman M.G. Younis et al. Egypt. J. Remote Sensing Space Sci. 26 (2023) 557–561

Declaration of Competing Interest

The authors declare that they have no known competing finan-

cial interests or personal relationships that could have appeared
to influence the work reported in this paper.

References

Alaoui, E.A., Agoujil, S., Hajar, M., Qaraai, Y., 2015. The performance of dtn routing
protocols: a comparative study. WSEAS Trans. Commun. 14, 121–130.
Fig. 7. The Delay for Tx power levels 10.5, 11.5, and 12.5dBw at NF equals 2, and Bohloulzadeh, A., Rajaei, M., 2020. A survey on congestion control pro- tocols in
different protocols. wireless sensor networks. Int. J. Wireless Inf. Networks 27, 365–384. https://doi.
org/10.1007/s10776-020-00479-3.
Bohloulzadeh, A., Rajaei, M.: Evaluation of spray based routing approaches in delay
tolerant networks. Journal of Communications Soft- ware and Systems. 10, 213–
220 (2014). doi: 10.24138/jcomss. v10i4.117.
Caini, C., 2021. Delay-tolerant Networks (DTNs) for Satellite Communications.
Elsevier, Bologna, Italy.
Cao, Y., Sun, Z., 2012. Routing in delay/disruption tolerant networks: a taxon- omy,
survey and challenges. IEEE Commun. Surv. Tutorials 15, 654–677. https://doi.
org/10.1109/SURV.2012.042512.00053.
Garg, P., Dixit, A., Sethi, P., 2021. Performance comparison of fresh and spray & wait
protocol through one simulator. IT in Industry 9 (2).
Ghaffari, A., 2015. Congestion control mechanisms in wireless sensor networks: A
survey. J. Netw. Comput. Appl. 52, 101–115. https://doi.org/10.1016/j.
jnca.2015.03.002.
Godha, P., Jadon, S., Patle, A., Gupta, I., Sharma, B., Singh, A.K.: Flooding and
forwarding based on efficient routing protocol. International Conference on
Innovative Computing and Communications: Proceedings of ICICC 2020,
Volume 2, Springer Singapore, 215-223 Jul (2021).
Fig. 8. The OH for Tx power levels 10.5, 11.5, and 12.5dBw at NF equals 2, and
Iranmanesh, S., Raad, R., Chin, K.-W.: A novel destination-based routing protocol
different protocols. (DBRP) in DTNs. 2012 International Symposium on Com- munications and
Information Technologies (ISCIT), Gold Coast, QLD, Australia, 325-330 Oct
(2012).
group of protocols, B-SaW achieved the best performance by Lakkakorpi, J., Ginzboorg, P.: ns-3 Module for routing and congestion control studies
in mobile opportunistic DTNs. 2013 International Sym- posium on Performance
achieving higher DR and better utilization of BO with decrease in Evaluation of Computer and Telecommunication Systems (SPECTS), Toronto,
OH compared with others. In addition, using the strategy of CC ON, Canada, 46-50 Jul (2013).
achieved a better DR utilization on the account of BO and traffic. Lakkakorpi, M.P. Jani, Ott., J., Using buffer space advertisements to avoid congestion
in mobile opportunistic DTNs. In Wired/Wireless Inter- net Communications:
In addition to reducing the NF causes some changes. In general, a 9th IFIP TC 6 International Conference WWIC 2011, Vilanova i la Geltrú, Spain,
small change in the BO, much higher utilization of the DR accom- 386-397 June (2011).
panied with an increase in the DP and the OH. It is worth notifying Madni, M.A.A., Iranmanesh, S., Raad, R., 2020. Dtn and non-dtn routing protocols for
inter-cubesat communications: a comprehensive survey. Electronics 9 (3), 482.
that, the applied strategies of DTN forwarding protocols may cause
Verma, A., Savita, Kumar, S., 2021. Routing protocols in delay tolerant networks:
a sever network congestion in both cases of high rate of ISLs inter- Comparative and empirical analysis. Wirel. Pers. Commun. 118 (1), 551–574.
mittent, or mega constellation of thousands of satellites. Therefore, Yulianti, D., Mandala, S., Nasien, D., Ngadi, A., Coulibaly, Y., 2020. Perfor- mance
for alleviating these limitations, the packets streaming using DTN comparison of epidemic, prophet, spray and wait, binary spray and wait, and
prophetv2. J. Emerging Technol. Innovative Res. 7, 882–893.
protocols must be terminated after a defined number of hops.

561