Fairness and Satisfaction Model for DTN Applications - HAL

Fairness and Satisfaction Model for DTN Applications
Using Various Transportation Means
Mohammed Al-Siyabi, Haitham Cruickshank, Zhili Sun, Godwin Ansa
To cite this version:
Mohammed Al-Siyabi, Haitham Cruickshank, Zhili Sun, Godwin Ansa. Fairness and Satisfaction
Model for DTN Applications Using Various Transportation Means. Xavier Masip-Bruin; Dominique
Verchere; Vassilis Tsaoussidis; Marcelo Yannuzzi. 9th Wired/Wireless Internet Communications
(WWIC), Jun 2011, Vilanova i la Geltrú, Spain. Springer, Lecture Notes in Computer Science,
LNCS-6649, pp.446-457, 2011, Wired/Wireless Internet Communications. <10.1007/978-3-642-215605_37>. <hal-01583639>
HAL Id: hal-01583639
https://hal.inria.fr/hal-01583639
Submitted on 7 Sep 2017
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
Distributed under a Creative Commons ttribution�ttribution 4.0 International License
Fairness and Satisfaction Model for DTN Applications
Using Various Transportation Means
Mohammed Al-Siyabi, Haitham Cruickshank, Zhili Sun and Godwin Ansa
Centre for Communication Systems Research
University of Surrey
Guildford, UK
{m.alsiyabi, h.cruickshank, z.sun, g.ansa }@ surrey.ac.uk
Abstract. Delay Tolerant Network suffers from lack of resources and
disconnected contact nature. In DTN, all possible methods are used to transmit
data including the physical transportations means. Aircrafts in commercial
routes have been proposed to carry data from ground users along their flying
routes. Delivery probability is compared when using aircrafts, buses and ferries.
Results show that aircrafts provide higher delivery probability which is up to
62% better compared with buses and ferries. Furthermore, when there is lack of
resources, it is difficult to satisfy all users’ demands for traffic. We propose a
Fairness and Satisfaction (FS) model to enhance the users' satisfaction during
DTN limited and scarce resources. Various scenarios are tested for the FS
model through intense simulations. FS model, when implemented, will improve
user’s satisfaction up to 18 % and DTN fairness up to 20 % compared with the
same scenario lacking the model.
Keywords: Delay Tolerant Network (DTN), Quality of service (QoS), Aircraft
and User’s Satisfaction.
1 Introduction
Delay tolerant networks work in the environment of limited communication
infrastructures where links are disconnected and might not be available for a long
time. In addition to using the standard transmission medium in these challenged
conditions, transportation means have been used in DTN literature to carry data for
DTN applications. Buses, ferries, bicycles, motorbikes and trains have been proposed
for DTN[1][2][3][4], sometime by physically transporting the saved data on a storage
medium such as USP memory or by wirelessly exchanging data, whenever a
transportation vehicle comes in range with the users. We have proposed using
aircrafts in commercial routes [5] because they have larger coverage area and fly over
remote areas where communication infrastructure is limited. In this paper, we will set
a scenario and compare between using buses, ferries and aircrafts. The results show
that using aircrafts will provide better delivery probabilities and more user
satisfaction. This is mainly because all requesting users are in range with the flying
aircraft at most of the time due to the flying high altitude whereas this is not the case
with buses and ferries.
Furthermore, these transportation means are considered as contacts and represent
the main resources in DTN. The more number of vehicles will mean more
opportunities to send data and therefore more resources. However, these resources are
normally scarce and limited in DTN which might degrade its ability to satisfy all
users’ requests of traffic. The users will compete with each other to get access to the
passing contact and this will lead to congestion. Whenever a contact exists, all users
will try to get access to it in order to transmit their data but unfortunately not all of
them will be able to send theirs due to the contact limited capacity and numbers.
However, after another wait, another contact might arrive, where the same
competition will be repeated among users and again not all of them will be able to
send their data. Therefore, there is a need for a system that should try to improve each
and every user chances of getting access to the contact resources especially during
severe shortage of resources. The system should be fair to allocate the resources
equally among the users and give a chance to as many users as possible to send their
data.
Throughout this paper, we will show the improvement in the level of user
satisfaction when applying the FS model compared with the same scenario lacking
our model. FS model will be applied to aircrafts, buses and ferries which act as
intermediate node/gateway to help it decide fairly which user to select. This decision
will be used by other vehicles to ensure that they satisfy as many users as possible.
2 Related Work
DTN has increasingly become a hot research topic and the IRTF DTNRG [6] is
leading the effort to expand the research of DTN in various topics. In DTN literature,
many transportations means are used to carry data for DTN applications. We
proposed using an aircraft as a DTN carrier at remote locations [5]. At UMass
Amherst [2], they used public buses travelling on scheduled routes to exchange data
using Wi-Fi nodes in a project called Diesel Net. Ref [3] proposed using ferries and
suggested that the users could schedule their movement to match the ferries
movements in order to send their data. We will compare between these three types of
transportations in DTN environment in the coming sections.
However, there are limited references that discuss the user satisfaction in DTN. Ref
[7] proposed a data delivery method for DTN for various users based on the relative
importance of their data requests, which is something like the users class of service
CoS classification in DTN. They proposed a carry-and-forward technique to enhance
the overall user satisfactions and achieved better users satisfactions compared with the
conventional first in first out FIFO method. Our FS model will also consider the class
of service CoS of the user, but furthermore, it will consider other metrics such as the
history of users previous admissions as we shall see later in the model discussion.
Furthermore, many references in the literature discuss the resources management in
DTN but each one is targeting a certain resource. Some of them discuss how to save
the buffer resources which is important for some DTN applications like the sensors
network. The concept of message ferrying is proposed in [8] where they proposed
allocating the buffer resources between users at different connections sessions. In our
aircraft application, we assume unlimited buffer space and we consider the aircraft
communication bandwidth as the only resource. The bandwidth will increase with the
longer time the aircraft stay in line of sight contact with the ground users. Also, more
aircrafts will mean more contacts and therefore more resources. Ref [9] also discusses
the resource management using adaptive optimal buffer management scheme for
various message sizes. It implements a dropping policy to enhance the delivery rate
and reduce the delivery delay. Our FS model will decide which users to accept by
managing the bandwidth resource of the aircraft.
Many references of DTN resource management are in the form of new routing
algorithms while our model is a complementary mechanism applied to the routing
algorithms. The Probabilistic Routing Protocol using History of Encounters and
Transitivity (PRoPHET) [10] eliminates the production of replications based on
history encounters in order to enhance the resource utilizations. Flooding is used in
Epidemic routing [11] where the nodes replicate and transmit the bundles to all other
nodes which do not have a copy of those bundles and this replication consumes many
of the network resources. Spray & Wait routing algorithm [12] sprays a number of
copies into the network and waits for at least one of these messages to arrive at
destination. The prioritizations concept is implemented in MaxProp [13] and is based
on scheduled transmitted and dropped packets where the historical data of past
meetings is used to decide the possibility of future meetings. MaxProb routing will be
used in our scenario because it the best routing algorithms among the other three to
provides better performance.
FS model can be applied to any DTN routing algorithm and we argue that it can
provide extra improvements in the performance as we shall see in the coming
sections.
3 Fairness and Satisfaction Model Concept
Routing can be considered as a data push operation where each node tries to
forward its data bundles to other nodes through various routes, which ensures faster
and optimum performance of data delivery from a source to a destination. In the case
of DTN, the process will be a store and forward one because the next nodes might not
be available for some time. On the other hand, FS model can be considered as a data
pull operation where the node decides which one to admit first and which to reject,
from a group of interested transmitting nodes.
As shown in Fig.1, which shows pull/push concept, there are 5 users (1 to 5 at area
X) who want to send/receive data from/to 5 other users (6-10 at area Y). There are 2
aircrafts (AA and CC) in direct contact with the users while 2 other aircrafts (BB and
DD) acting as intermediate nodes. The aircrafts are considered mobile intermediate
routers/gateways for the users. FS model is applied at those aircrafts that while
receiving the data, they can apply the model to select which to accept among the
interested users. This is considered as a pull operation. Now the aircraft wants to
forward the data using the routing algorithm to the next contact. This is considered a
push operation. For example, aircraft AA applies the pull concept (FS model) from
users 1,2,3,4 and 5, while it applies the push concept (Routing) to forward it to
intermediate aircrafts BB and DD.
Fig. 1. Push/Pull concept of the FS model
FS model will apply fairness during the selection process and for that we will use
Jains fairness index [14] to evaluate the model performance. Jains index will evaluate
the fairness of distributed resources for a group of users competing for a common
bottleneck channel at a particular time.
3.1 FS Model Design
FS model is working based on collecting feedback administration metrics data
from all users of their class of service CoS, Time of their request, number of past
acceptances and number of past rejections. FS model will assign a weight value for
each metric and decide which user to admit first based on the highest metrics weight.
Each contact will use the latest feedback administration data for his decision and these
data will be updated after each contact session. The model will allocate the whole
bandwidth to users one after another and will not distribute it among users. During
scarce resources, if the contact carries a fragment from all users, it might end up - in
the worst cases- that none of them is able to process his data because they all haven’t
received all their fragments. Therefore, FS model is providing fair and better
resources allocation using this concept which is more useful at scarce resources only.
3.2
QoS Contact-to-Contact /End-to-End
There are requirements to provide end-to-end QoS and hop-to-hop QoS in
traditional networks. However, in DTN the links are not guaranteed and the next hop
arrival might not be known and there might be other local hops for the node within its
area. Therefore, providing this concept might be difficult and impractical in DTN
environments. Instead, we think the QoS in DTN disconnected environment should be
provided per contacts because it is the main resource. Whenever a contact becomes
available, it will implement the QoS requirements through collaborations with other
contacts by referring to the past data recovered from the nodes. This way, providing
the QoS contact-to-contact in DTN is more visible and this can be illustrated in Fig. 1
as users 1 to 5 and aircraft AA or users 6-10 and aircraft CC. Furthermore, end-to-end
QoS can be provided based on the performance of the routing algorithms and the
availability of contacts and this can be illustrated in Fig. 1 as users 1 to 5 and users 610 exchanging data using intermediate aircrafts AA, BB, CC and DD.
We will have two sets of results. The first is contact-to-contact which considers the
average results of all contacts applying the model within a group of users’ nodes and a
particular contact. This set will evaluate the FS model performance results with
respect to the satisfied users and fairness. The second set of results will be based on
the end-to-end satisfied users and fairness from source to destination. It might
consider multiple contacts and the routing algorithms among nodes and contacts.
4 Simulation Results
We will evaluate the performance of using aircrafts compared with using buses and
ferries. Furthermore, we will evaluate the performance of FS model.
4.1
Simulation Scenario
The scenario assumes three remote locations, A, B and C where each area has a
group of users. These locations have limited communication infrastructure and they
use all types of available links including the transportation means of buses, ferries and
aircrafts. As shown in Fig. 2, the areas are at the shore of a big lake where buses,
aircrafts and ferries operate between the three locations. Whenever any of the vehicles
(bus or aircraft or ferry) approaches the area, the ground user will try to send their
data to the passing vehicle which in turn will deliver it to the next area and take other
data to another area and so on.
Fig.2. FS Model Senario
4.2
Simulation Configuration
Fig. 3. Scenarios configurations in ONE simulator
We have used Opportunistic Networking Environment (ONE) simulator [15] which
is a DTN simulator platform for testing the routing and applications protocols. FS
model is configured in ONE simulator and the cluster movement model is used for the
model scenario as shown in Fig. 3 which illustrates the simulator GUI. The scenario
assumed three locations, i.e. A, B and C, which appears in Fig.3 as the group of dotes.
The transportation vehicle, which appears in Fig.3 as the circle in route between the
clusters locations, is moving between them and collecting data from ground user
nodes to send them to other nodes in another area. Every time the vehicle passes by
cluster nodes, the decisions of which nodes are selected or rejected will be updated
and used by the next vehicle to enable it to take acceptance decision and so on until
all vehicles pass throughout the set simulation time.
Simulation time is in seconds and it is assumed that a full day of 24 hrs is
equivalent to 5000 seconds simulation time. 24 hrs is chosen because the
transportation means normally have daily schedules of their journeys routes which
will be repeated every day. All users and vehicles have equal data rate and ground
users create and try to send messages only when the vehicle contact is their LOS
range. It is also assumed that all transportation means are in continuous movements in
their routes for the whole day and they only stop at the end of the day.
Table 1. Simulator ONE settings
Transportation
Mean
Max Real Speed
(Km/hr)
Bus
120
Aircra
ft
900
Ferry
Equivalent
Simulation speed
(m/s)
Real Wait Time
(min)
Equivalent
Simulation Wait
Time (s)
Range
Km
33.3
250
60
(32.4
knot)
16.6
5
30
10
17.35
104.1
34.7
5
100
5
All scenarios are run for 5000 seconds which represent 24 hrs in the scenario time.
The scenarios are set for 20 users distributed randomly per location and 2 vehicles for
each transportation type. Table 1 shows the setting of the various transportations
means. Different vehicle speeds are used where the aircraft is the highest of 900
km/hr and the ferry is the lowest with 60 km/hr. Transportations vehicles are set to
stop and wait at each cluster, i.e. the duration time they are in LOS with the ground
users. Aircrafts have longer wait time because they fly at high altitude of around
11000 m; therefore, they will have longer contact time and larger coverage range.
Each real value will have an equivalent simulator value based on the assumption that
24 hr is equal to 5000 seconds. The performance metric is the delivery probability,
user’s satisfactions, average processing delay and the fairness. It will be used to
evaluate the performance of FS model for various scenarios.
4.3 Simulation Results
The simulator is run for various simulation times. From 1000 to 5000 seconds and
the results are obtained as follows:
4.3.1 Transportation Means Comparison
Fig. 4 shows the obtained delivery probability results for a scenario of 2 vehicles
and 20 users per location when FS model is enabled. The delivery probability is the
overall delivered messages out of the generated one regardless of the users. It shows
that aircrafts have higher delivery probability between 85-95 % compared with buses
and ferries whose values range between 20 to 30%. Overall, aircrafts are having up to
62 % improvements over buses and ferries. This is due to that all users are within the
aircraft range; therefore, all have better opportunities to send their data. On the other
hand, buses and ferries have smaller range because they are at ground level; therefore,
not all the users are in their range. Furthermore, the wait time i.e. contact LOS, is
larger with the aircrafts because they are at high altitude and therefore can stay longer
time over the ground users to enable them to send more data, while the buses and
ferries are staying shorter time and they become out of range. This shows the potential
of using aircrafts in this special situation and its usefulness at remote locations.
Fig. 4. Delivery probabilty of various transportation means
4.3.2 FS Model Comparison
The previous section compared the results of aircrafts with ferries and buses when FS
model is implemented. However, in this section we will compare the results between
the three transportation means with and without FS model.
4.3.2.1
Contact-to-Contact Results
The first set of results is between the ground users and the approaching contacts.
The results are obtained for users’ satisfaction within contacts. It does not consider
the end-to-end routing of data or delivering the data to its final destination. All it
considers is the transfer of the user’s messages to the passing contact. Users
satisfactions is a measure of how many users successfully delivered all their generated
messages out of the overall available users. Fig. 5 shows user satisfaction
improvements in applying FS model compared with the lack of it for the three
transportation means which varies between 2- 18%. Transportation type has no major
impact on the model results because the outcome will mainly depend on how long the
contact is available for the ground users. As mentioned earlier, messages are
generated during contact time only. Therefore, buses have the best user satisfaction
because they have the shortest wait time and therefore have less number of generated
messages per user, which increases the user’s chances of acceptance. Aircrafts have
the second best user satisfaction level because they have the highest speed which
means more wait time occasions for the ground users.
Fig. 5. Contact-to-contact users satisfactions
Fig. 6. Contact-to-contact fairness index
Fig. 6 shows Jains fairness index results for allocation of the available resources.
The results show improvements in the fairness between 5-20 % compared with the
lacking of FS model. Buses and ferries have better fairness results because their users
generated fewer messages during the same simulation time compared with aircrafts.
However, FS model provides fairness improvements regardless of the transportation
means used.
4.3.2.2
End-to-End Results
Fig. 7. End-to-end users satisfactions
Fig. 8. End-to-end fairness index
The second set of results applies for the end-to-end concept which utilizes the
routing to deliver messages to their final destinations. MaxProp [13] DTN routing is
used to route the messages for the three transportation means. The obtained results in
contact-to-contact section are relatively close to each other between the three
transportation means. However, it is not the case with the end-to-end results which
indicates a very high priority for using aircrafts compared with buses and ferries.
Fig. 7 shows the user’s satisfaction of aircrafts reaching up to 90 % and
improvement varies between 5-18 % compared with the lack of FS model. The
improvements are higher at the early runs of simulation. On the other hand, buses and
ferries have very limited user satisfaction, less than 10 %. These results come to an
agreement with the results of the overall delivery probability shown in Fig. 4, but the
difference is that Fig. 7 results are per users who delivered all their messages, while
Fig. 4 is per overall generated messages regardless of who generated them. Aircrafts
have provided higher users satisfactions due to their large coverage which can reach
all users, while this is not always the case with buses and ferries. Furthermore,
applying FS model has improved this satisfaction by providing better allocations of
resources for more users.
Fig. 8 shows the improvments in end-to-end fairness index for the various
transportation means and FS model has improved the fairness for the end-to-end also
by 5-18 % compared with the lack of it.
Fig. 9. End-to-end average processing delay
On the other hand, there is a small drawback for the model represented in the slight
increase in the processing average delay which is the average time in seconds between
message creation and delivery. Applying FS model will sometimes cause the delay to
increase between 0 second in the best cases to 100 second in the worst case. Aircrafts
have the smallest amount of delay due to their high speed and therefore will mean
quicker delivery time and hence less processing time. It is not a very critical increase,
and DTN is delay tolerant and hence the extra delay is considered insignificant.
5 Conclusions and Future Work
Various transportation means have been proposed for carrying DTN data due to the
scarce and limited resources of its environments. Aircrafts in commercial routes are
compared with buses and ferries. The simulation results showed higher delivery
probability for using aircrafts due to their large coverage range. Furthermore, FS model
is proposed to allocate the resources among users in DTN conditions. The model
showed improvements in user’s satisfactions and better fairness compared with the
same scenario lacking the model. However, FS model will sometimes cause a slight
increase in the average processing delay which is not very significant in DTN situation.
As part of our future work, we will implement the aircraft concept together with the
FS model in various and more complicated scenarios for a larger number of vehicles
and users to analyze its performance.
References
1. S. Farrell and V. Cahill, Delay-and Disruption- Tolerant Networking. London: ARTECH
HOUSE, Ch3, p 65, 2006.
2. X. Zhang, J. Kurose, B. N. Levine, D. Towsley, and H. Zhang, “Study of a bus-based
disruption-tolerant network: mobility modeling and impact on routing,” in Proceedings of
the 13th annual ACM international conference on Mobile computing and networking, pp.
195–206, 2007.
3. W. Zhao, M. Ammar, and E. Zegura, “A message ferrying approach for data delivery in
sparse mobile ad hoc networks,” in Proceedings of the 5th ACM international symposium
on Mobile ad hoc networking and computing, pp. 187–198, 2004.
4. S. Jain, K. Fall, and R. Patra, “Routing in a delay tolerant network,” in Proceedings of the
2004 conference on Applications, technologies, architectures, and protocols for computer
communications, pp. 145–158, 2004.
5. M. A. AL Siyabi, Cruickshank, H, and Z. Sun, “Delay/Disruption Tolerant Network
Architecture for Aircraft Datalink on Scheduled Routes,” presented at the PSAT10, Rome,
Italy, PSAT10, 2010.
6. “Home - Delay Tolerant Networking Research Group.” [Online]. Available:
http://www.dtnrg.org/wiki. [Accessed: 20-Jan-2011].
7. Yasuhiro Ishimaru, Weihua Sun, Keiichi Yasumoto, and Minoru Ito, “DTN-based Delivery
of Word-of-Mouth Information with Priority and Deadline,” The Fifth International
Conference on Mobile Computing and Ubiquitous Networking (ICMU 2010), 26-Apr2010.
8. Y. Li, M. Qian, D. Jin, L. Su, and L. Zeng, “Adaptive Optimal Buffer Management Policies
for Realistic DTN,” in GLOBECOM 2009 - 2009 IEEE Global Telecommunications
Conference, pp. 1-5, 2009.
9. M. Chuah and W. Ma, “Integrated Buffer and Route Management in a DTN with Message
Ferry,” in MILCOM 2006, pp. 1-7, 2006.
10. T. Huang, C. Lee, and L. Chen, “PRoPHET+: An Adaptive PRoPHET-Based Routing
Protocol for Opportunistic Network,” in 2010 24th IEEE International Conference on
Advanced Information Networking and Applications, pp. 112-119, 2010.
11. A. Vahdat and D. Becker, Epidemic routing for partially connected ad hoc networks.
Citeseer, 2000.
12. T. Spyropoulos, K. Psounis, and C. S. Raghavendra, “Spray and wait: an efficient routing
scheme for intermittently connected mobile networks,” in Proceedings of the 2005 ACM
SIGCOMM workshop on Delay-tolerant networking, pp. 252–259, 2005.
13. J. Burgess, B. Gallagher, D. Jensen, and B. N. Levine, “Maxprop: Routing for vehiclebased disruption-tolerant networks,” in Proc. IEEE Infocom, vol. 6, 2006.
14. Jain, R., Chiu, D.M, and Hawe, W, “A Quantitative Measure of Fairness and
Discrimination for Resource Allocation in Shared Systems,” DEC Research Report TR301, 1984.
15. A. Ker\änen, J. Ott, and T. K\ärkk\äinen, “The ONE simulator for DTN protocol
evaluation,” in Proceedings of the 2nd International Conference on Simulation Tools and
Techniques, pp. 1–10, 2009.