Multicasting in Delay Tolerant Networks: Implementation and Performance Analysis

CS6250 PROJECT, GEORGIA INSTITUTE OF TECHNOLOGY, DECEMBER 12, 2016 1
Multicasting in Delay Tolerant Networks:
Implementation and Performance Analysis
Harsh Maniar, Neha Milind Raje, Nagendra Posani, Swarnim Vyas, Venkatasivaram Charan B
Abstract—Delay Tolerant Networks(DTN) are a class of emerg-
ing networks which experience intermittent connectivity and lack
end-to-end paths due to absence of well-defined infrastructure.
In this paper we explore the nuances of multicasting in DTNs.
Multicasting enables efficient distribution of messages to a group
of users, a paradigm that can be applicable in the context of
DTNs. While multicasting in internet and ad-hoc networks has
been studied extensively, realizing the same in DTNs is non-
trivial given that many factors have to be considered. This paper,
presents an implementation of multicast routing for various
protocols in DTNs using ONE simulator. It also provides the
analysis and performance results for the various protocols studies
against the different movement models.
Keywords—Group based communication, routing protocols, mul-
ticast, ONE simulator, movement model
I. INTRODUCTION
Delay tolerant networks are named so since the connections
between nodes are intermittent and hence lack instantaneous
end-to-end sender-to-receiver paths. This is particularly appli-
cable to systems that suffer disruption in infrastructure leading
to huge delays in message delivery. There can be many real-
life applications of delay tolerant networking like emergency
response, military systems, vehicular systems and deep-space
communication.
In such scenarios, there is no network infrastructure to
establish an end-to-end communication path between two
nodes. The messages have to be relayed via intermediate
nodes that hold on to the message till they meet another
node. Distribution of a message to a group of recipients can
be useful in such an environment. For example, in a disaster
recovery scene, it is vital to disseminate information about
the victims and potential hazards among the rescue workers
where there is no central arbiter. In battlefield, soldiers in
a squad need to inform each other about their surrounding
environment.
To disseminate messages in an efficient manner, various
protocols are defined for unicast routing in DTNs. This
paper studies a subset of various protocols for delay tolerant
networks: First Contact, Direct Delivery, Epidemic, Spray-
and-Wait and PRoPHET. A property that sets these routing
protocols apart from the ones widely applicable in connected
networks is that they use a store and forward approach to
transmit the message to the intended end recipient. These
protocols have been chosen in such a manner that few of them
are single copy protocols - first contact and direct delivery,
Thanks to Prof. Mostafa Ammar for guiding us through out the project
epidemic uses flooding as the underlying approach, spray and
wait can be classified as controlled flooding and PRoPHET is
the probability based protocol that uses delivery predictability
as a metric.
Various approaches can be taken to enable and implement
multicasting in the protocols discussed. A unicast based
multicast is the most straightforward way of doing it, but
can have potential issues. Other approaches include having
an Oracle or a logical group based identifier. Based on the
advantages offered by the logical unique identifier approach,
it was chosen and multicast in delay tolerant networks was
implemented for various protocols. Implementation was
done using the ONE(Opportunistic Network Environment)
simulator which has been designed for emulating delay
tolerant networks and provides support for various protocols
and mobility models. Leveraging the power of ONE, we
analysed and evaluated various scenarios to measure the
performance of the protocols based on different parameters.
The rest of the paper is organized as follows, section 2
examines the various protocols in the context of delay tolerant
networks, section 3 describes the approach taken for enabling
multicast in DTN, section 4 provides the implementation and
evaluation details, section 5 describes performance analysis
along with the simulation results and section 6 states the
conclusions drawn and future work.
II. PROTOCOLS
A. Epidemic
Epidemic routing [1] is flooding-based in nature, as nodes
continuously replicate and transmit messages to newly dis-
covered contacts that do not already possess a copy of the
message. In this protocol, messages are quickly distributed
through connected portions of the network. The overall goal
of Epidemic Routing is to maximize message delivery rate and
minimize message delivery latency, while also minimizing the
aggregate system resources consumed in message delivery.
B. Spray and Wait
Spray and Wait [2] combines the speed of epidemic routing
with the simplicity and thriftiness of direct transmission. Spray
and Wait routing decouples the number of copies generated per
message, and therefore the number of transmissions performed,
from the network size. It consists of two phases:
• Spray phase: for every message originating at a source
node, a fixed number(L) of message copies are initially

spread forwarded by the source and possibly other nodes
receiving a copy to L distinct relays.
• Wait phase: If the destination is not found in the spraying
phase, each of the L nodes carrying a message copy per-
forms direct transmission i.e. will forward the message
only to its destination.
The source of a message initially starts with L copies; any
node A that has n (n being greater than 1) message copies
(source or relay), and encounters another node B (with no
copies), hands over to B n/2 copies and keeps n/2 copies for
itself; when it is left with only one copy, it switches to direct
transmission. This called Binary Spray and Wait routing. Spray
and Wait is simple but outperforms all flooding based schemes
with respect to number of transmissions and delivery delays.
It also achieves comparable delays to an optimal scheme, and
is very scalable as the size of the network or connectivity level
increases.
C. First Contact
In this single copy routing protocol [2], node transmits the
message node randomly using any available contact. Hence,
only a single copy of every message in the network. If none of
the nodes are available, then the message waits for a particular
path until become available and is assigned to first available
contact. There is no guarantee that the first node that is met is a
better candidate than the previous node carrying the message,
so First Contact does not exactly maximize on the reachability
criteria.
D. Direct Delivery
The source only transmits the message if it comes into
contact with the destination. The set of relays contains only
the destination hence this is the degenerate case of the
flooding family. It is simple because it does not consume
many resources, and it uses exactly one message transmission.
However, it only works if the source contacts the destination
[3].
E. PRoPHET
Probabilistic Routing Protocol using History of Encounters
and Transitivity (PRoPHET) [4] protocol uses an algorithm
that takes advantage of the non-randomness of real-world
encounters. It maintains a set of probabilities (delivery pre-
dictability) for successful delivery to known destinations in
the DTN. The delivery predictability indicates how likely
it is that a node will be able to deliver a message to the
destination node. When two nodes meet, the summary vectors
are exchanged. The summary vectors contain the delivery
predictability information stored at the nodes. This information
is used to update the internal delivery predictability vector, and
then the information in the summary vector is used to decide
which messages to request from the other node based on the
forwarding strategy used. The messages are forwarded during
opportunistic encounters only if the node that does not have
the message appears to have a better chance of delivering it.
Hence with time, there seems to have formed a virtual path
through the nodes that has the best delivery probability.
III. MULTICAST ROUTING IN DTN PROTOCOLS
As discussed, there are multiple applications that may
benefit from a group based message delivery approach in
the realm of delay tolerant networking. Communicating to a
predefined subset of nodes can be made possible in various
ways. Establishing effective group based communication in
DTNs and evaluating it is the main motivation of the work.
Multicasting can be achieved using a simple unicast
based approach the sender can create a separate copy of
the message for each user and send it using the unicasting
based nethodology. This can highly overwhelm the network
specifically in the context of flooding based protocols as
multiple copies of the same message with different destination
nodes (but of same group) exist in the network at the same
time. This can lead to congestion and high network overhead,
hence is highly inefficient.
An Oracle based approach [5] for multicast is one where in
there is a central knowledge base that maintains the mapping
between the groups and the members of the group. Each time
a node is encountered, a query would be needed to be made
to the Oracle to determine if the node encountered belongs
to the targeted group so as to determine if the message needs
to be transmitted. This additional call to the Oracle for every
transmission can be an overhead which can be avoided by
using the approach discussed next.
Another way to achieve this is by having each message
carry the list of targeted recipients of the message and on
every encounter with a node iterate through the list to check
whether the node encountered belongs to the group list and
take further actions accordingly. This approach is reduces the
overhead of the calls made to the Oracle, but can definitely
be improved in terms of performance by slightly changing
the way group lists are defined in a node.
The final one that was chosen to be implemented is also the
most optimum one amongst those discussed here. Each node
carries information about its group identifier and whenever it
meets another node it takes routing decisions based on the
intended recipient of the message and the group identifier of
the other node. This logical identifier uniquely identifies a
group in a set of nodes. The action(s) which each node takes
on an encounter is defined by the protocol being implemented
as discussed above. This way addition of a node from a group
is as simple as assigning the specific group identifier to the
newly added node. The message headers can now remain the
same since the destination need not be a list of nodes, but a
single logical group identifier.
The paper describes next how the ONE simulator was used
to implement the aforementioned multicasting approach.
IV. EVALUATION METHODOLOGY
ONE Simulator: The ONE stands for Opportunistic
Network Environment simulator. It enables generation of
node movement using different movement models. ONEs

simulation environment is capable of routing messages
between nodes with various DTN routing algorithms, and also
between different sender and receiver types. Visualizing both
mobility and message passing in real time is also possible
using its graphical user interface. ONE can import mobility
data from real-world traces or other mobility generators. It
can also produce a variety of reports from node movement to
message passing and general statistics.
The ONE simulator provides support for modifying a lot
of parameters that govern the delivery of the message from
source to destination. The parameters that were played with
for evaluating our implementation of multicast routing are
elaborated as follows.
A. Protocol
The implementation of 5 of the routing protocols in ONE
simulator for DTN, namely Epidemic, First Contact, Direct
Delivery, Spray and Wait and PRoPHET were modified so as
to be extended for multicast routing support. These protocols
were chosen to cover the most important classes of DTN
routing protocols single-copy, n-copy and unlimited-copy pro-
tocols, as well as estimation based protocols. Since a protocol
controls the delivery of the message to the node, path to reach
the node, and how many copies of the message are forwarded,
thereby affecting the congestion in the network, the choice of
protocol greatly influences our results.
B. Movement model
Movement models define the algorithms and rules that
generate the node movement paths. The movement models are
either random movement based, map-constrained movement
based and human behaviour movement based. To evaluate the
interplay between movement models, we chose the following 3
models: Random Waypoint movement, Map Based movement
and Cluster movement. Since the movement model dictates the
mobility of nodes in the network, simulation results depend
directly on the nuances of the model chosen.
C. Number of nodes
The number of nodes determine how dense the network
is. More the number of nodes, greater is the probability of a
message getting delivered (as there are more message carriers)
- for most of the protocols. We shall see the interesting cases
in the results section. Also, more number of nodes in a group
within a group and more messages are delivered on average.
The only trade-off is the number of duplicate messages in the
network and the increased congestion in the network.
D. Number of groups
The simulation was performed for 2 groups multicasting
simultaneously. Hence messages of both the chosen groups
were routed in parallel with each other. From the group types
supported by ONE, the groups chosen for our simulation were
those of pedestrians and cars. In the cases where we have just
one group, we have used cars as our default group.
E. TTL
TTL is the time to live, or the maximum number of nodes
a message can be sent to before it is dropped. A higher TTL
value means that a message can now traverse more number
of nodes, thereby positively affecting the reachability of the
message and hence the percentage of messages received in a
group. More number of destination nodes end up receiving
the message if TTL value is increased. Again, as in the
case of number of nodes, the improvements come at the cost
of increased congestion and duplication of messages in the
network.
F. Warmup Time
This parameter only affects probability based protocols, i.e.
PRoPHET. Such protocols need an initial warmup time to
interact and develop the probabilities of a node reaching any
other node in the network, so that a forwarding node can
choose the node with highest probability of communicating
with the destination. Higher the warmup time, more developed
are the probabilities. This implies that the the percentage of
messages received depends directly on warmup time.
V. RESULTS
The comparative analysis of various protocols in the
context of different movement models provides a yardstick
for choosing the appropriate routing protocol for a given
scenario. The results from of each algorithm is detailed in the
below sections.
Number of nodes and Protocols: Increase in the number
of nodes has a direct impact on the percentage number
of messages delivered. As can be seen in Figure 1, where
for most of the protocols, the percentage of the messages
delivered increases with increase in number of nodes. This is
true for most of the protocols, except for Direct Delivery and
PRoPHET. In case of direct delivery, since the protocol aims
to deliver the message directly to the destination, increase in
the node density does not necessarily help the source node
reach the destination node directly - this results in a flat line
for the protocol. We see a similar behaviour in PRoPHET
- irrespective of number of nodes, knowledge base with the
nodes remains same as we have used the same warm-up time
for all the simulations. Hence the reachability probabilities for
the nodes remain fairly constant leading a similar percentage
of messages delivered even in the case of increased node
density.
Groups: In the simulations seen till now, we have used
only one group as a reference. Figure 2 shows the results
when two groups are multicasting parallely in the simulation.
The overall picture gives the impression to be similar to the
one group scenario wherein we just have x+y number of
nodes in the multicast group. In this specific case the two
node kinds being cars and pedestrians.
Latency: Increase in the number of nodes decreases the
average latency for protocols that are flooding based since they

Fig. 1. Percentage of messages delivered vs. nodes in Random Waypoint
Fig. 2. Percentage of messages delivered vs. nodes in Map Based Movement
Fig. 3. Average latency vs. nodes in Map Based Movement
Fig. 4. Average HOP count vs. Nodes in Map Based Movement
Fig. 5. Average HOP count vs. Nodes in Random Waypoint
Fig. 6. Percentage of messages delivered vs. TTL in Random Waypoint
Fig. 7. Average Latency vs. Warmup Time in Map Based Movement

are able to utilize the node density to their advantage(Figure
3). For Direct delivery and PRoPHET both of which depend
on the reachability path to the final destination, increase in
number of nodes increases the latency since more nodes add
noise to the network. For First Contact, the latency again
increases since there are now more nodes that need to relay
the message that are between the source and destination.
Average HOP Count: In Figure4 & 5, we see the effect
of various mobility models on the hop count. Increase in
number of nodes increases the hop count for flooding based
protocols. For path based protocols like PRoPHET and direct
delivery, there is only a slight increase.
Time To Live: Increase in the TTL value results in the
increase in the percentage of messages delivered. Flooding
based protocols keep propagating the messages till the TTL
value expires or the message gets dropped. Hence a direct
correlation exists. But for nodes that might be partitioned in
the network in such a manner that they are never reachable
given the mobility model, further increase in TTL will not
help. In this Fig-6, if we were to increase the TTL beyond
200, the percentage of messages delivered does not increase
indicating a saturation. Direct Delivery and PRoPHET are
unaffected by TTL values since the region in which the
multicast message is reachable still remains the same and no
matter how long the message can hop around, it might not be
able to reach a sparsely connected destination region where
the node is present.
Warmup Time: An optimum value of warm up gives best
results for the PRoPHET protocol. It can be seen in Figure
7 that the increase in the number of nodes decreases the
average latency since now we have a more well defined path
between the source and the destination. For simulations where
the warm-up time was kept at zero, it took much longer for the
messages to reach the destination since clearly defined paths
were not available for the same.
VI. CONCLUSIONS AND FUTURE WORK
Effective group based communication in real time scenarios
can be utilized in various applications in delay tolerant
networks. Having implemented the multicast functionality for
various protocols in ONE simulator has helped us analyze the
results of the simulations in various scenarios under varying
movement models and carrier nodes. The results are highly
dependent on the scenario in which particular protocol runs
and there is no one single protocol which stands out in
terms of all performance metrics mentioned in the evalutation
section.
Probability based protocols can be very effective in
scenarios where the knowledge base of each node to
determine the probability of meeting every other node is
accurate. PRoPHET exhibited overall decreased latency when
it was given sufficient warm up time to establish the accurate
knowledge. In many scenarios where there is no sufficient
time to establish this knowledge base the performance of
PRoPHET exhibits no significant added advantage. Scenarios
like that of emergency where there is no sufficient time
to establish knowledge base this protocol may not work
optimally. However, in cases where this knowledge base can
be established, like in case of battalion of Army, PRoPHETs
performance could be utilized for effective communication.
Flood based protocols are highly reliant methods that
make sure the message goes through but they can highly
overwhelm a network and in real life scenarios this can lead
to congestion and eventually communication failure for all the
nodes. Controlled flooding can be an optimisation over such
an approach to control the network traffic. Direct Delivery
and First Contact can also be used depending on the need
of the application and urgency of the message to get delivered.
As a part of what can be done next, one may consider
incorporating the dynamic nature of node for the purpose of
analysis. A node can changes its nature (group id) within
a single simulation. This may affect the performance metric
severely specially in cases of probability based protocols which
determine their knowledge with a single target node group
in mind. Also one could imagine scenarios where few nodes
belong to more than one group and its effects could be
observed by simulating such scenarios once implementation
support for the same has been provided.
VII. ACKNOWLEDGMENT
We would like to thank Professor Mostafa Ammar for
guiding us through the project. We would also like to thank
Karim Habak for the valuable advise at times of perplexity.
We also would like to thank Barun Saha and Sahil Gupta for
their detailed blogs on ONE simulator.
REFERENCES
[1] Amin Vahdat and David Becker, Epidemic Routing for Partially-
Connected Ad Hoc Networks. Duke Technical Report CS-2000-06
[2] Monika Aneja and Vishal Garg Simulation of Epidemic, Spray and Wait
and First Contact Routing Protocols in Delay Tolerant Network, IOSR
Journal of Electronics and Communication Engineering, 2015.
[3] Evan P.C. Jones and Paul A.S. Ward Routing Strategies for Delay-
Tolerant Networks, University of Waterloo Survey 2007.
[4] Anders Lindgren, Avri Doria and Olov Schelen Probabilistic Routing in
Intermittently Connected Networks, Electronics and Telecommunications
Research Institute (ETRI).
[5] Wenrui Zhao, Mostafa Ammar, Ellen Zegura Multicasting in delay
tolerant networks: semantic models and routing algorithms, WDTN ’05
Proceedings of the 2005 ACM SIGCOMM
[6] Kernen, A., Ott, J., Krkkinen, T. The ONE simulator for DTN protocol
evaluation, Proceedings of the Second International ICST Conference on
Simulation Tools and Techniques - 2009.
[7] Chuah, M. C., Yang, P. Context aware multicast routing scheme for
Disruption Tolerant Networks. International Journal of Ad Hoc and
Ubiquitous Computing - 2009.

Multicasting in Delay Tolerant Networks: Implementation and Performance Analysis

Related slideshows

More Related Content

Multicasting in Delay Tolerant Networks: Implementation and Performance Analysis