A SIMULATION-BASED PERFORMANCE COMPARISON OF MANETS CDS CREATION ALGORITHMS USING IDEAL MAC AND IEEE 802.11 MAC

Sundarapandian et al. (Eds) : ICAITA, SAI, SEAS, CDKP, CMCA-2013
pp. 55–64, 2013. © CS & IT-CSCP 2013 DOI : 10.5121/csit.2013.3806
A SIMULATION-BASED PERFORMANCE
COMPARISON OF MANETS CDS CREATION
ALGORITHMS USING IDEAL MAC AND IEEE
802.11 MAC
Khalid A. Almahorg1
, Mohamed S. Elbuni2
, Elmahdi M. Abousetta3
,
Ahmed Arara4
1
Department of Computer Engineering, University of Tripoli, Tripoli-Libya
khalid.almahrog@gmail.com
2
mselbuni@gmail.com
3
drelmahdiabousetta@gmail.com
4
ahmedarara@yahoo.fr
ABSTRACT
Mobile Ad Hoc networks (MANETs) are gaining increased interest due to their wide range of
potential applications in civilian and military sectors. The self-control, self-organization,
topology dynamism, and bandwidth limitation of the wireless communication channel make
implementation of MANETs a challenging task. The Connected Dominating Set (CDS) has been
proposed to facilitate MANETs realization. Minimizing the CDS size has several advantages;
however, this minimization is NP complete problem; therefore, approximation algorithms are
used to tackle this problem. The fastest CDS creation algorithm is Wu and Li algorithm;
however, it generates a relatively high signaling overhead. Utilizing the location information of
network members reduces the signaling overhead of Wu and Li algorithm. In this paper, we
compare the performance of Wu and Li algorithm with its Location-Information-Based version
under two types of Medium Access Control protocols, and several network sizes. The MAC
protocols used are: a virtual ideal MAC protocol, and the IEEE 802.11 MAC protocol. The use
of a virtual ideal MAC enables us to investigate how the real-world performance of these
algorithms deviates from their ideal-conditions counterpart. The simulator used in this research
is the ns-2 network simulator.
KEYWORDS
Network Protocols, Ad Hoc Networks, Mobile Networks, Network Simulator
1. INTRODUCTION
The multi-hop radio packet networks, also known as mobile Ad Hoc networks (MANETs), are
proposed by researchers to allow for the instantaneous creation of reconfigurable temporary
wireless communications networks. As the name implies, MANETs are created by the mobile
hosts, a.k.a. nodes, on an Ad Hoc basis without any support or administration provided by any

Computer Science & Information Technology (CS & IT) 56
fixed or pre-installed communications infrastructure. The self-control, self-organization, and Ad
Hoc properties make MANETs applicable in both civilian and military sectors. Their applications
range from providing temporary and instantly deployable communications networks in
conference rooms, class rooms, and exhibitions to providing reliable inter-vehicle
communications, and communications networks in rural areas, devastated areas, and battle fields
[1].
Along with their attractive properties, MANETs have other properties that make their realization
a challenging task. Due to the mobile nature and the limited transmission ranges of MANETs’
nodes, communication links between nodes are established and torn down in unpredictable
fashion; therefore, MANETs are characterized by dynamic topologies. The self- control property,
a powerful feature of MANETs, obliges nodes to perform routing and other network
administrative tasks cooperatively. The routing algorithms used in traditional wired networks
generate a huge amount of overhead traffic in order to discover and maintain routes in such
dynamic topologies. The limited bandwidth of the wireless channel renders all these algorithms
non-feasible for MANETs routing.
The Connected Dominating Set (CDS), a.k.a. virtual backbone or spine, is a proposed solution for
routing, broadcasting, and establishing a dynamic infrastructure for distributed location databases
in MANETs [5], [8], [11], [12]. Minimizing the CDS cardinality simplifies the MANET’s
abstracted topology, allows for using shorter routes, and reduces the total number of
retransmissions in broadcasting scenarios. It is proven that finding the minimum CDS (MCDS) in
general graphs is NP-complete [7]; therefore, approximation algorithms and heuristics are used to
tackle this problem.
The literature is rich of many CDS approximation algorithms competing in terms of CDS size,
running time, and signaling overhead [3], [6], [9], [10], [12], [14], [16]. It has been reported that,
among all competing algorithms, localized CDS creation algorithms are the fastest and the
lightest in terms of signaling overhead [2]. Examples of these algorithms are Wu and Li algorithm
[9], the MPR algorithm [13], and Alzoubi algorithm [4]. The simplest and the fastest among all
these algorithms is Wu and Li algorithm; however, this algorithm has a relatively high signalling
overhead. To reduce the signaling overhead, proposals were made that are based on utilizing
nodes’ location information [11], [20].
In this paper we compare the performance of the original Wu and Li algorithm [9] with that of its
Location-Information-Based version given in [20] under two different types of Medium Access
Control protocols (MAC protocols) and several network sizes. The MAC protocols used are: a
virtual ideal MAC protocol, and the IEEE 802.11 MAC protocol [21]. While the ideal MAC
protocol adheres to the design assumption made by the algorithms designers regarding the
guaranteed delivery of broadcasted packets, the IEEE 802.11 MAC violates this assumption since
broadcasted packets may not be correctly delivered due to packet collisions. The use of a virtual
ideal MAC gives us the ability to investigate how the real-world performance of these algorithms
deviates from their ideal-conditions counterpart.
The rest of this paper is organized as follows: Section 2 gives a description of the working
mechanics of Wu and Li algorithm [9] and its Location-Information-Based version [20]. Section
3 deals with the experiment setup. In section 4, the results are presented and discussed. Finally,
section 5 concludes this paper.

57 Computer Science & Information Technology (CS & IT)
2. ALGORITHMS DESCRIPTION
This section gives a description of the CDS creation and maintenance mechanisms of the studied
algorithms.
2.1. Wu and Li Algorithm
Initially the CDS is empty. Every node starts by broadcasting a list that contains the identities of
its 1-hop neighbors. Exchanging this information with neighbors allows every node to know the
identities of all nodes in its 2-hop neighborhood and the connections among its 1-hop neighbors.
Based on this knowledge, every node decides whether to join the CDS or not and it informs its
neighbors about this decision. A node, आ, joins the CDS if it has at least two neighbors that are
not directly connected to each other. After receiving the decisions of all its neighbors, a CDS
member, υ, makes its final decision regarding whether to withdraw from the CDS or not, and it
informs its neighbors about this decision. υ withdraws from the CDS in one of two cases; the first
case is when υ has a neighbor u such that N[υ] ⊆ N[u] and id(υ) < id(u), where N[υ] is the closed
1-hop neighborhood of node υ and id(υ) is υ’s ID; the other case is when υ has two neighbors u
and w such that u and w are directly connected and N[υ] ⊆ N[u] ∪ N[w] and id(υ) = min(id(υ),
id(u), id(w)). To maintain the CDS in face of topology changes, every node, आ, exchanges its 1-
hop neighborhood information with its neighbors whenever it discovers a topology change. Every
node u ∈ N[आ] updates its role according to the connectivity of its current 1-hop neighborhood by
applying the rules explained above.
2.2. Location-Information-Based Wu and Li Algorithm
To reduce the signaling overhead of Wu and Li algorithm, Stojmenovic et al. proposed that nodes
have to broadcast their location information, i.e. X and Y coordinates, instead of broadcasting
their lists of neighbors [11]. Using the location information of its neighbors, a node, आ, can
discover if any two of its neighbors are directly connected or not; nodes u and w are directly
connected if the Euclidean distance between them is shorter than the transmission range;
otherwise, they are not connected; however, in [20], it was shown that utilizing the location
information does not guarantee the reduction of signaling overhead. The argument made in [20] is
that since the node ID is a four-byte integer number while its location information, i.e. its X and
Y coordinates, is two eight-byte floating point numbers, broadcasting the location information of
a given node is more costly, in terms of signaling overhead, than broadcasting its list of neighbors
if this node has three neighbors or less. Based on this argument, it was shown that in order to
reduce the signaling overhead of Wu and Li algorithm every node has to broadcast its X and Y
coordinates only if it has more than four neighbors; otherwise, it must broadcast its neighbors list.
3. EXPERIMENT SETUP
The simulator used in this study is the ns-2 network simulator [17], a widely used discrete event
simulator targeted at network research. Nodes use Hello messages to discover their neighborhood;
every node broadcasts a Hello message every second. The transmission range of every node is set
to 250 meters. To reduce the effect of the short-live connections that are formed when fast
moving nodes come into the transmission range of each other for very short periods of time, node
υ does not add node u to its neighbors list until it receives the third consecutive Hello message
from u.
The network sizes simulated range from 30 nodes to 100 nodes with increments of 10. The
mobility model used is the Random Waypoint mobility model [18] and a field of 2000 m × 2000
m was used with this model. The speed of movement was uniformly distributed between 20 and

25 m/s. For every network size, two different MAC protocols were used: a virtual ideal MAC
protocol, and the IEEE 802.11 MAC protocol. For every network size and MAC protocol, fifteen
different movement scenarios were simulated. Every simulation lasted for 300 seconds. The
average of the results of every set of fifteen simulations is calculated and used to represent the
algorithm’s performance under the given network size and MAC protocol.
4. EXPERIMENT RESULTS
The results obtained from the conducted simulations are presented in the following figures. These
results include the CDS size, the unicast signaling overhead, the broadcast signaling overhead, the
CDS establishment time, and the total running time.
4.1. The CDS size
The average size of the CDS produced by each algorithm as a function of the network size for
each MAC protocol is given in figures 1 and 2. It can be seen that, in both algorithms, using the
IEEE 802.11 MAC leads to creating smaller CDS compared to that created by using the virtual
ideal MAC. The fact that in broadcasting scenarios the IEEE 802.11 MAC does not eliminate
collisions is the reason for this decrease in CDS size. Losing packets due to collisions prevents
nodes from having accurate knowledge about changes in their 1-hop neighborhood and this
inaccurate knowledge prevents nodes that are supposed to join the CDS from joining it and
decreases the CDS size.
Figure 1. Size of the CDS created by Wu and Li algorithm for different number of nodes and MAC
protocols.

Figure 2. Size of the CDS created by Location-Information-Based Wu and Li algorithm for different
number of nodes and MAC protocols.
4.2. Unicast Signaling Overhead
Unicast signaling refers to messages sent from a sender to a specific receiver. In this type of
signaling, the IEEE 802.11 MAC uses a handshaking mechanism between the sender and the
receiver to avoid collisions; however, this handshaking mechanism does not eliminate collisions
completely, since unicasted packets may collide with packets broadcasted by hidden nodes [19].
Figures 3 and 4 show the average number of bytes unicasted by every node during the CDS
creation and maintenance phases of each algorithm as a function of the network size for each
MAC protocol. It can be noticed that, in both algorithms, using the IEEE 802.11 MAC leads to
generating less unicast signaling compared to that generated by using the virtual ideal MAC. The
fact that the IEEE 802.11 MAC does not eliminate collisions completely is the reason for this
decrease in unicast signaling. Due to collisions, nodes may not get updated about changes in their
1-hop neighborhood; therefore, they do not react to these changes and this leads to generating less
unicast signaling overhead.
4.3. Broadcast Signaling Overhead
Broadcast signaling refers to messages sent from a sender to all its neighbors. In this type of
signaling, the IEEE 802.11 MAC does not use a handshaking mechanism between the sender and
the receiver so that the chances of collisions are high; hence, broadcast signaling is considered a
problematic type of signaling. Figures 5 and 6 show the average number of bytes broadcasted by
every node during the CDS creation and maintenance phases of each algorithm as a function of
the network size for each MAC protocol. It can be noticed that, in both algorithms, using the
IEEE 802.11 MAC leads to generating less broadcast signaling compared to that generated by
using the virtual ideal MAC. This decrease in broadcast signaling overhead is due to lack of
updates about neighborhood changes that results from losing of packets because of collisions.
5. THE CDS ESTABLISHMENT TIME
The CDS establishment time is the time at which every node knows whether it is a CDS member
or not. Figures 7 and 8 show the average CDS establishment time for each algorithm as a
function of network size for each MAC protocol. It can be noticed from these figures that, in both

algorithms, using the IEEE 802.11 MAC leads to longer CDS establishment time compared to the
time spent in CDS establishment by using the virtual ideal MAC. Losing of packets due to
collisions forces nodes to keep waiting for retransmissions of the lost packets and this results in
longer CDS establishment time.
Figure 3. Unicast signaling overhead generated by Wu and Li algorithm for different number of nodes and
MAC protocols.
Figure 4. Unicast signaling generated by Location-Information-Based Wu and Li algorithm for different
network sizes and MAC protocols.

Figure 5. Broadcast signaling overhead generated by Wu and Li algorithm for different number of nodes
and MAC protocols.
6. THE TOTAL RUNNING TIME
The total running time is the time spent during CDS creation and maintenance phases. The
average total time of each algorithm as a function of the network size for each MAC protocol is
shown in figures 9 and 10. Again, we notice that using the IEEE 802.11 MAC leads to longer
total time compared to the time spent by using the virtual ideal MAC, and the reason for this is
the loss of packets due to collisions that keeps nodes waiting for retransmissions of the lost
packets.
Figure 6. Broadcast signaling generated by Location-Information-Based Wu and Li algorithm for different
network sizes and MAC protocols.

Figure 7. Time to establish the CDS by Wu and Li algorithm for
different number of nodes and MAC protocols.
7. CONCLUSION
In this paper the performance of the original Wu and Li algorithm is compared with that of its
Location-Information-Based version under a virtual ideal MAC protocol and the IEEE 802.11
MAC protocol. The performance is measured in terms of CDS size, signaling overhead, and run
time. The results obtained showed that using the IEEE 802.11 MAC protocol leads to smaller
CDS size, lighter signaling overhead, and longer run times compared to those obtained by using
the virtual ideal MAC protocol. The reason of these deviations is the loss of packets due to
collisions. This loss of packets deprives nodes of acquiring accurate knowledge and fast updates
about changes of their 1-hop neighborhoods, this lack of accurate and up-to-date information may
lead nodes to refrain from joining the CDS erroneously, not broadcast information unaware of
neighborhood changes, and spend longer time waiting for information.
Figure 8. Time to establish the CDS by Location-Information-Based Wu and Li algorithm for
different number of nodes and MAC protocols.

Figure 9. Total running time by Wu and Li algorithm for different
number of nodes and MAC protocols.
Figure 10. Total running time by Location-Information-Based Wu and Li algorithm for different number of
nodes and MAC protocols.
REFERENCES
[1] C. E. Perkins (2001) Ad Hoc Networking, Addison-Wesley.
[2] S. Basagni, M. Mastrogiovanni, A. Panconesi and C. Petrioli , (2006) “Localized Protocols for Ad
Hoc clustering and Backbone Formation: A Performance Comparison”, IEEE Trans. Parallel Distrib.
Syst., Vol. 17, No. 4, pp292–306.
[3] P. Wan, K. Alzoubi, and O. Frieder, (2004) “Distributed Construction of Connected Dominating Set
in Wireless Ad Hoc Networks”, MONET Vol. 9, No. 2, pp141–149.
[4] K. Alzoubi, (2002) “Virtual Backbones in Wireless Ad Hoc Networks”, PhD thesis, Illinois Institute
of Technology.
[5] R. Sivakumar, P.Sinha, and V. Bharghavan, (1999) “CEDAR: A Core-Extraction Distributed Ad Hoc
Routing Algorithm”, IEEE Journal on Selected Areas in Communications, Vol. 17, pp1454–1465.

[6] Jeremy, B. M. Ding, A. Thaeler, and X. Cheng, (2004) “Connected Dominating Set in Sensor
Networks and MANETs”, Handbook of Combinatorial Optimization D.-Z. Du and P. Pardalos (Eds.),
pp329–369, Kluwer Academic Publishers.
[7] S. Guha, and S. Khuller, (1998), “Approximation algorithms for connected dominating sets,
Algorithmica”, Vol. 20, pp374–387.
[8] B. Das, and V. Bharghavan, (1997), “Routing in Ad Hoc Networks Using Minimum Connected
Dominating Sets”, International Conference on Communications, Montreal, Canada.
[9] J. Wu and H. Li, (1999), “On calculating connected dominating set for efficient routing in Ad Hoc
wireless networks”, in DIAL M’99, Seattle, pp7–14.
[10] L. Jia, R. Rajaraman, and T. Suel (2001), “An Efficient Distributed Algorithm for Constructing Small
Dominating Sets”, in Proceedings of the 20th ACM Symposium on Principles of Distributed
Computing (PODC’01).
[11] I. Stojmenovic, M. Seddigh, and J. Zunic, (2002), “Dominating Sets and Neighbor Elimination-Based
Broadcasting Algorithms in Wireless Networks”, IEEE Transactions on Parallel and Distributed
Systems, Vol. 13, pp14–25.
[12] B. Liang and Z. J. Haas, (2000), “Virtual Backbone Generation and Maintenance in Ad Hoc Network
Mobility Management”, in INFOCOM, pp1293–1302.
[13] J. Wu, (2003), “An Enhanced Approach to Determine A Small Forward Node Set Based on
Multipoint Relays”, in Proc. of IEEE VTC, Vol. 4, pp2774–2777.
[14] D. Dubhashi, A. Mei, A. Panconesi, J. Radhakrishnan, and A. Srinivasan, (2003), “Fast Distributed
Algorithms for (Weakly) Connected Dominating Sets and Linear-Size Skeletons”, in Proc. ACM-
SIAM Symposium on Discrete Algorithms (SODA), pp717–724.
[15] K. M. Alzoubi, P.-J. Wan, and O. Frieder, (2002), “Message-Optimal Connected Dominating-Set
Construction for Routing in Mobile Ad Hoc Networks”, in 3rd ACM International Symposium on
Mobile Ad Hoc Networking and Computing (MobiHoc02).
[16] S. Parthasarathy and R. Gandhi, (2004), “Fast Distributed Well Connected Dominating Sets for Ad
Hoc Networks, Technical Report CS-TR-4559, University of Maryland, Computer Science
Department.
[17] K. Fall and K. Varadhan: The ns Manual (formerly ns Notes and Documentation),The VINT Project.
[18] D. Johnson, D. Maltz, J. Jetcheva, (2002), “The Dynamic Source Routing Protocol for Mobile Ad
Hoc Networks”, Internet Draft, draft-ietf-manet-dsr-07.txt, work in progress,.
[19] https://en.wikipedia.org/wiki/Hidden node problem
[20] K. Almahorg, S. Naik, and X. Shen, (2010), “Efficient Localized Protocols to Compute Connected
Dominating Sets for Ad Hoc Networks”, IEEE GLOBECOM,.
[21] http://grouper.ieee.org/groups/802/11/main.html.

A SIMULATION-BASED PERFORMANCE COMPARISON OF MANETS CDS CREATION ALGORITHMS USING IDEAL MAC AND IEEE 802.11 MAC

More Related Content

What's hot (16)

Similar to A SIMULATION-BASED PERFORMANCE COMPARISON OF MANETS CDS CREATION ALGORITHMS USING IDEAL MAC AND IEEE 802.11 MAC (20)

More from cscpconf (20)

Recently uploaded (20)

A SIMULATION-BASED PERFORMANCE COMPARISON OF MANETS CDS CREATION ALGORITHMS USING IDEAL MAC AND IEEE 802.11 MAC