Sunteți pe pagina 1din 7

(IJCNS) International Journal of Computer and Network Security, 113

Vol. 2, No. 6, June 2010

Study the Effect of Node Distribution on QoS


Multicast Framework in Mobile Ad hoc Networks
Mohammed Saghir

Computer Department, Hodeidah University, Yemen


Information Center, Hodeidah Governorate, Yemen
mshargee@yahoo.com

Abstract: The distribution of nodes in mobile ad hoc


networks affects connectivity, network capacity, group member 2. Literature Review
and route length. In this paper, the effect of node distribution on It has been proven that the capacity of network does not
the QoS multicast framework (FQM) is studied. In order to do
increase with its size when nodes are stationary [2]. On the
this, extensive simulations are performed for FQM with two
well-known node placement models: The Random placement other hand, it has been proven that the mobility increases
model and Uniform placement model. the capacity of mobile ad hoc networks [3]. The
The performance of FQM with the placement models is studied performances of some placement modes are studied in
under different node mobility and node density. The analysis of previous work. In [4], a new algorithm (a Node Placement
simulation results have shown that there is a difference between Algorithm for Realistic Topologies-NPART) was proposed
the performance of FQM-Uniform and FQM-Random when to create realistic network topologies. This algorithm
mobility is zero. This difference is decreased while mobility
increased.
provides realistic topology with different input data. The
NPART algorithm is compared with the uniform placement
Key words: MANET, RANDOM, UNIFORM. model. The result of simulation shows that NPART
algorithm reflects the properties for user initiated network.
1. Introduction The researchers in [5] modified the Particle Swarm
Wireless networking and multimedia applications are Optimization (PSO) using genetic algorithm to optimize
growing in importance rapidly. The motivation for node density and improve QoS in sensing coverage. The
supporting QoS multicasting in MANET is the fact that nodes in simulation area are divided into stationary nodes
multimedia applications are becoming important for group and mobile nodes and the study focuses on how to optimize
communication. Among types of wireless networks, mobile nodes distribution to improve QoS in sensing
MANET provides flexible communication with low cost. All coverage in sensor network.
communications are done over wireless media without the The properties of the random waypoint mobility model
help of wired base stations. The environment for MANETs (RWP) are studied in [6], [7], [8], [9], [10], [11], [12] and
is very volatile so connections can be dropped at any the bugs that might occur when using this model is
moment. Distant nodes communicate over multiple hops highlighted. The researchers concluded that the nodes
and nodes must cooperate with each other to provide distributions after long time of simulation are different from
routing. The challenges in MANET are attributed to the initial nodes distributions. In addition, the random
mobility of intermediate nodes, absence of routing waypoint model and Brownian-like model are studied in [8]
infrastructure, low bandwidth and computational capacity of and concluded that the concentration of nodes in the center
mobile nodes. The network traffic is distributed through of simulation area in RWP model is based on the choice of
multi-hop paths and the construction of these paths is mobility parameters. Moreover, the effect of RWP model on
affected by node distribution. In addition, the node the node distribution is studied in square and circle area
distribution affects connectivity, network capacity, group [13]. The behavior of mobile nodes in RWP mobility model
member and route length in mobile ad hoc networks. is outlined and analyzed. Some parameters for distribution
Moreover, if mobile nodes are not distributed uniformly, in the RWP model to accurate the movement of mobile
some area of ad hoc network may not be covered. In some nodes in a square area are defined.
cases, mobile nodes are located in the middle of simulation
area and as a result, they have higher average connectivity The previous studies are focused on studying and updating
degree than nodes at border [1]. the node placement models and mobility models while in
this paper, we study the effect of node placement models on
This paper is structured as follows: Section 2 gives an the performance of FQM framework as the node distribution
overview on the previous work whereas Section 3 describes affects the group member and construction of multi-hop
the QoS multicast framework FQM and defines the uniform paths communication in the QoS multicast routing
and random placement models. In Section 4, the simulation protocols.
results of implementing FQM with two different placement
models are presented. Finally, Section 5 concludes this
paper and makes mention of future work.
114 (IJCNS) International Journal of Computer and Network Security,
Vol. 2, No. 6, June 2010
3. The FQM with two Placement Models multiplexed when it is necessary to reach two or more
destinations on disjoint paths. This advantage conserves
3.1 The FQM QoS multicast framework bandwidth and network resources [14]. In our previous work
Multicast routing is more efficient in MANETs because it is [15], we propose a cross-layer Framework FQM to support
inherently ready for multicast due to their broadcast nature QoS multicast applications for MANETs. Figure 1 gives an
that avoids duplicate transmission. Packets are only overview on the cross-layer framework FQM.

Figure 1. An Overview on the Functionalities of the Cross-Layer Framework While Receiving Flows

The first component of the framework is a new and efficient requirements. Second, a distributed admission control which
QoS multicast routing protocol (QMR). The QMR protocol used to prevent nodes from being overloaded by rejecting the
is used to find and maintain the paths that meet the QoS request for new flows that will affect the ongoing flows.
(IJCNS) International Journal of Computer and Network Security, 115
Vol. 2, No. 6, June 2010

Third, an efficient way to estimate the available bandwidth performance metrics:


and provides the information of the available bandwidth for • Packet delivery ratio: the average of the ratio
other QoS schemes. Fourth, a source based admission between the number of data packets received and
control which used to prevent new sources from a affecting the number of data packets that should have been
the ongoing sources if there is not enough available received at each destination. This metric indicates
bandwidth for sending to the all members in the multicast the reliability of the proposed framework.
group. Fifth, a cross-layer design with many QoS scheme: • Control overhead: number of transmitted control
classifier, shaper, dynamic rate control and priority queue. packet (request, reply, acknowledgment) per data
These schemes work together to support real-time packet delivered. Control packets are counted at
applications. each hop. The available bandwidth in MANETs is
limited so it is very sensitive to the control
The traffic is classified and processed based on its priority. overhead.
Control packets and real-time packets will bypass the shaper • Average latency: the average end-to-end delivery
and be sent directly to the interface queue at MAC layer. delay is computed by subtracting packet generation
The best-effort packets should be regulated based on the time at the source node from the packet arrival time
dynamic rate control. In terms of queue priority, control at each destination. The multimedia applications
packets and real-time packets have higher priority than best- are very sensitive to the packet delay; if the packet
effort packets. All components in the framework are takes long time to arrive at destinations, it will be
cooperating to provide the required level of services. useless.
3.2 The Uniform Placement Model • Jitter: it is defined as a variation in the latency of
received packets. It is determined by calculating the
In Uniform placement model, the simulation area is divided
standard deviation of the latency [17]. This is an
into a number of cells based on the number of mobile nodes
important metric for multimedia applications and
in the simulation area. Within each cell, a node is placed
should be kept to a minimum value.
randomly. The uniform placement model is set to create
topologies with an equaled average node degree and this • Group Reliability: it is defined as the ratio of number
improves connectivity. of packets received at 95% of destination and
number of packets should be received. This means
3.3 The Random Placement Model that the packet is considered to be received only if it
In the Random placement model, mobile nodes are placed is received by 95% of the number of multicast
randomly in a given area according to a probability group.
distribution of mobile nodes. Based on the probability
4.1 The Performance of FQM under Different Mobility
distribution, the mobile nodes density is different from one
area to another in the simulation area of ad hoc networks. In this section, we study the performance of FQM with the
Actually, the Random distribution is suitable as it reflects uniform and random placement models under different
the real behavior of mobile nodes in ad hoc networks. mobility.
4.1.1 Packet Delivery Ratio (PDR)
4. Performance Evaluation 100
90 FQM-Random FQM-Uniform
We have conducted experiments using GLOMOSIM [16] to
study the effect of node distribution on the QoS multicast 80

framework (FQM). The main concern of these experiments 70

is to study the effect of the random placement model and the 60


PDR

uniform placement model on FQM framework while 50


supporting QoS multicast applications. This simulation was 40
run using a MANET with different number of nodes moving 30
over a rectangular 1000 m × 1000 m area for over 900 20
seconds of simulation time. Nodes in the simulation move 10
according to the Random Waypoint mobility model provided 0 5 10 15 20
by GLOMOSIM. Mobility speed is ranged from 0-20 m/s Mobility (m/s)
and the pause time is 0 s. We used one multicast source
sending to 15 multicast destinations in all experiments Figure2. Performance of PDR vs. mobility.
(assuming all destinations were interested to receive from
the source node). The radio transmission range is 250 M
and the channel capacity is 2Mbit/s. Each data point in this In a stationary network, the nodes always remain either in
simulation represents the average result of ten runs with the range of each other or out of range. When mobility is
different initial seeds. increased, the positions of mobile nodes are changed. The
The performance of FQM with uniform and random performance of PDR vs. increasing mobility is given in
placement models is studied through the following Figure2. The PDR in FQM-Uniform is significantly higher
116 (IJCNS) International Journal of Computer and Network Security,
Vol. 2, No. 6, June 2010

than that in FQM-Random. In FQM-Uniform, the mobile show that AL for FQM-Uniform is relatively lower than
nodes are distributed uniformly so the degree of neighbor FQM-Random. In FQM-Uniform, the traffic is distributed
nodes is equaled and as a result the traffic load is balanced through different paths while in FQM-Random the traffic is
through intermediate nodes. congested and as a result, the AL is increased. In addition,
when mobility is increased, the uniform distribution of
In FQM-Random, the mobile nodes are distributed
nodes is changed and as a result, the AL in FQM-Uniform is
randomly so the degrees of nodes are different from one
increased.
node to another and as a result, the traffic load may congest
through some intermediate nodes. When mobility is 4.1.1 Jitter
increased, the distributions of nodes are affected and as a 100

FQM-Random FQM-Uniform
result the difference of PDR between FQM-Uniform and 90

FQM-Random are decreased. 80

4.1.2 Control Overhead (OH) 70

Jitter (ms)
60
0.39
0.36 FQM-Random FQM-Uniform 50
0.33
0.3 40
OH per Packet Delivery

0.27
30
0.24
0.21 20
0.18
0.15 10

0.12
0
0.09 0 5 10 15 20
0.06
Mobility (m/s)
0.03
0
Figure 5. Performance of jitter vs. mobility.
0 5 10 15 20
Jitter occurs due to temporally lack of wireless connections
Mobility (m/s)
and scheduling issues on the link layer [18]. The number of
Figure3. Performance of OH vs. mobility. hops in the path is affected by the node distribution and as a
Figure 3 shows the Control OH vs. increasing mobility. The result, the jitter is affected. Frequently changing routes
results show that control OH for FQM-Uniform is lower could increase the jitter since the time for selecting forward
than FQM-Random when mobile nodes are static; this is nodes and the delay variation between the old and new
because the number of data packets that received in FQM- routes increase the jitter. Figure 5 gives an overview on the
Uniform was higher than that received in FQM-Random. As performance of jitter vs. increasing mobility. The results
mobility increased, the number of data packets that received show that the jitter for FQM-Uniform is relatively less than
in FQM-Uniform decreased and this affect the percentage of FQM-Random when mobile nodes are static. When mobility
control OH per packet. As a result of this, the differences increased, the uniform node distribution is affected and as a
between FQM-Uniform and FQM-Random are decreased. result the differences between FQM-Uniform and FQM-
Random in jitter are decreased.
4.1.3 Average latency (AL)
100 4.1.5 Group Reliability (GR)
90 FQM-Random FQM-Uniform
The Group Reliability vs. increasing mobility is given in
80
Figure 6. The GR for FQM-Uniform is higher than that in
70
FQM-Random when mobility is zero. As mobility is
60
increased, the difference between FQM-Uniform and FQM-
Random in group reliability is decreased as discussed in
AL (ms)

50

40
section 4.1.1.
100
30
FQM-Random FQM-Uniform
90
20
80
10
70

0
60
0 5 10 15 20
GR

50

Mobility (m/s)
40

30
Figure4. the Performance of AL vs. mobility.
20

In wireless networks such as IEEE 802.11, the mobile nodes 10

share the same channel and use the contention mechanism 0


0 5 10 15 20

to capture the channel so the bandwidth availability affected Mobility (m/s)

by the number of nodes in the surrounding and this


increases network delay and jitter. The results in Figures 4 Figure 6. Performance of GR vs. mobility.
(IJCNS) International Journal of Computer and Network Security, 117
Vol. 2, No. 6, June 2010

4.2 The Performance of FQM under Different Nodes


300
Density
270
In this section we study the performance of FQM with the FQM-Random FQM-Uniform
240
uniform and random placement models under different 210
nodes density with no mobility to focus on the effect of node 180

AL (ms)
density. 150
120
In a denser network, the probability of mobile nodes to sense
90
the activities of its neighbor nodes increases so the packet
60
collision that is coming due to hidden terminals is reduced
30
[19]. In addition, when the network density increases, the
0
number of connection increases so packets can finds paths to 100 150 200 250
arrive at destinations. Number of nodes

In sparser network, the connectivity will be small so the


Figure 8. AL vs. network density
number of delivered data packets is few due to lack of routes
[20]. On the other side, when the network density is very 4.2.3 Jitter
high, the interference between nodes increases and this The performance of FQM as a function of network density is
increases the collisions and reduces the channel access described in Figure 9. The results describe the difference
which leads to drop data packets. between the jitter in FQM-Random and FQM-Uniform.

4.2.1 Packet Delivery Ratio (PDR) 400

The packet delivery ratio as a function of network density 360


FQM-Random FQM-Uniform
for FQM-Uniform and FQM-Random is given in Figure 7. 320

The figure shows that there is a difference between PDR 280

FQM-Uniform and PDR FQM-Random. The difference is 240


Jitter (ms)

increased when the node density is increased. This is 200

because in FQM-Uniform, the mobile nodes uniformly 160

distributed so when node density is increased, the number of 120

paths increased and traffic load is balanced and as a result 80

the PDR increased. 40


0
In FQM-Random, when node density is increased with 100 150 200 250
random distribution, the traffic is congested and the Number of nodes
available bandwidth is reduced as a result of increase
contention and collision between mobile nodes. Figure 9. Jitter vs. network density
100 FQM-Random FQM-Uniform 4.2.4 Group Reliability (GR)
90 Figure 10 shows the performance of the group reliability vs.
80 increasing network density. From the Figure, there is a
70
difference between GR in FQM-Uniform and FQM-Random
60
and this is because PDR in FQM-Uniform is higher than
PDR

PDR in FQM-Random as discussed before in Section 4.2.1.


50

40
100
30 FQM-Random FQM-Uniform
90
20
80
10
70
100 150 200 250
Number of nodes 60
GR

50

Figure 7. PDR vs. network density 40


30
4.2.2 Average Latency (AL)
20
The number of nodes in the neighborhood affects the
10
available bandwidth and this increases the network delay
0
and jitter. Figure 8 reflects the difference between the AL in
100 150 200 250
FQM-Uniform and FQM-Random. The AL in FQM- Number of nodes
Random is slightly increased while network density
increased.
Figure 10. GR vs. network density
118 (IJCNS) International Journal of Computer and Network Security,
Vol. 2, No. 6, June 2010

4.3 The Performance of FQM under High while mobility is increased; this is because the average
Density and High Mobility latency is changed while mobility is increased.
The effect of mobility with low node density is discussed in
400
details in section 4.1. In this Section, the effect of high
360 FQM-Random FQM-Uniform
mobility with high density on the performance of FQM with
320
uniform and random placement models is studied. The node
280
density is 100 mobile nodes and mobility is 20 m/s.

Jitter (ms)
240
4.3.1. Packet Delivery Ratio (PDR) 200
The packet delivery ratio as a function of mobility with high
160
density for FQM-Uniform and FQM-Random is given in
120
Figure 11. The figure shows that the difference between
80
PDR for FQM-Uniform and FQM-Random is decreased
40
while mobility increased even the node density is high. This
is because the uniformly distribution is affected with high 0
0 20
mobility and as a result, the traffic is congested and
M obility (m/s)
available bandwidth is reduced.
100 FQM-Random FQM-Uniform Figure 13. Jitter as a Function of mobility with high density
90
4.3.4. Group Reliability (GR)
80
The performance of the group reliability with high mobility
70 and high node density is described in Figure 14. The Figure
60 shows that the difference between GR in FQM-Uniform and
PDR

50 FQM-Random is decreased while mobility is increased; this


is because the uniformly distribution is affected with high
40
mobility and as a result, the traffic is congested and
30
available bandwidth is reduced as discussed before in
20 Section 4.3.1.
10
0 20 100
Mobility (m/s) FQM-Random FQM-Uniform
90

80
Figure 11. PDR as a Function of mobility with high density 70

60
4.3.2. Average Latency (AL)
GR

The number of nodes in the neighborhood affects the 50

available bandwidth and this increases the network delay 40

and jitter. Figure 12 shows that the difference between the 30

AL in FQM-Uniform and FQM-Random is decreased while 20

mobility is increased even the node density is high. 10

0
0 20
300
Mobility (m/s)
270 FQM-Random FQM-Uniform

240
210 Figure 14. GR as a Function of mobility with high density
180
AL (ms)

150
120 5. Conclusion and Future work.
90
60 In this paper, we have studied the performance of FQM
30 framework with two placement models under different node
0 mobility and node density. From the results, the
0 20 performance of the QoS multicast framework FQM with
Mobility (m/s)
Uniform placement model (FQM-Uniform) is better than the
Figure 12. AL as a Function of mobility with high density performance of FQM with Random placement model (FQM-
Random) when the mobility of nodes is zero. Although the
4.3.3. Jitter Uniform placement model is superior Random placement
The results in Figure 13 reflect that the difference between model, the Random placement model is more suitable to
the jitter in FQM-Random and FQM-Uniform is decreased reflect the real behavior of nodes in mobile ad hoc networks.
(IJCNS) International Journal of Computer and Network Security, 119
Vol. 2, No. 6, June 2010

This is because the mobility is the main characteristic of [12] J. Yoon, M. Liu, and B. Noble, “Random Waypoint
mobile ad hoc network. Moreover, the Uniform placement Considered Harmful,” In the Proceeding of the IEEE
model is suitable in some applications of sensor networks INFOCOM, 2003.
where static sensors are used. [13] C. Bettstetter, G. Resta and P. Santi, “The Node
Distribution of the Random Waypoint Mobility Model
The analysis of simulation results shows that the mobility for Wireless Ad Hoc Networks,” IEEE Transaction on
model has the most effect on the performance of the FQM Mobile Computing, Vol, 2, No. 3, JULY-
QoS multicast as it changes the distribution of mobile nodes SEPTEMBER, 2003.
and as a result, it affects the group member and network [14] M. Hasana and L. Hoda, Multicast Routing in Mobile
capacity. In future work, we intend to study the performance Ad Hoc Networks: Kluwer Academic Publishers, 2004.
of the FQM QoS multicast framework with different [15] M. Saghir, T. C. Wan, and R. Budiarto, “A New Cross-
mobility models. Layer Framework for QoS Multicast Applications in
Mobile Ad hoc Networks,” International Journal of
Computer Science and Network Security, (IJCSNS),
References vol. 6, pp. 142-151, 2006.
[16] http://pcl.cs.ucla.edu/projects/glomosim.
[1] C. Bettstetter and C. Wagner, “The Spatial Node [17] K. Farkas, D. Budke, B. Plattner, O. Wellnitz, and L.
Distribution of the Random Waypoint Model,” In Wolf, “QoS Extensions to Mobile Ad Hoc Routing
Preceding the First German Workshop Mobile Ad Hoc Supporting Real-Time Applications,” In the
Networks, 2002. proceeding of the 4th ACS/IEEE International
[2] P.gupat and P. Kumer, “the capacity of ad hoc Conference on Computer Systems and Applications,
networks”, IEEE transaction information theory, V 64, Dubai-UAE, 2006.
no 2, pp388-404. [18] O. Farkasa, M. Dickb, X. Gub, M. Bussec, W.
[3] M. grossglauser and D.Tse, “mobility increase capacity Effelsbergc, Y. Rebahid, D. Sisalemd, D. Grigorase, K.
of ad hoc networks”, In the proceeding of the IEEE Stefanidisf, and D. Serpanosf, “Real-time service
INFOCOM, pp1360-1369,2001. provisioning for mobile and wireless networks,”
[4] B. Milic and M. Malek, “NPART - Node Placement Computer Communications, vol. 29, pp. 540-550,
Algorithm for Realistic Topologies in Wireless Multi- 2006.
hop Network Simulation”, In Proceedings of the 2nd [19] C. Lin, H. Dong, U. Madhow, A. Gersho, “Supporting
International Conference on Simulation Tools and real-time speech on wireless ad hoc networks: inter-
Techniques , 2009 packet redundancy, path diversity, and multiple
[5] P. Song, J. Li, K. Li and L. Sui, “Researching on description coding,” In Proceedings of the 2nd ACM
Optimal Distribution of Mobile Nodes in Wireless international workshop on Wireless mobile
Sensor Networks being Deployed Randomly,” applications and services on WLAN hotspots, USA,
International Conference on Computer Science and 2004.
Information Technology, 2008. [20] A. NILSSON, “Performance Analysis of Traffic Load
[6] C. Bettstetter, “Mobility Modeling in Wireless and Node Density in Ad hoc Networks,” In proceeding
Networks: Categorization, Smooth Movement, and the Fifth European Wireless Conference, Spain, 2004.
order Effects,” ACM Mobile Comp. and Comm. Rev.,
vol. 5, no. 3, 2001. Author Profile
[7] C. Bettstetter, H. Hartenstein, and X. Perez-Costa,
“Stochastic Properties of the Random Waypoint Mohammed Saghir received his B.S
Mobility Model,” ACM/Kluwer Wireless Networks, from Technology University, Iraq in
2004. 1998, M.Sc. from Al Al-Bayt
[8] D. M. Blough, G. Resta, and P. Santi, “A Statistical University, Jordan in 2004 and his
Analysis of the Long-Run Node Spatial Distribution in Ph.D from University Sains Malaysia
Mobile Ad Hoc Networks,” In the Proceeding ACM 2008 in Computer Science. He is
Int’l Workshop Modeling, Analysis, and Simulations working as a lecturer in Hodeidah
University, Yemen. His current
of Wireless and Mobile Systems (MSWiM), 2002.
research interests include, Mobile Ad
[9] T. Camp, J. Boleng, and V. Davies, “A Survey of hoc networks, QoS multicast routing
Mobility Models for Ad Hoc Network Research,” in MANETs, WiMax.
Wireless Comm. & Mobile Computing (WCMC), vol.
2, no. 5, pp. 483-502, 2002.
[10] E. Royer, P. Melliar-Smith, and L. Moser, “An
Analysis of the Optimum Node Density for Ad Hoc
Mobile Networks,” In the Proceeding of the IEEE Int’l
Conf. Comm. (ICC), 2001.
[11] J. Song and L. Miller, “Empirical Analysis of the
Mobility Factor for the Random Waypoint Model,” In
Proceeding of the OPNETWORK, 2002.

S-ar putea să vă placă și