You are on page 1of 12

A Comprehensive Performance Study of OPNET Modeler

For ZigBee Wireless Sensor Networks


I. S. Hammoodi
Caledonian College of
Engineering, Muscat, Oman
irfan@caledonian.edu.om

B. G. Stewart
Glasgow Caledonian
University, Glasgow, UK
b.stewart@gcal.ac.uk

Abstract - OPNET has been widely used as a network


simulator, but not much emphasis has been given on the
performance of this simulator for ZigBee wireless sensor
networks (WSN). Simulation of WSNs is a challenging task
due to the nature of hardware design, energy limitations,
and deployment of a vast number of nodes. An inclusive
study and analysis of the QoS performance evaluation of the
ZigBee protocol within the OPNET simulator for different
WSN topologies and routing schemes is presented here.
Based on simulation and analysis of results this paper can be
considered as a guide for researchers in evaluating OPNET
Modeler as a WSN simulator for Zigbee networks. Some
enhancements needed in OPNET Modeler to be more
suitable for the simulation of ZigBee WSNs are discussed.
Keywords: OPNET; Wireless Sensor Networks; QoS; ZigBee

I.

INTRODUCTION

A wireless sensor network (WSN) comprises spatially


distributed autonomous devices using sensors to
cooperatively monitor physical or environmental
conditions, such as temperature, sound, vibration,
pressure, motion etc. at different locations. Subsequent to
collaborative sensing of a given environment, the nodes
perform in-network computation and communicate with a
base station when a targeted event takes place [1]. A WSN
has a number of exclusive characteristics when compared
with conventional wireless networks. These include
limited bandwidth, limited computation capability of
individual nodes, and limited energy supply. Selforganization, dynamic network topology, and multi-hop
routing are additional key possible features of a WSN,
which make them important for many applications. It is
advantageous to perform precise simulations or to develop
models before deploying WSNs in the field. This is
because WSNs may be deployed randomly in an ad-hoc
manner with a large number of tiny nodes. Simulations
help in the validation and evaluation of the performance
of sensor networks within certain application
environments, something which was not possible to
achieve a number of years ago. Consequently, simulation
of sensor networks is therefore gaining greater demand
because of their capabilities, lower energy constraints and
the use of a larger number of nodes compared to
conventional wireless networks. ZigBee (a set of
specifications built around the IEEE 802.15.4 wireless
protocol) is a common platform for WSNs. Some
published work exists on the evaluation of WSN
simulation software. As an example reference [2]
evaluates NS-2 as a simulator for ZigBee WSNs and
concludes that NS-2 itself does not perform well in terms
1Most of the work was performed when the author was with Birla

A. Kocian1
University of Rome "Tor
Vergata", Rome, Italy
kocian@ieee.org

S. G. McMeekin
Glasgow Caledonian
University, Glasgow, UK
scott.mcmeekin@gcal.ac.uk

of efficiency and support for real world applications. In


[3] a comparative study between OPNET Modeler and
NS-2 has been reported in which the accuracy of NS-2
and OPNET Modeler in a wired network is compared
using Constant-Bit-Rate (CBR) data traffic and a File
Transfer Protocol (FTP) session. It was found that NS-2
provides very similar results compared to OPNET
Modeler in terms of throughput. However the
performance evaluation or comparison in [3] does not
cover the performance evaluation of wireless networks or
WSNs in particular. A survey of simulation within sensor
networks has also been covered in [4]. Here different
kinds of sensor networks simulators including OPNET
were briefly discussed. The constraints and limitations of
each sensor simulator were identified as well as focusing
on open research issues in WSNs. However no simulation
results or actual analysis of the simulation tools were
presented.
To provide further clarity on software simulation
performance of WSNs, the intention of this paper is to
present a QoS (Quality of Service) performance
investigation of OPNET Modeler for three different
ZigBee WSN topologies and to ascertain its general
capabilities for these situations. The effect of the number
of nodes on the MAC throughput and end-to-end delay is
inclusively evaluated for these topologies. Further, the
effect of handshaking on the end-to-end delay between the
nodes is also investigated. Possible improvements to
OPNET are also proposed to make simulation of ZigBee
WSNs more suitable within this software. To the best of
the authors knowledge no work has previously been
presented which focuses on the performance evaluation of
OPNET Modeler as a WSN simulator for the ZigBee
protocol. Therefore this work may assist researchers in
evaluating and validating OPNET Modeler as an
appropriate WSN simulation tool.
The paper is structured as follows. Section 2 presents an
overview of the performance evaluation summarizing
QoS parameters and the simulation scenarios considered.
Section 3 presents the results and a critical analysis of the
different scenarios adopted. In section 4, improvements to
OPNET are suggested and in section 5, conclusions are
drawn.
II.

PERFORMANCE EVALUATION

Based on several criteria such as availability,


reliability, response time, load, throughput, bandwidth
capacity, and packet loss ratio, a real network can be
evaluated. These parameters can be summarized as QoS.

QoS can be regarded as a low level networking device


visible attribute such as bandwidth, delay, jitter and
packet loss rate, or as a high level user observable like
the quality of voice communication or video
communication [5]. The low level attributes are applicable
for most WSNs.
A network simulator performance refers to its ability and
accuracy to simulate and imitate an actual network while
maintaining an accurate log of the QoS attributes
discussed above. Additionally, a number of other factors
such as network protocol, routing, topology, energy
model, etc., have considerable influence on the operation
of a simulator and hence on its performance. Further,
architecture, data structure, and the algorithms
implemented within the simulator software also have an
effect on its performance. Clearly the accurate simulation
of a WSN is a complex task.
In order to provide a general performance evaluation of
OPNET Modeler for ZigBee WSNs, the performance
criteria presented in this paper will include end-to-end
delay, number of employed hops, throughput, data
dropped and simulation time. In general these parameters
give a clear prospective of the QoS parameters in WSNs.
A. Introduction to OPNET Modeler
The OPNET Modeler environment includes tools for
all phases of a study, including model design, simulation,
data collection, and data analysis [6]. OPNET Modeler
provides a comprehensive development environment
supporting the modelling of communication networks and
distributed systems. Both behaviour and performance of a
model can be analyzed by performing discrete event
simulations. A Graphical User Interface (GUI) supports
the configuration of the scenarios and the development of
network models. Three hierarchical levels for
configuration are differentiated: i) the network level
creating the topology of the network under investigation,
ii) the node level defining the behaviour of the node and
controlling the flow of data between different functional
elements inside the node, and iii) the process level,
describing the underlying protocols, represented by finite
state machines (FSMs) and are created with states and
transitions between states. The source code is based on
C/C++. The analysis of simulated data is supported by a
variety of built-in functions [7]. Different graphical
presentations for the simulation results are available and
node mobility can be easily implemented in different
kinds of nodes i.e. ZigBee coordinator, end device and
router nodes.
The OPNET ZigBee model uses four process models:
ZigBee MAC model which implements a model of the
IEEE 802.15.4 MAC protocol. The model implements
channel scanning, joining and failure/recovery process
of the protocol in the unslotted operation mode.
ZigBee Application model which represents a low
fidelity version of the ZigBee Application Layer as
specified in the ZigBee Specification. The process
model initiates network joins and formations,
generates and receives traffic and generates different
simulation reports.

ZigBee Carrier Sense Multiple Access/Collision


Avoidance (CSMA/CA) model which implements the
media access protocol of the MAC layer.
ZigBee Network model which implements the ZigBee
Network Layer as specified in the ZigBee
specification. This model is responsible for routing
traffic, process network join, formation requests and
generating beacons [6].

B. Simulation Scenarios
To evaluate the performance of OPNET in simulating
ZigBee WSNs, the three common topologies of WSNs
will be investigated, namely: star, mesh and tree
topologies. In star topology, nodes are connected to a
single hub node. The hub coordinator requires greater
message handling, routing, and decision-making
capabilities than the other nodes or end devices. If a
communication link is cut, it only affects one node.
However, if the coordinator fails the network is destroyed.
In mesh topology nodes are regularly distributed to allow
transmission only to a nodes nearest neighbours. The
nodes in these networks are generally identical, so that
mesh nets are also referred to as peer-to-peer nets. Mesh
nets can be good models for large-scale networks of
wireless sensors that are distributed over a geographic
region. An advantage of mesh topologies is that, while all
nodes are possibly identical and have the same computing
and transmission capabilities, certain nodes can be
designated as coordinators that take on additional
functions. If a coordinator fails, another node can then
take over these duties. In tree topology the coordinator
node is connected to one or more other nodes that are one
level lower in the hierarchy with a point-to-point link
between each of the end nodes and the coordinator node.
Also each of the end nodes that are connected to the
coordinator node will have one or more other nodes that
are one level lower in the hierarchy connected to it with a
point-to-point. The initial star topology scenario
considered here consists of 8 ZigBee end devices (reduced
function devices) and 1 coordinator (full function devices)
as shown in Fig. 1. The initial mesh and tree topology
scenarios considered consist of 8 ZigBee end devices, 6
ZigBee routers and 1 coordinator as shown in Fig. 2. The
selection of the number of ZigBee routers in the tree and
mesh topologies is made to ensure that each topology can
handle the increase in the number of end devices in the
simulation scenarios that follow without rapid increase in
the delay. The mesh and the tree routing scenarios
normally use the ad-hoc on-demand distance vector
(AODV) routing protocol. All the three topologies will be
simulated with and without the use of Request-to-Send/
Clear-to-Send (RTS/CTS) handshaking to study the effect
of handshaking on the delay and other QoS parameters. In
sensor networks hundreds to several thousands of nodes
could be deployed throughout the sensor field. Therefore
consideration of the number of nodes was included in the
simulation evaluation process. Simulation was performed
for scenarios of 8 nodes up to scenarios of 200 nodes in
the three selected WSNs topologies.

All sensor nodes were configured with CBR traffic, and


for evaluation purposes, all the nodes in a single scenario
were assumed to be in the same personal area network
(i.e. have the same Personal Area Network (PAN) ID).
For simplicity, the node will choose a random destination
node within its own PAN. The altitude of all the nodes in
all the scenarios is set to be 1 meter above ground level.
This altitude is the default value of the OPNET ZigBee
node model and can be varied according to the application
being simulated. Normally in WSNs one of the
coordinators duties is to dictate the topology of that
network, therefore in each topology the type of network
(star/tree/mesh) is set at the coordinator node. The
simulation run time for all the scenarios is set at 30
minutes. Table 1 shows the network parameters set at the
coordinator nodes only. Table 2 shows the MAC, Physical
and Application layer simulation parameters that have
been implemented in the simulation scenarios.

higher delays in the network with the mesh topology when


compared with the tree topology; this is basically due to
the differences in the routing techniques and the size of
the routing table used in mesh routing. It is seen that the
ETE delay in the star topology is higher than the other two
topologies. This is due to the fact that the star topology is
a single hop topology and there is only one path from one
node to another through the coordinator. Also it is seen
that the delay increases proportionally with the increase in
the number of nodes; this is as expected since increasing
the node numbers in WSNs will lead to higher traffic and
hence higher delay. The general trend of the simulation
results agree with the results presented in [7] therefore
demonstrates that OPNET is providing acceptable results.
Fig. 4 shows the simulation results of average ETE delay
of the WSN star topology with and without using
RTS/CTS handshaking. In general the ETE delay
confirms the behaviour in Fig. 3 in that for small node
numbers the ETE delay is very low then it increases
exponentially for higher node numbers. It is also shown
that the delay increases rapidly when RTS/CTS is not
used. This is due to excessive collisions when there is no
congestion control leading to very much higher delays.
However, for low node numbers, the delay in the case of
not utilizing RTS/CTS is slightly lower. This occurs since
there is no need for congestion control for small numbers
of nodes and data packets are processed quicker when
handshaking is not in use.

Figure 1 Initial scenario of star topology.


TABLE 1 COORDINATORS NETWORK LAYER PARAMETERS
Coordinators Network Layer Parameters
Maximum number of end devices and routers in one PAN

250

Maximum number of routers in a single PAN

Route discovery timeout (sec)


The duration of route discovery entries remaining in the
table before they are removed. (Only used in mesh
networks).

10

TABLE 2 THE MAC, PHYSICAL AND APPLICATION LAYER


SIMULATION PARAMETERS
MAC Layer Parameters

Figure 2 Initial scenario of mesh and tree topologies.

III.

SIMULATION AND ANALYSIS

A. End-to-End Delay
The end-to-end delay (ETE) is defined as the end-toend delay of all the packets received by the 802.15.4
MACs of all WPAN nodes in the network and forwarded
to the higher layer. As the number of nodes in the WANs
increase the delay obviously will increase. A simulation of
ETE delay of the three topologies with increasing number
of nodes was undertaken. Fig. 3 shows the simulation
results of the average ETE delay for the mesh, tree and
star topologies as a function of the number of nodes.
In this simulation scenario the RTS/CTS handshake is
enabled. It is seen that the difference in delays between
the mesh and the tree topology is small even when the
number of nodes increase. However, there are slightly

Acknowledge wait duration (sec)

0.05

Maximum Number of retransmissions


Minimum value of the back-off exponent in the
CSMA/CA (if this value is set to 0, collision avoidance is
disabled during the first iteration of the algorithm)
Maximum number of back-offs the CSMA/CA algorithm
will attempt before declaring a channel access failure.
Channel sensing duration (sec)

5
3
4
0.1

Physical Layer Parameters


Data rate (kbps)

250

Receiver sensitivity (dB)

-85

Transmission band (GHz)

2.4

Transmission power (W)

0.05

Application Layer Parameters


Packet interval time/ type (sec/constant )

Packet size/type

1024

(bits/constant)

B. Number of Hops
Fig. 5 shows the average number of hops taken by
application traffic sent by a particular node for different
numbers of sensor nodes for the three test scenarios. For
mesh and tree topologies, the minimum hop number is
around two. However for the star topology the number of
hops remains at 2 as the number of nodes increase. The
tree topology obviously exhibits a higher number of hops
than the other two topologies. This is due to the longer
route taken for the data from source to destination. There
are no alternative routes in this topology hence the path is
longer. In mesh topology the number of hops is lower than
the tree topology because there is always an alternative
route to reach the destination and this route is based on the
shortest path to destination. The maximum number of
hops in tree topology is expected to reach as high as 7
hops at 100 nodes but since the average number of hops is
considered it shows a maximum of 4. In star topology the
hop count is maintained at 2 and this number is the
average over the entire simulation run. It is obvious that as
node number increases the average hops will increase.
However, it is seen that when there are more than 100
sensor nodes, the hop number decreases. It appears that
the network is performing better as the number of nodes
increase. In fact, this takes place since higher data
collisions occur in crowded networks thus only some data
packets possess a probability of delivery to allow them to
reach the destination and hence only few nodes can send
data to the receiver successfully. Again the general trends
of these results agree with the results obtained in [2].
C. Global MAC Throughput
Global MAC throughput is the total data traffic in
bits/sec successfully received and forwarded to the higher
layer by the 802.15.4 MAC in all the nodes of the WSN. It
is known that throughput usually depends on many
aspects of networks such as power control, scheduling
strategies, routing schemes and network topology [8]. Fig.
6 shows the average global MAC throughput against the
number of nodes for all 3 simulation topologies. It can
clearly be seen that when the number of nodes increases
the MAC throughput increases. This is correct because the
data being received by the MAC layer increases.

Figure 4 Simulation results of average ETE delay of the WSN star


topology with and without RTS/CTS handshaking.

Figure 5 Average number of hops traveled by application traffic.

Figure 6 Average global MAC throughputs against the number of nodes.

Figure 3 Simulation of the average ETE delay for the mesh, star and
tree topologies.

This behaviour agrees in general with the results


presented in [8]. However when the number of nodes
increases above 60 in mesh and tree topologies, more
collisions will take place as the MAC layer cannot handle
the increased number of nodes. The throughput decreases
sharply if the number of nodes increases above 80 due to
access collisions. This throughput drop (10-40kbps) also
agrees in general with the results presented in [5]. There is
a slight difference between the throughput of the mesh

and tree topologies for the same number of nodes, this is


because the mesh topology supports multipath routes
which can cause a minor increase in the throughput. In the
star topology the throughput increases until it reaches a
maximum at 100 nodes and drops rapidly for higher node
numbers. This difference in throughput between star
topology and the other two topologies is not related to the
ETE delay explained earlier. This difference appears in
node numbers between 80 and 100. This is due to
excessive demand from all the nodes within their MAC
layers which will cause the MAC throughput to increase
momentary then after 100 nodes, due to excessive
collisions, the throughput drops rapidly.
D. Simulation Time
Time to run the simulation, generate reports and
statistics in OPNET varies as the numbers of sensor nodes
increase or decrease. It is apparent that increasing the
number of nodes leads to an increase in running time. This
increase in simulation time starts slowly when the number
of nodes is low then takes an exponential manner as the
number of nodes increase - this relationship is illustrated
in Fig. 7. The results indicate that a network with more
nodes is much more complex internally than a network
with fewer nodes and requires much more time to
generate reports by OPNET. At certain levels, simulation
times may take hours. Even in some real application
environments, 180 nodes is not considered a high number.
Comparing the simulation times of OPNET with NS-2 [2]
reveals that the latter has a much faster simulation time
than that of OPNET. However this does not mean that
OPNET Modeler lacks performance in this aspect but this
is due to the significant amount of data collected in
reports and statistics. The simulation time analysis will
assist in deciding which simulation tool is faster
especially when there are a very high number of nodes in
the WSN to be simulated.

IV.

IMPROVEMENTS SUGGESTED FOR OPNET MODELER

Simulation results show that OPNET is suitable for


simulating ZigBee WSNs with a wide range of available
statistics. Features that are useful for the applications of
WSNs include large number of reports and statistics that
are available across different network layers of the ZigBee
WSN, especially at the MAC layer where the quantity of
reports generated was not found in any other ZigBee
WSN simulator. However from this investigation it may
be suggested that it would be beneficial if a number of
features of OPNET were enhanced to make it more
suitable for the simulations and analysis of WSNs.
Improvements proposed for OPNET ZigBee model in
simulating WSNs include; multicast traffic, indirect
transmission, security and a model for energy. These are
currently the key aspects related to WSN implementation.
A. Multicast and Unicast Traffic
Multicasting means transmitting a single message
from one node to a selected group of nodes. Unicasting
means transmitting a single message from one node to
another node, e.g. coordinator node to a specific node [9].
In OPNET ZigBee WSN simulation only the message
broadcast is available, thus any option for multicasting or
unicasting is not available even though OPNET Modeler
configures multicasting and unicasting in the
IEEE802.11a, b protocols. Multicasting is essentially
important in simulating applications where control of the
coordinator over certain nodes is required. The results
presented in this paper represent generally the QoS
parameters using broadcast traffic. However the
evaluation of the ETE delay and throughput using
multicasting and unicasting could not be accomplished.
The option of multicasting and broadcasting traffic can be
done by adding or modifying the kernel procedure of the
OPNET ZigBee node model at the network layer so that a
particular node can transmit to one or more nodes in a
network with the same network ID.
B. Security Model

Figure 7 Simulation time for different number of sensor nodes

In some WSN applications simulating security issues


is essential and required especially when the simulated
WSN handles critical information. However, OPNET
simulator does not support or give an option for involving
security models in the simulation of WSN. An encryption
algorithm such as AES (Advanced Encryption Standard)
[5] can be used to encrypt data packets within the same
PAN. Having such encryption algorithm in a ZigBee
WSN will have an impact on the QoS performance of the
entire network since WSNs in general have limited data
rates and bandwidth recourses. The ETE delay and the
throughput simulation results presented here could not be
evaluated with the presence of AES. AES can be included
as an attribute in the application layer of OPNET Modeler
attributes by adapting the application layer kernel of the
OPNET ZigBee model, hence the affect of the AES
algorithm on the QoS parameters could be investigated
and studied in the simulated network.

C. Energy Model
Unlike NS-2 and other WSN simulation tools the
OPNET model for ZigBee networks does not support
energy models or the simulation of any energy related
aspects of WSNs [2]. Evaluating and simulating energy
resources sometimes is an essential factor for estimating
the life-time of a sensor node since the full operation of
the wireless node depends on internal battery power. The
simulation results presented here generally discuss the
QoS parameters with varying number of nodes. However
the WSN throughput will have an effect energy consumed
in the entire network. This evaluation could not be done
with OPNET Modeler due to the non availability of
energy model. The new energy model can be introduced
at the OPNET ZigBee process level at the MAC layer by
including a new kernel procedure for energy which
evaluates the energy consumed when the node is
accessing the channel and when it is in idle state. In so
doing the operator will then have a clear understanding of
node life-time under certain operating conditions and how
the QoS parameters are affected by the available energy
resources.
D. Contention Free Operation Mode
The OPNET model for ZigBee networks does not
support contention-free operation and slotted operation
mode. Generally, schedule-based protocols are
contention-free and hence energy waste caused by
collisions is eliminated. In addition to that, sensor nodes
require turning their RF transmitters and receivers on
during slots where data is to be transmitted or received
(i.e. carrier sensing). The sensor node can turn off its
transmitter and receiver in all other slots, thereby avoiding
overhearing. This results in low-duty-cycle node
operations, which may significantly extend the network
lifetime. Schedule-based MAC protocols have several
disadvantages [10] such as missing a reporting event
while the node is a sleep or the incorporated delays that
occur between the wakeup and sleep times. The effect of
contention free mode of operation on the QoS
parameters was not included here. The contention free
mode of operation can be included by adding a new
function at the MAC layer process model of OPNET
ZigBee model which allows the selection between the two
modes of operation.
V.

CONCLUSION

The purpose of this paper was to investigate the


performance capabilities of OPNET Modeler in
simulating ZigBee WSNs. In general the results presented
here show consistency with other software simulators of
WSNs. It can be concluded that OPNET has good
potential in simulating ZigBee WSNs since it can provide
a vast variety of reports and statistics at different network
layers (particularly at the MAC layer) for an individual
node or for the entire WSN. Further, it was established
that ZigBee WSNs are somewhat easier to deploy and
configure compared to other WSN simulators. The effect
of varying the number of nodes and the use of
handshaking on the performance the ZigBee WSN was

also demonstrated. However it was found that OPNET


ZigBee WSN does not perform well in the physical and
application layers since the essential energy and security
models are not incorporated in the simulation of the
ZigBee WSNs. Potential improvements were proposed to
further develop OPNET Modeler to compete with other
well known WSNs simulators. These improvements will
enhance OPNET Modeler to cover all aspects of WSNs
simulations and investigations for both researchers and
network operators.
REFERENCES
[1]
[2]

[3]

[4]
[5]
[6]
[7]

[8]

[9]

[10]

Becker et al, Comparative Simulations of WSN, ICT-Mobile


Summit 2008, Cunningham, P., 2008.
Hue et al, Performance Evaluation of NS-2 Simulator for
Wireless Sensor Networks, Proceedings of Canadian Conference
on Electrical and Computer Engineering, CCECE 2007,
Vancouver, BC, 2007, pp.1372-1375.
G. Lucio et al "OPNET Modeler and ns-2 - comparing the
accuracy of network simulators for packet-level analysis using a
network testbed", wseas transactions on computers, Vol. 2, No. 3.
(July 2003), pp. 700-707.
Curren, "A survey of simulation in sensor networks",
http://www.cs.binghamton.edu/ kang/teaching/cs580s/david.pdf.
Karl, Willig, "Protocols and Architectures for Wireless Sensor
Networks", John Wiley & Sons, Ltd, 2006.
OPNET official website, http://www.opnet.com.
S. Wu, Y. Tseng, "Wireless AD HOC Networking PersonalArea, Local Area, and Sensory-Area Networks", Auerbach
Publications, 2007.
Mathioudakis, I et al, "Wireless Sensor Networks: A case study
for Energy Efficient Environmental Monitoring", Proceedings of
Eurosensors conference 2008, Dresden, Germany, September
2008.
K Sohraby, D Minoli, T Znati, Wireless Sensor Networks
Technology, Protocols, and Applications, John Wiley & Sons,
2007.
Huang et al, OPNET Simulation of a Multi-hop Self-organizing
Wireless Sensor Network, Proceedings of OPNETWORK 2002
conference, Washington D.C, 2002.

You might also like