International Journal of Mobile Network Communications & Telematics (IJMNCT) Vol.2, No.
5, October 2012
DOI : 10.5121/ijmnct.2012.2501 1
ANADAPTIVE THROUGHPUT, SPECTRAL, ENERGY
EFFICIENT AND FAIR NETWORK CONFIGURATION
INPERVASIVE ENVIRONMENT
Debasish Chakraborty
1
, Basab B Purkayastha
1,2
and Kandarpa Kumar Sarma
1
1
Department of Electronics and Communication Technology, Gauhati University,
Guwahati, India
debasish0711@gmail.com, kandarpaks@gmail.com
2
Department of Physics, Indian Institute of Technology Guwahati, India
basab.bijoy@gmail.com
ABSTRACT
In this work, we propose a mechanism in which power consumption in a sensor network can be reduced
greatly by taking into account the fairness factors and to increase the probability of nodes for successful
packet transmission with lowest possible Signal to Noise Ratio (SNR) in a crowded environment. In other
words we propose a high throughput and spectrally efficient network. The power consumption of the
network reduces drastically when the clustered approach is adopted, which has been confirmed using the
Friss transmission formula. Here, we have used analytical expressions from renewal reward theorem to
deduce network throughput, probability of collisions and back-off periods upon collision in a random
multiple access environment. Using those expressions we have made an analysis on network throughput,
probability of collisions, bandwidth availability per node and estimation of back-off intervals upon
collisions. The total number of nodes in the network is kept constant for observational purposes. The
number of clusters has been varied to take observation of network throughput, probability of collisions,
bandwidth availability per node and estimation of back-off intervals upon collisions. The analytical results
obtained show that the power requirement of the network reduces drastically as the number of clusters
increases.
KEYWORDS
CSMA/CA, sensor network, throughput, collision probability, backoff time, power reduction ratio, clustered
network approach
1. INTRODUCTION
An ad-hoc network is a network which comprises of a number of nodes which are self dependent
and do not associate with any infrastructure based network. Ad-Hoc networks provide a variety of
services such as sharing of data and distributed services while sensor networks are used in
consumer as well as industrial applications such as process monitoring and control, environment
and habitat monitoring, healthcare applications home automation and traffic control [1]. They can
be deployed anywhere and their cost of implementation is low. An important characteristic of ad-
hoc networks is that they can configure themselves without any centralized control; a feature
which distinguishes itself from other networks which have infrastructure dependency. Taking this
International Journal of Mobile Network Communications & Telematics (IJMNCT) Vol.2, No.5, October 2012
2
into consideration we are proposing mechanisms that avoid problems associated with single point
failures and infrastructure based dependencies in sensor networks.
In this work, we have focused on the identification of problems in both ad-hoc and sensor
networks and provide such as low link throughput, power consumption, dependency of nodes on
one particular super node. In sensor networks, we need cluster-heads and pre-defined network
configuration among the clusters [13]. This often leads to common node failure. For e.g. if a
sensor head fails then no sensor belonging to it can forward the data. So to eliminate this
centralized dependency we have suggested a solution in which each node gets an opportunity to
act as the sensor head there by using distributed architecture. We describe methods to optimize
power required for transmission. Note that battery power is very important for a sensor as its life
is equal to the life of its battery. There are some related work [14-17] which deals with efficient use of
energy through clustering technique. Here we propose solutions on how to increase throughput of
such networks taking into account the fairness factors.
As the number of nodes in an ad-hoc network increases the probability of any particular node to
successfully communicate with acceptable bit error reduces significantly. The waiting period in
the underlying random access protocols tends to increase drastically [2]. Efficient switching
techniques are necessary to tackle such type of problems. In light of such problems we have
formulated measures to provide simple realizable solutions to tackle such problems without
bearing much cost and complexity in network design. We present here a mechanism in wireless
ad-hoc/sensor networks that reduces power consumption of the network drastically and at the
same time increase throughput significantly. Some of the relevant literatures are [2, 3, 4].
2. BACKGROUND PRINCIPLES
Ad-hoc networks using wireless communication are generally mobile which implies that the
primary energy resource is the battery. When the number of sensor nodes increases, the network
topology needs to be updated more frequently. If the frequency of a route is comparable to the
frequency of updates, then considerable amount of resource is drained. When the nodes of the
network are mobile the problem is compounded. In order to conserve power and spatial reuse, the
nodes reduce their transmission range.
Figure 1: Conventional Ad-Hoc Network
This, however, increases the hop count and hence there can be long delays for the packets to
reach their destination [5, 6]. When the range is too low, the network can be disconnected due to
International Journal of Mobile Network Communications & Telematics (IJMNCT) Vol.2, No.5, October 2012
3
increase in hop count. Therefore, the range must be sufficiently high to keep the network
connected. Apparently if the operational area of the network is kept fixed then the transmission
range is a function of the number of nodes in the network. There are many factors which cause
inefficiencies in a conventional ad-hoc network which increase tremendously when the number of
nodes increases [1, 6]. In light of the factors, we have made an attempt to propose a mechanism in
which all the problems can be addressed in a simple manner at a low cost of computation. The
diagram of a conventional ad-hoc network is shown in the Figure 1.
In ad-hoc networks, as the number of communicating node increases the probability that a node
will able to send packets decreases and in order to complete a successful transmission without
collision causes significant delay. So in order to achieve good link throughput the number of
nodes have to be reduced somehow. This is not a solution as it would lead to denying connection
to users who might be in an emergency. One way in which this problem can be tackled is to
reduce the transmission power of the nodes. This directly increases the reuse factor as greater
number of users can be accommodated in the available spectrum. This also drastically increases
throughput of the network. But the option of reducing the transmission range of the nodes
generates another problem which being the network overhead, ultimately reduces the packet
efficiency.
3. THE PROPOSED SENSOR NETWORK USING THE CLUSTERED APPROACH
Wireless sensor networks are those in which autonomous sensors monitor physical and
environmental conditions. The sensors pass their data cooperatively through the network to a
main location. The sensor networks are similar to ad-hoc networks and have no infrastructure.
The sensors are tiny and have very low processing power and memory. In sensor network
application, we need cluster-heads and pre-defined network configuration among the clusters. At
present the networks are bidirectional, enabling also to control the activity of the sensors. Energy
is the scarcest resource of wireless sensor networks and it determines the lifetime of the sensor
network.
Our main approach is to find an optimization of throughput, range reusability such that the nodes
in the network are able to transmit maximum number of packets with minimum power
consumption with reliability of communication being one of the foremost matters of importance.
Our main proposal is that the network be divided into clusters with each cluster containing a part
of nodes of the whole network. The range of each node is confined to the cluster boundaries and
each one is identified as belonging to a particular cluster. Among all the nodes in a cluster there
exists a special node which has a range higher than those belonging to that cluster and is called
the primary node. The rest are called secondary nodes. Thus the node distinction comes from the
range variability concept. Similarly all the clusters have the same story. The primary node of each
cluster communicates with the primary nodes of its neighbouring clusters and secondary nodes of
its own cluster and not with anyone else. The primary nodes make inter-cluster communication
possible and thus functions as a gateway forwarding packets of secondary nodes of its own
cluster. The primary nodes receive information from neighbouring primary nodes and relay the
relevant information to its secondary nodes. This approach takes care of the network overhead
problem that arises due to the breaking up of the network into clusters. It happens because the
primary nodes come into being when distance between the communicating nodes is high and
reduce the hop count substantially.
Each cluster has lesser number of nodes in comparison with number of nodes in the entire
network and hence the probability that a node will able to generate a packet and its successful
delivery will increase. Apparently, the bandwidth available per user (node) will increase which
leads to enormous increase of throughput of the whole network when the throughput of all the
International Journal of Mobile Network Communications & Telematics (IJMNCT) Vol.2, No.5, October 2012
4
clusters are taken in as a whole. Therefore, increasing the number of clusters increases the
throughput significantly. Some of the relevant literatures are [3] [4].
The main problem in implementing this scheme is the reluctance of secondary nodes to become
the primary node as it will have to forward packets of other node, thereby, hindering the
transmission of its own packets. So we have suggested all nodes in the cluster would take turns in
functioning as the primary node. This is accomplished by using a Voting Request Algorithm
(VRA).
The steps of the algorithm are:-
1. All nodes would advertise their remaining battery power and processor utilization at
regular intervals.
2. All other nodes would receive it and find out the node with the fittest remains.
3. All the nodes will share the fittest node information with all other nodes.
4. Node receiving maximum vote will be selected as the next cluster-head (primary node)
and the current head will hand over the charges to the new head.
Thus, there would be a periodic rotation of cluster-heads which would eliminate the problem of
single node failure and dependency on infrastructure based network.
Figure 2: Suggested Network Topology
The protocol used at the link layer to manage communications is CSMA/CA. The 802.11
specification, distributed random access standard called the distributed coordination function
(DCF) is used in our network for multiple accesses [7, 8]. The DCF random access procedures are
based on CSMA/CA mechanisms. The physical layer standard used for communication is
802.11a. Different band of frequencies are used for intra and inter cluster communication to
prevent interference between primary and secondary for inter traffic. The frequency band used for
secondary traffic is same for all the clusters. The primary nodes use a separate band from the
secondary nodes. All the primary nodes use same frequency for inter cluster communication and
each primary node communicates only with the neighbouring primary nodes to avoid
interference. This increases the spectral efficiency. When specified that each node is identified as
International Journal of Mobile Network Communications & Telematics (IJMNCT) Vol.2, No.5, October 2012
5
belonging to a particular cluster means that each cluster is assigned a particular range of IP
addresses. In the network, there would be a pre-defined number of clusters. The primary node
will possess all the IP addresses belonging to that cluster. There will be two categories of IP
address. The first category is used to identify the primary node and to carry inter-cluster traffic
while the second category to carry intra-cluster traffic. This means that only the primary node will
have two categories of IP addresses for inter and intra cluster communication while the secondary
nodes have one IP address for intra cluster communication. The nodes that arrive afterwards in a
particular cluster are allocated IP address from the primary node for intra cluster communication.
The distance vector routing [12] protocol will be used for secondary traffic routing and On-
demand routing for primary traffic routing which is based on the flooding mechanism [11]. Each
node will have two routing protocols for inter and intra cluster communication. The primary node
will be having two routing protocols activated at the same time and the secondary nodes will have
one routing protocol active and the second inactive.
When there is an exchange of node status from primary to secondary or vice-versa in a particular
cluster, then that particular cluster will be invisible momentarily for other primary nodes present
in the network. As a result, the primary traffic will be interrupted but the secondary traffic will
continue as usual. When a particular node moves from one cluster to the other then it will have to
surrender its IP address to the primary node of the cluster in which it belonged previously. When
a primary node detects a newly arrived node, it allocates an IP address to it and updates its
secondary routing table. The period during which a node moves from one cluster to another, the
entire communication whether uplink or downlink is completely terminated. However, the node
can resume communication and re-establish itself to its previous state if an application for the
same purpose is installed in it.
In sensor networks, energy is the scarcest resource and hence preserving it should be the primary
objective. The life of a wireless sensor node depends upon the life of its battery. Because of the
division of networks into clusters the distance between the transmitter and receiver has become
much lesser and the transmission range reduced. Correspondingly, the power requirement of the
nodes will be reduced according to the inverse square law of power variation with distance as
well as the power received by a node will be much higher in comparison to the conventional
approach.
Thus we have proposed a mechanism which improves the link throughput of the network, reduces
power consumption of the nodes, improves power gain and hence enhances the lifetime of the
nodes, improves the fairness of the system such that no node is over burdened and increases the
reliability of the network. These features are obtained by dividing the network into clusters.
4. MATHEMATICAL ANALYSIS
In this section, we specify the mathematical tools [8, 9, 13] and the expressions which we shall
use to calculate the results to demonstrate the validity of the claims that we have made in previous
section.
4.1. Throughput
We shall assume that the network is busy all the time and nodes are exchanging packets with one
another. We also assume that the back off time is exponentially distributed with mean 1/ .
Therefore, the average back off time is 1/ .
International Journal of Mobile Network Communications & Telematics (IJMNCT) Vol.2, No.5, October 2012
6
Let us consider T
0
as the packet over head time then in 802.11 standard :
o
T = RTS + CTS + ACK + 3SIFS + DIFS. (1)
where RTS(Request to send), CTS(Clear to send), ACK(Acknoledgement), SIFS(Short
Interframe Space) and DIFS (Distributed Coordination Function Interframe Space). If T
c
be the
time spent in a collision before next backoff period then in 802.11 standard:
c
T = RTS + DIFS. (2)
Now, let us consider n be the number of nodes, r the data rate, L be the packet length used by all
the nodes, as slot time and as the parameter of exponential backoff duration. Since we have
assumed that the backoff period is exponentially distributed the residual and fresh backoff times
are also exponentially distributed. Hence the time until the first backoff time is completed is
exponentially distributed with mean 1/n . The collision probability ( ) in a distributed random
acess standard distributed coorination function (DCF) is given by [6]:
( 1)
1
n
e
(3)
Now the mean time between successive renewal times is; the residual or fresh backoff time,
packet overhead time and payload time if there is a successful transmission, collision time if there
is a collision stated mathematically [6]:
T =
0
1
(1 )
c
L
T T
n r
_
+ + +
,
Hence, network throughput is given by [6]:
( )
0
(1 )
,
1
(1 )
c
L
L
T T
n r
_
+ + +
,
(4)
Now the throughput is maximized for:
4 1
1 1
2 ( 1)
c
c
nT
nT n
_
+
,
(5)
4.2. Back off Time
Let K be the maximum chances for a packet to attempt transmission before being discarded and
K
b the mean back off duration of a node after k
th
collision, then the random amount of cumulative
backoff time between successive successful packet transmissions or packet discards will be:
( ) ( ) ( ) ( ) ( ) 0 1 2 K
b b b b + + + .
The number of attempts a node makes between successive successful packet transmissions or
packet discards is given by:
2
1
K
+ + + + .
Therefore, the attempts parameter becomes:
2
2
0 1 2
1
K
K
K
b b b b
+ + + +
+ + + +
(6)
International Journal of Mobile Network Communications & Telematics (IJMNCT) Vol.2, No.5, October 2012
7
Assuming that K = and m such that:
min
2 1
. ,
2
k
K
CW
b
_
,
for 0 1 k m and
,
_
2
1 2
min
CW
b
m
k
for k m then:
( )
( ) ( )
min min
2 1 2 1
1 2 1 1 (2 )
m
CW CW
( +
(7)
The back off time is given by:
( )( ) ( ) ,
( )
2 1 2
2 1 1 2 1
.
1
min min
m
CW CW
(8)
4.3. Power Saving Formula
Let P
t
be the transmitted power, P
r
be the received power, A
er
the receiving antenna aperture, A
et
be the transmitting antenna aperture and G
t
as transmitting antenna gain. Then according to Friss
transmission formula the ratio of power received to power transmitted is given by [10]:
2 2
et er r
t
A A P
P R
(9)
When the network is divided into clusters R becomes R / M, where M is the number of clusters in
the axis of transmission. The power ratio of the clustered network is:
2
2 2
rc et er
tc
P M A A
P R
(10)
Therefore the power increases by a factor:
2 r c r
t c t
P P
M
P P
(11)
Therefore the power with which a node has to transmit for successful packet delivery is reduced
by 1/M
2
times in comparison with the conventional type network as P
tc
=P
t
/M
2
5. RESULTS AND DISCUSSION
The number of nodes is kept fixed at 600. Maximum number of nodes in one cluster assumed to
be 20. We also assume that the number of nodes in all the clusters is equally distributed. We have
analytically computed the network throughput, backoff time, collision probability, bandwidth
availability per using the formulas stated in the previous section. The observations were made by
varying the number of nodes. Since the total number of nodes in the network has been kept fixed
varying the number of nodes essentially means increasing the number of clusters.
The values considered for the following variables during computation are:
International Journal of Mobile Network Communications & Telematics (IJMNCT) Vol.2, No.5, October 2012
8
Data rate (r)=18 MBPS, packet length used by all the nodes (L) =1500 Bytes, slot time ( ) = 9s,
packet overhead time (T
0
) =598s,
c
T :RTS + DIFS =240 s, CWmin: = 32.
5.1. Collision Probability
The collision probability is calculated using eqn.(3). We assume that the throughput is
maximized, hence, is calculated using eqn.(5).
Figure 3: Collision probability vs. Number of clusters
In Figure 3, the curve shows how the collision probability is changing with respect to the number
of nodes and clusters. When the number of nodes reduces and the number of cluster increases
then the collision probability comes down.
5.2. Throughput
The throughput is calculated using equation (4). We have considered that is set to maximize
the throughput. The values of collision probability computed using (3) previously are used in the
equation (4) to compute the throughput.
Table 1: Network Throughput for different numbers of clusters
Number of Nodes Network Throughput (MBPS) Number of clusters
20 1.9100 30
25 1.9084 24
40 1.9062 15
50 1.9067 12
100 1.9035 6
200 1.9033 3
300 1.9028 2
600 1.9025 0
International Journal of Mobile Network Communications & Telematics (IJMNCT) Vol.2, No.5, October 2012
9
Figure 4:Throughput vs number of clusters
In Figure 4, the graph shows the throughput variation agaist the number of clusters. As the
number of clusters increase the number of nodes reduces and this in turn increases the probability
with which a node can complete a successful data transfer increases and this results in an
increased throughput.
5.3. Backoff Period
The backoff period is calculated using eqn.(8). The values of collision probability computed using
eqn.(3) are used in the equation (8) to compute the back off period.
Table2: Back off Periods for different numbers of clusters
Number of Nodes Backoff Period ( s) Number of clusters
20 172.744 30
25 172.950 24
40 173.200 15
50 173.370 12
100 173.538 6
200 173.630 3
300 173.700 2
600 173.754 0
International Journal of Mobile Network Communications & Telematics (IJMNCT) Vol.2, No.5, October 2012
10
Figure 5: Backoff period vs number of clusters
In Figure 5, the graph shows that the backoff period of the colliding nodes is getting shorter as the
number of nodes reduces. This is possible only when the number of clusters is increased.
5.4. Power Saving
The power reduction ratio is given by:
2
1
tc
t
P
P M
As the number of clusters increase the power requirement in the clustered network is greatly
reduced in comparison with the conventional network.
Figure 6: Power reduction ratio vs number of clusters in the axis of transmission
In Figure 6, from this graph it is evident that that power requirement of the network reduces
drastically as the number of clusters increase. This is obvious because the secondary nodes which
are more in number have a much lesser transmission range when compared with the conventional
network. However the fact to be taken care of is increasing the number of clusters increases the
number of primary nodes which dissipate more power. Hence the number of clusters must be
appropriately chosen so that the reduced power requirement of our proposal holds true.
International Journal of Mobile Network Communications & Telematics (IJMNCT) Vol.2, No.5, October 2012
11
6. CONCLUSIONS
In this paper a mechanism has been proposed for an ad-hoc/sensor network in which they are
broken up into clusters with each clusters consisting of two types of nodes. The node
classification is on the basis of range variation. Each cluster has a primary node and many
secondary nodes. The secondary nodes communicate among each other and use the primary node
which has a larger transmission range to carry their inter-cluster traffic. The proposal being made
increases the network throughput, reduce power consumption, improve reliability and increase the
fairness of the system as well. These claims have been verified by the results obtained by taking
help from the derived mathematical analysis. An important trade-off between the numbers of
clusters versus effect of interference on throughput has also been discovered which can be a very
important subject of study. This work can also be extended by making an analysis of throughput
in ad-hoc/Sensor networks using MIMO technology and using efficient scheduling algorithms.
The paper is pursued to specifically identify problems existing in wireless Ad-hoc sensor
networks and some solutions proposed to overcome such problems. However there are some
limitations that have been identified in the project work. The results prove that when the network
is spitted the benefits are enormous. The power consumption is reduced, the network throughput
and the bandwidth available per user increases. But the splitting cannot continue indefinitely
because these will lead to increase in the number of primary nodes.
With the increase in number of primary nodes, the power consumption of the network increases
as they have a higher transmission range and also forward packets from other nodes. Too many
primary nodes also increase the network overhead and this leads to the degradation of throughput
efficiency. Intra-cluster network is adversely affected because they will be processing requests
from nearby primary nodes to relay or receive packets. The secondary nodes in each cluster
which rely on the primary nodes for communication with nodes in other clusters have longer
delays for processing of their packets.
The above discussion points that the number of primary nodes in the network needs to be
controlled. Investigation of the adequate number of nodes can be pursued in the future. The
effects of introducing MIMO technology in the primary nodes can be pursued. When the nodes
are equipped with MIMO then the primary nodes can be increased and they will be able to
process more requests from the secondary nodes and at the same time take part in communicating
with nodes from neighbouring primary nodes. The spatial reuse increases significantly which in
turn increases throughput of the network as a whole. However in order to utilize these advantages
an efficient antenna orientation algorithm needs to be proposed which will result in minimum
collision among signals of various nodes.
REFERENCES
[1] HET-NETs 04: Second International Working Conference In Performance modelling and Evaluation
of Heterogeneous Networks Ilkley, West Yorkshire, U.K., 26-28th July, 2004,
[2] N. K. Ray and A. K. Turuk: A Review on Energy Efficient MAC Protocols for Wireless LANs,
Proceedings of the 4th International Conference on Industrial and Information Systems, ICIIS-2009,
pp. 137-142, Dec., 2009.
[3] D. Yu, K. T. Kim, B. Y. Jung, and H. Y. Youn: An Energy Efficient Chain-Based Clustering
Routing Protocol for Wireless Sensor Networks, Proc. International Conference on Advanced
Information Networking and Applications Workshops, pp. 383-388, Bradford, 2009.
[4] N.Vlajic and D.Xia. Wireless Sensor Networks: To Cluster or Not to cluster?, IEEE international
Symposium on world of Wireless, Mobile and Multimedia Networks (wowmom06), pp258-268,
Niagara-Falls/Buffalo, NY, June 2006.
International Journal of Mobile Network Communications & Telematics (IJMNCT) Vol.2, No.5, October 2012
12
[5] K. Pahlavan and A. H. Levesque, Wireless data communications, Proc. IEEE, vol. 82, no.9, pp.
1398-1430, Sept. 1994.
[6] Anurag Kumar, D. Manjunath, Joy Kuri, Communication Networking: An Analytical Approach,
Morgan Kaufman Series in Networking, (an imprint of Elsevier Science), May 2004
[7] IEEE Standard for Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY)
Specifications, Nov. 1997.
[8] T. S. Ho and K. C. Chen, Performance evaluation and enhancement of the CSMA/CA MAC protocol
for 802.11 Wireless LANs, in Proc. IEEE PIMRC, Taipei, Taiwan, , pp. 392396, Oct. 1996.
[9] G. Bianchi. Performance analysis of the IEEE 802.11 Distributed Coordination Function, IEEE
Journal on Selected Areas in Communications, vol.18, no.3, pp: 535547, March 2000.
[10] Antennas 2nd Edition by John D Kraus.
[11] Taek Jin Kwon; Gerla, M.; Varma, V.K.; Barton, M.; Hsing, T.R. Efficient flooding with passive
clustering-an overhead- free selective forward mechanism for ad hoc/sensor networks
Proceedings of the IEEE Volume: 91 , Issue: 8 Publication Year:2003,pp.1210-1220
[12] Perkins, C.E., Royer, E.M.: Ad-hoc on-demand distance vector routing, Second IEEE Workshop on
Mobile Computing Systems and Applications, WMCSA '99, pp.90-100, 1999.
[13] Martalo, M.; Ferrari, A simple information-theoretic analysis of clustered sensor networks with
decentralized detection, Communications Letters, IEEE, vol.14,issue. 6, pp.560 562, 2010
[14] Dali Wei, Yichao Jin,; Vural, S., Moessner, K. and Tafazolli, R.: An Energy-Efficient Clustering
Solution for Wireless Sensor Networks, Wireless Communications, IEEE Transactions on, vol.10,
no.11, pp. 3973-3983, Nov.2011.
[15] Ningbo Wang and Hao Zhu: An Energy Efficient Algorithm Based on LEACH Protocol,
International Conference on Computer Science and Electronics Engineering (ICCSEE), vol.2, pp. 339
342, China 2012.
[16] J. Kamimura, et. al., Energy-Efficient Clustering Method for Data Gathering in Sensor Networks," in
the Annual International Conference on Broadband Networks, October 2004.
[17] M. Ye, C. Li, G. H. Chen, and J. Wu, "EECS: An Energy Efficient Clustering Scheme in Wireless
Sensor Networks," International Journal of Ad Hoc & Sensor Wireless Networks, pp. 535-540, April
2005.
