Content uploaded by Hossein Anisi
Author content
All content in this area was uploaded by Hossein Anisi on Nov 09, 2017
Content may be subject to copyright.
Mobile Netw Appl
DOI 10.1007/s11036-016-0780-y
Mitigation of Packet Loss Using Data Rate Adaptation
Scheme in MANETs
Muhammad Saleem Khan1·Saira Waris1·Ihsan Ali2·Majid I. Khan1·
Mohammad Hossein Anisi2
© Springer Science+Business Media New York 2016
Abstract Node’s mobility, bursty data traffic, and dynamic
nature of the network make congestion avoidance and
control a challenging task in Mobile Adhoc Networks
(MANETs). Congestion results in high packet loss rate,
increased delays, and wastage of network resources due to
re-transmissions. In this paper, we propose In-route data rate
adaptation to avoid packet loss. Proposed scheme is based
on the analysis of queue length of the forwarding nodes,
number of data source nodes, and rate of link changes.
In proposed technique, queue length of forwarding nodes
is communicated periodically to the neighbor nodes using
existing control messages of the underlying routing proto-
col. Keeping in view the queue length of forwarding nodes,
number of data source nodes, and rate of link changes,
initially the intermediate nodes buffer the incoming data
Muhammad Saleem Khan
skhan.ciit@gmail.com
Mohammad Hossein Anisi
anisi@um.edu.my
Saira Waris
syra054@yahoo.com
Ihsan Ali
ihsanalichd@siswa.um.edu.my
Majid I. Khan
majid iqbal@comsats.edu.pk
1Department of Computer Science, COMSATS Institute
of Information Technology, Islamabad, Pakistan
2Faculty of Computer Science and Information Technology,
University of Malaya, Malaya, Malaysia
packets upto some threshold and then, gradually shift the
effect of congestion to the data source nodes. Then, the
source node adapts its sending data rate to avoid conges-
tion and to ensure reliable data communication. We have
performed simulations in NS-2 simulator by varying dif-
ferent network metrics such as data rate, number of source
nodes, and node speed. Results show that proposed tech-
nique improves network performance in terms of packet
delivery ratio upto 15 %, reduction of average end-to-end
delay and packet loss due to interface queue overflow upto
25 % and 14 % respectively, as compared to the static rate
adaptation scheme.
Keywords MANETs ·Congestion ·Packet loss ·Data rate
adaptation ·DRAS
1 Introduction
1.1 Background
Mobile Ad Hoc Networks (MANETs) consist of mobile
devices that are connected by wireless links. MANETs are
self-organizing networks which do not rely on any fixed
infrastructure. In MANETs, nodes can communicate with
other nodes directly which are in the communication ranges.
Nodes that do not lie in the transmission ranges of each oth-
ers communicate through forwarding nodes. The movement
of the nodes present in MANETs is arbitrary i.e. nodes can
leave and join the network anytime that leads to frequent
changes in topology of the network.
Due to mobility, lack of continues end-to-end connectiv-
ity, and dynamic network topology, reliable data delivery
becomes a challenging task in MANETs. When source
Mobile Netw Appl
node transmits data packets to the destination, any inter-
mediate node can suffer from congestion due to limited
resources. Congestion will prompt high packet loss, long
delays, and wastage of network resources. Issues like con-
gestion and non-availability of next hop is more common
in MANETs as nodes are free to move individually in any
direction resulting in frequent changing topology [1–4]. So,
the reasons for packet loss may be due to node mobility,
non-availability of next hop nodes, interface queue over-
flow, and so on. Moreover, as multiple sources are sending
frequent data, queue of forwarding node may overflow caus-
ing packet drops which leads to degradation of the network
performance.
1.2 Problem statement and motivation
Various techniques have been proposed to avoid packet loss,
such as adapting alternate route [5–8], congestion-adaptive
routing using bypass concept, multi-agent routing and rate-
based congestion control. Although these schemes find the
alternate path in case of congestion at nodes in the cur-
rent path. However, run-time calculation of alternate path
is an overhead in these schemes. Similarly, sending control
packets for congestion notification is itself a overhead for
a congested network. Moreover, congestion reports used in
alternate route adaptation techniques may be delayed which
can effect the network performance. Therefore, an efficient
congestion control mechanism is of vital significance in net-
works like MANETs. The essential targets of congestion
control mechanisms should be the best utilization of net-
work resources, reduction of delays, and improving network
performance. Moreover, the required mechanism should be
able to adapt the data rate at source nodes based on the
run-time network conditions. Similarly, in case of under
utilization, the required mechanism should also be able to
adapt the data rate to efficiently utilize the available channel
bandwidth.
Our Contributions In this paper, we propose In-Route
Data Rate Adaptation Scheme (IR-DRAS) to avoid packet
loss due to interface queue overflow. The proposed IR-
DRAS, avoids congestion before it actually happens. In the
proposed scheme, when source/neighbor node detect packet
loss due to congestion at forwarding node, data packets
are buffered at neighbor node of the congested node to
some pre-defined threshold and then the effect of conges-
tion is gradually shifted towards the source nodes. Based
on the adaptation factor which is computed using the run-
time network conditions, source node adapts the data rate
accordingly. Moreover, our proposed technique also con-
sider the channel under utilization case. If the channel is
under utilized, proposed IR-DRAS adapts the data rate by
increasing data rate at source nodes to better utilize the
available channel.
Our key contributions in this paper can be summarized as
follows:
– we evaluate how individual network condition parame-
ters affect the packet loss rate and data rate adaptation;
– we propose a distributed scheme for data rate adapta-
tion, based on several network parameters;
– we provide an implementation of our model in a real
routing protocol, the Optimized Link State Routing
(OLSR) protocol;
– we evaluate and compare our IR-DRAS with the
recently proposed trust scheme by Thakur et al. [20],
showing that the IR-DRAS outperforms such scheme.
The rest of the paper is organized as follows. Section 2
presents related work. Section 3describes our proposed
technique. Performance evaluation is discussed in Section 4
and finally conclusion and future work is presented in
Section 5.
2 Related work
Packet loss due to congestion or queue overflow is a severe
problem in MANETs. Various techniques have been pro-
posed to avoid congestion which can be classified from
different aspects. Based on the methodology these schemes
use to avoid congestion, can be classified as alternate path
based [5–7], data rate adaptation based and pausing control
messages [9,10] schemes.
Techniques proposed in [5–8] work on the basis of
finding alternate path after detecting congestion.
A technique to control congestion on the basis of prior-
ity has been introduced in [5]. In the proposed technique,
the overall network traffic is divided into different classes,
such as real time traffic and normal network traffic. Differ-
ent queues for each type of traffic is implemented. Every
queue has different data rate adjustment to control conges-
tion so that data with high priority (real time traffic) will be
send first.
Data rate adaptation based techniques is the another cate-
gory of the congestion control in MANETs. This methodol-
ogy works for congestion control and based on rate control
mechanism. Several techniques are proposed for conges-
tion control using data rate adaptation. These techniques
can be further categorized as (a) Feedback-based data rate
adaptation and (b) Queue-based data rate adaptation. In the
following, we discuss these schemes in details.
Techniques proposed in feedback-based data rate adap-
tation category are based on feedback mechanism for
adaptation of data rate. These techniques either use con-
secutive ACKs or explicit notification bits in headers for
Mobile Netw Appl
congestion awareness and congestion information is sent to
source node using ACK packet. To control congestion and
improve fairness in the network, Congestion Control and
Fairness (CCF)scheme was proposed by Ee and Bajcy [11].
In the aforementioned protocol, to fairly share the avail-
able bandwidth among adjacent downstream links, a tree
structure is used at the root of a routing sub-tree. Conges-
tion is prevented in CCF as the senders are not allowed to
exceed its assigned bandwidth. One of the shortcomings of
the aforementioned scheme is that a simplistic technique
is used for fairness which may not be applicable to prior-
ity based multimedia traffic. In [12] data rate is increased
if there are 10 consecutive successful transmissions and
decreased after 2 consecutive transmission failures. Success
and failure is evaluated on the basis of received ACK pack-
ets. This scheme is not taking into account the reasons of
transmission failures. Data rate is adapted whatever is the
reason for transmission failure. Auto-rate fallback (ARF)
technique [13] is based on the reception of consecutive
ACKs. If two consecutive ACKs are not correctly received,
then lower data rate is used for next re-transmission. In
case when 10 consecutive ACKs are received success-
fully, then next transmission takes place at higher data rate.
This scheme is easy to implement as it is only based on
ACKs and timer. However, data rate is adapted without tak-
ing into account the reason of transmission failure. ARF
technique was enhanced and adaptive multi-rate auto rate
feedback (AMARF) technique was proposed in [14]. In
this technique, success threshold is assigned to each data
rate. Threshold is dynamically changed according to run-
ning conditions like channel parameters and packet length.
Success threshold is used to shift between different data
rates. Simulation results show that this scheme achieves
good throughput. But this scheme does not take into account
the competing nodes, so fairness is not considered, hence,
not applicable for MANETs. End-to-end congestion control
mechanism [1,15,16] are proposed to control congestion
in MANETs. In end-to-end congestion control mechanism,
ACK packets are send by destination node to communicate
congestion status to the source nodes.
Another way of providing congestion feedback to source
node is explicit congestion notification bits in header of
the data packet. Techniques proposed in [17,18] are using
the explicit congestion notification (ECN). In these tech-
niques, once the data is reached at the router, load factor
is calculated by the router for all the links. Based on the
computed load factor, congestion region is identified in the
network. If calculated load factor is comparatively higher,
then congestion status is updated by overwriting ECN bits.
Technique proposed in [18] uses two ECN bits for improv-
ing results. In this technique, data rate is changed by a static
factor. Another technique in which successful packet trans-
mission and buffer threshold both are considered is additive
increase and multiplicative decrease (AIMD) scheme [19]
is proposed. In this scheme, for its every successful packet
transmission, data rate is increased with increasing param-
eter and continue until buffer threshold is received from
other side and data rate is decreased when packet trans-
mission failed. Failure of data transmission is measured by
not receiving ACK packets. Due to congested route, deliv-
ery of ACK packet might be delayed that cannot represent
the actual status of route. Several techniques have been pro-
posed to avoid congestion by sending control messages from
the forwarding nodes to source nodes. These control mes-
sages are sent by sensing queue length at node level. Queue-
based data rate adaptation is performed in [20–23]. These
techniques use queue length as a parameter to judge conges-
tion. Probability of accessing the communication channel is
calculated by each node based on the number of unsuccess-
ful transmissions in [23]. In addition, each node receives a
hello message periodically from its neighbors containing the
channel access probability, transmission rate, and the esti-
mated traffic load. Reinforcement learning is used by each
node to analyze the channel access probability. Thus, previ-
ous actions are used to decide either it is necessary to update
the transmission rate or not. Negative impact of updating the
transmission rate unnecessarily is mitigated using this tech-
nique. Moreover this technique also take in account the load
on each node calculated by its queue length and then this
information is used to decide whether to increase, decrease
or keep its transmission rate unchanged. In this paper, no
specific mechanism is discussed about factor by which data
rate is increased or decreased. We assume it a static factor.
In [21], a technique is proposed to control congestion in
proactive protocols by generating a Packet Error Announc-
ing Message called PEAM messages. The limitation of
these techniques is initiating control message on congested
route is itself a overhead for the network.
To adapt data rate at sender node, a technique is proposed
in [20]. This technique is based on queue length analysis at
intermediate nodes. In this technique, queue length is cat-
egorized in classes to adapt data rate accordingly. When
congestion occurs, the intermediate nodes notify the source
node by sending Congestion Indication Packet (CI Packet)
explicitly. According to CI Packet, the sender node ulti-
mately reduces its sending data rate as indicated in packet.
In this way, congestion can be avoided and a reliable com-
munication within MANETs is ensured. Although, afore-
mentioned technique overcomes the problem of queue over-
flow, however, the data rate is adapted with fixed percentage
without analyzing the other network conditions. Data rate is
adapted at source nodes only and no adaptation at interme-
diate nodes. Similarly, channel under utilization case is not
considered in this scheme. Moreover, to communicate con-
gestion to the source node, extra control packets are used
which is an overhead in the already congested network. To
Mobile Netw Appl
overcome the limitation of proposed techniques, we have
proposed packet loss avoidance technique which dynami-
cally adapts data rate at initially at intermediate nodes on
the basis of current queue length and then gradually shift the
congestion effect to source nodes. In the proposed scheme,
we are considering under utilization of forwarding node’s
queue to better utilize the network resources. Our proposed
DRAS scheme reduces or maximizes the data rate based on
the adaptation factor computed using the run-time network
conditions in the network. Also, the data rate adaptation
does not take place at the source nodes directly when con-
gestion occurs but the effect is gradually shifted towards
the source node by queuing the packet at neighbor nodes
of congested node up to pre-defined threshold. Moreover,
existing periodic control messages are used to communicate
the congestion notification without using the extra control
messages to avoid the overhead.
We originally proposed the DRAS in [24]. In our previ-
ous work, we proposed data rate adaptation scheme which
adapts the data rate directly based on the run-time network
conditions at the source node upon receiving congestion
notification. Moreover, the previous work does not consider
the channel under utilization case. In this work, we address
some shortcomings and extend our previous work by (i)
providing the in-route data rate adaptation at intermediate
nodes which gradually shifts the effect of congestion to the
source nodes, (ii) consider the channel under utilization case
by increasing the data rate at source nodes, (iii) providing
experimental comparison with the state-of-the-art protocol
proposed by Thakur et al. [20], and (iv) providing additional
experiments for performance analysis.
3 In-route data rate adaptation scheme
(IR-DRAS)
Techniques using data rate adaptation at source node results
in wastage of network resources. To avoid the negative influ-
ence of these techniques on network, an In-Route Data Rate
Adaptation Scheme (IR-DRAS) is proposed to avoid packet
loss caused due to interface queue overflow in MANETs.
In this scheme, unlike feedback-based rate adaptation, data
rate is adapted only if necessary as we are using queue
length of the nodes as a metric for data rate adaptation
which is an indication of congestion. Moreover, data rate is
adapted at intermediate nodes for better utilization of net-
work resources. Data rate is adapted dynamically by consid-
ering the network parameters having significant influence
on the data rate. In this paper, we have identified the net-
work parameters that are critical in order to adapt the data
rate and analyze their relationship to the network dynamics.
Then, we discuss how such network parameters affect the
data rate.
3.1 Network dynamics and data rate adaptation
In this section, we first identify the various network param-
eters that have significant effect on the data rate of nodes
and then leverage these parameters to compute the data rate
adaptation factor ρ.Theρis a factor which means how
much data rate should be adapted (increase or decrease)
based on the run-time network conditions.
Packet loss caused due to interface queue overflow is
influenced by the factor of mobility, queue length, and num-
ber of data source nodes in the network. We are using term
queue length which refers to the number of packets in node’s
queue or buffer at time t).
3.1.1 Queue length (ψ)
In the proposed scheme, queue length of a forwarding node
is being used as data rate adaptation parameter by immedi-
ate neighbors and source nodes. As mentioned previously,
forwarding nodes will periodically exchange their queue
status with neighbors. With the increasing data rate, there
are high chances that node’s queue may overflow which
results in high packet loss ratio. So, to avoid packet loss,
data rate should be reduced as the node’s queue length is
increasing. In the following equation, we define the relation-
ship between queue length and data rate adaptaion factor ρ:
ρ∝1
ψcurrent
,(1)
where ψcurrent represents the current queue length of a
forwarding node. The minimum possible value for queue
length is 0 which means that node’s queue is empty. Fol-
lowing formula helps us to find optimal rate adaptation
value at source/immediate neighbor nodes with respect to
the received queue length from forwarding node.
ρψ=1−ψcurrent
ψmax
,(2)
where ψmax is the maximum possible queue length of a
forwarding node.
3.1.2 Number of sources (ξ)
Number of data source nodes is another parameter for data
rate adaptation. Number of source nodes means that how
many data source nodes are there sending data to a particular
forwarding node. When a particular node is not a forwarding
node of any single source node, this parameter has minimum
value of 0. On the other hand, a forwarding node has max-
imum number of source nodes if all one-hop neighbors are
sending data to it. To find the optimal rate adaptation value
at node N for avoiding packet loss caused due to queue over-
Mobile Netw Appl
flow with respect to number of source nodes, we use the
following expression:
ρξ=1−SourcesN
nbmax
(3)
where SourcesNis current number of source nodes of a node
Nand nbmax is maximum possible source nodes of a node N.
3.1.3 Rate of link changes (η)
Rate of link changes in the evaluating node neighborhood is
one of the basic parameters used in our proposed scheme to
adapt the data rate. In one of our previous works [25,26], we
used the rate of link changes parameter for the adaptation
of trust threshold and trust update frequency. In this work,
we use the rate of link changes parameter to compute the
optimal data rate adaptation factor.
Link changes in MANETs at particular node occur when
a node arrives or leaves the neighborhood. Most of the
times, link changes are caused by the node mobility in the
network. To compute the neighborhood dynamicity of a par-
ticular node E, rate of link changes (η) [27] is an important
parameter. Rate of link changes at node Ecan be computed
using the following equation:
ηE=λE+μE,(4)
where λErepresents the number of new links established
and μEis the number of links that are broken experi-
enced by the node Eduring the particular time interval
[28]. According to the above equation, minimum rate of link
changes ηEmin will be 0, if a node Eexperiences no new
arrival and no link breakage. In this case, the network will
tends to static with no node mobility. Similarly, maximum
possible link breakage μEmax occurs at node Ewhen all
the 1-hop neighbors move out of the node Eneighborhood.
Based on the evaluation presented by Samar et al. [28],
the link arrival rate λEmax is equal to the link breakage
rate in MANETs. In view of the aforementioned findings,
maximum link change rate (ηEmax)is formulated as:
λEmax +μEmax =2·σE,
where σErepresents the node degree (number of neighbor
nodes) of node E. Based on the above formulation for rate
of link changes, we can compute the optimal factor for data
rate adaptation in terms of rate of link changes ρηusing the
following equation:
ρη=ηE
2·σE
.(5)
3.2 Mathematical model
We can combine the equations introduced so far into a math-
ematical model to compute data rate adaptation factor ρto
avoid packet loss caused due to interface queue overflow.
By combining Equations 2, 3 and 5, we obtain:
ρ=αρψ+βρξ+γρη
α+β+γ,whereα +β+γ=3.(6)
In above equation, α,β,andγare the weights assigned
to each data rate adaptation parameter, discussed previously.
As we are considering every parameter equally important
for data rate adaptation, weights assigned to each parameter
is equal, i.e. 1.
3.3 Working of IR-DRAS protocol
Algorithm 1 presents the working functionality of proposed
IR-DRAS. In proposed IR-DRAS based protocol, data rate
is adapted if the reason for packet loss is queue overflow
as we are monitoring queue length and considering it as an
indication of congestion in the network. In IR-DRAS, queue
length of each node is communicated to the neighboring
nodes periodically. On the basis of the queue length, sta-
tus of congestion at forwarding nodes is monitored. Queue
length is communicated to neighbor nodes periodically by
using reserved bits in hello messages of underlying rout-
ing protocol. In case of congestion, our scheme adapt data
rate initially at intermediate nodes by buffering the packets
and slowing down the en-queuing rate. When intermediate
node’s queue about to overflow, the effect of this conges-
tion is shifted gradually towards the source node. Source
node then adapt the data rate upto the optimal rate adap-
tation factor which is computed based on the run-time
network conditions. A minimum and maximum threshold is
set for queue length. Queue length is compared to thresh-
old and if queue length lies in range (between maximum
and minimum threshold) then data transmission is contin-
ued unhindered. Otherwise, data rate is increased in case of
under utilization of queue length and decreased when queue
length received from neighbor nodes is beyond maximum
threshold (queue overflow scenario).
Tabl e 1 Simulation parameters
Parameter Value
Simulation time 1000 seconds
Number of nodes 50
Network size 1000m ×1000m
Transmission range 250m
Packet size 512b
Queue length 50
Mobility model Random way point
Traffic type Constant bit rate (CBR)
Max speed 1–10m/s
Source-Destination pairs 10–50%
Data rate 2–10 packets/sec
Mobile Netw Appl
Please Replace the whole algorithm with the the follow-
ing given tex code:
4 Performance evaluation
In this section, we present the simulation setup, performance
metrics, and simulation results to evaluate the proposed
scheme in comparison to one of the recent data rate adap-
tation scheme [20]. Proposed technique is evaluated on
Network Simulator (NS2). The Optimized Link State Rout-
ing (OLSR) protocol is used as routing protocol. Table 1
shows the simulation parameters.
We have evaluated our proposed In-Route Data Rate
Adaptation scheme referred as “IR-DRA-scheme” in graphs
in comparison to our previously proposed scheme [24]
referred as “DRA-scheme, to a baseline protocol without
any data rate adaptation scheme referred as “WRA-scheme”
and to a static rate adaptation scheme [20] referred as
“SRA-scheme”. We have evaluated our proposed scheme by
varying network parameters i.e., under varying data rate and
node speed. The data rate has a significant effect on over-
all network performance. High data rates can cause queue
overflow and packet loss which leads to the degradation
of the network performance. On the other hand, low data
rates result in low throughput and long delays. So routing
protocol should utilize the network resources efficiently by
adapting the data rate for better network performance. Sim-
ilarly, to evaluate the impact of mobility, we have analyzed
the proposed scheme with varying node mobility. Mobility
in the network has a significant effect on overall network
performance. High mobility can cause more packet loss
which leads low packet delivery ratio.
4.1 Packet delivery ratio (PDR)
Figure 1a shows the effect of increasing data rate on PDR.
As data rate increases, chances for packet loss are higher
because queues of forwarding node overflows frequently.
So, delivery probability decreases with increased data rate.
The WRA scheme is not adapting data rate to avoid packet
loss, while SRA scheme adapts the data rate statically
without considering other network conditions. Moreover, in
DRA scheme, the congestion notification is sent directly to
the source nodes without adapting the rate at intermediate
nodes. On the other hand, IR-DRA scheme initially adapts
the data rate at intermediate nodes by queuing the packets
to maximum bearable threshold and then gradually shift the
effect towards the source nodes. This phenomena not only
avoids the packet loss at forwarding nodes but also keep the
data rate constant as the source nodes are not required to
slow down the data rate immediately. Then, source nodes
adapt the data rate based on the factor computed on the
run-time network conditions. So due to the in-route storage
mechanism and gradual shift of congestion effect to source
nodes, IR-DRA scheme performs better and achieves higher
PDR as compared to the other schemes.
Similarly, Fig. 1b shows the effect of node mobility on
PDR. For increasing node mobility, it results in frequent for-
warding node queues overflow as nodes change their posi-
tion frequently and forwarding nodes need to store packets
for longer time in their queues. So, the chances of packet
loss become higher. As there is no data rate adaptation tech-
nique in the WRA scheme, with increasing mobility, there
is more packet loss which cause lower PDR. The IR-DRA
scheme has bit higher PDR as compared to DRA scheme
and SRA scheme. The reason is that in-route storage mech-
anism in IR-DRA scheme. Due to this mechanism, source
nodes are not immediately required to cut-off the data rate
upon congestion notification, so more data packets are sent,
hence more PDR as compared to the other schemes.
As shown in Fig. 1c, IR-DRA scheme achieves higher
PDR as compared to other schemes. The reasons are the
same as already explained in Fig. 1a. The IR-DRA scheme
is considering the run-time network parameters which
results in better data rate adaptation and avoid packet loss
efficiently. Overall, PDR is decreasing with increasing num-
ber of source nodes because increased number of source
nodes results in queues overflow of forwarding nodes as
more data sources are sending data to forwarding nodes. So,
the chances of packet loss become higher which results in
low PDR.
4.2 Throughput
As data rate increases, throughput is also increases because
increased data rate means number of packets generated
per unit time increases. As shown in Fig. 2a, initially, IR-
DRA scheme shows the high throughput as compared to
other schemes. The reason is the use of mechanism which
consider the channel under utilization case. Due to this
mechanism, at low data rate such as 2–4 packet/second, the
Mobile Netw Appl
50
55
60
65
70
75
80
85
2 4 6 8 10
PDR (%)
Data rate (packet/sec)
WRA-scheme
SRA-scheme
DRA-scheme
IR-DRA-scheme
(a)
50
55
60
65
70
75
80
85
2 4 6 8 10
PDR (%)
Maximum node speed (m/sec)
WRA-scheme
SRA-scheme
DRA-scheme
IR-DRA-scheme
(b)
50
55
60
65
70
75
80
85
90
10 20 30 40 50
PDR (%)
No. of sources (%)
WRA-scheme
SRA-scheme
DRA-scheme
IR-DRA-scheme
(c)
Fig. 1 Effect of data rate, node speed, and No. of sources on throughput
queues of forwarding nodes tend to empty and still there is
more capacity to forward the data packet at higher data rate.
So, as per the current status of forwarding nodes queues,
data rate at source nodes are increased to fully utilize
the underlying communication channel capacity. When the
forwarding nodes queues overflow at higher data rate, such
as 6–10 packet/second, the data rate is decreased at source
nodes, hence low throughput, but still comparable to other
schemes. Moreover, WRA shows highest throughput as
compared to the DRA scheme because it sends data with
constant data rate without taking congestion in account. If
congestion occurs, WRA scheme does not have any mech-
anism to deal with such situation. Similarly, SRA scheme
minimizes the congestion effect to some extent using static
parameters. On the other hand, IR-DRA adds more parame-
ters and adapts data rate in a better way. In adaptation phase,
DRA is not utilizing whole bandwidth of the channel which
results in decreased throughput that is why throughput in
DRA scheme is slightly lower as compared to the WRA
scheme.
Similarly, Fig. 2b shows the average throughput for
increasing node speed. Due to increasing node mobility for-
warding nodes may unable to send packets due to unreach-
able destination nodes. As shown in aforementioned figure,
the IR-DRA scheme has higher throughput at low node
speed as compared to other schemes. The reason is the con-
sideration of channel under utilization case in the IR-DRA
scheme. Overall, the throughput decreases with increasing
node mobility. Moreover, the WRA scheme shows high-
est throughput at higher node speed, such as 6–10m/sec,
as compared to other schemes because WRA scheme sends
data with constant data rate without taking congestion into
the account and rate adaptation. Due to data rate adaptation
mechanism in the IR-DRA, DRA and SRA schemes, the
data rate is decreased at the source nodes to avoid packet
loss.
Figure 2c shows throughput for increasing number of
data sources. As shown in figure, when initially, the num-
ber of data sources are small, IR-DRA scheme achieves
comparable throughput with WRA scheme and higher
than SRA and DRA scheme. The reason is that with
smaller number of data sources, the queues are under uti-
lized, so the IR-DRA scheme increase the data rate to
overcome the channel under utilization case. However,
throughput in DRA scheme is slightly lower then WRA
and IR-DRA scheme because with increasing number of
source nodes, data rate is reduced to avoid packet loss.
In adaptation phase, DRA is not utilizing the whole band-
width of the channel which results in lower throughput
values.
60
80
100
120
140
160
180
200
220
240
2 4 6 8 10
Average Throughput (kbps)
Data rate (packet/sec)
WRA-scheme
SRA-scheme
DRA-scheme
IR-DRA-scheme
(a)
140
160
180
200
220
240
260
2 4 6 8 10
Average throughput (kbps)
Maximum node speed(m/sec)
WRA-scheme
SRA-scheme
DRA-scheme
IR-DRA-scheme
(b)
60
80
100
120
140
160
180
200
220
10 20 30 40 50
Average Throughput (kbps)
No. of Sources (%)
WRA-scheme
SRA-scheme
DRA-scheme
IR-DRA-scheme
(c)
Fig. 2 Effect of data rate, node speed, and No. of sources on PDR
Mobile Netw Appl
4.3 End-to-end delay
End-to-end delay increases for increasing number of
packet loss as packet loss results in large number of re-
transmissions, hence increases end-to-end delay. As shown
in Fig. 3a, the end-to-end-delay is comparatively higher
in IR-DRA scheme at high data rate, such as 6–10
packet/second. The reason is that at high data rate, there
are more congestion at forwarding nodes, so due to in-
route storage mechanism at intermediate nodes in IR-WRA
scheme causes bit more delay in comparison to the SRA and
DRA scheme but still lower than WRA scheme. In WRA
scheme, as packet loss is high, so it results in long end-to-
end delays as shown in Fig. 3a. The DRA scheme has lower
end-to-end delay as compared to WRA and SRA scheme
because data rate is efficiently adapted with low packet loss.
Moreover, as DRA scheme is adapting data rate to avoid
packet loss which means smaller number of packets in net-
work will be communicated, hence avoid congestion. As
congestion is avoided, so end-to-end delay also decreases in
proposed scheme due to smaller number of re-transmissions
for lost packets.
Similarly, Fig. 3b shows the impact of mobility on aver-
age end-to-end delay. Due to high mobility, data is not
delivered to distant nodes. Where as, data is delivered speed-
ily to nearer nodes that is why end-to-end delay decreases
with increasing node mobility. The IR-DRA scheme reduces
delivery delay as compared to the DRA and SRA scheme
as shown in Fig. 3b because data rate is efficiently adapted
which reduces packet loss. Hence, large number of re-
transmissions are not required which effects end-to-end
delay positively. The WRA scheme shows highest end-to-
end delay among other schemes as these schemes are not
considering rate adaptation to avoid packet loss. Hence,
large number of re-transmissions are required which ulti-
mately results in long delays. Also, in WRA scheme, send-
ing data with constant rate without adapting any mechanism
to avoid congestion. So, with occurrence of congestion,
end-to-end delay increases.
Impact of increasing number of data source nodes on
end-to-end delay is shown in Fig. 3c. The IR-DRA and
DRA schemes have almost comparable end-to-end delay.
Although, the in-route storage mechanism in IR-DRA
scheme increase the end-to-end delay, however, impact is
normalized because of low number of packet loss, hence
lower re-transmissions. In the IR-DRA scheme, data rate is
efficiently adapted which reduces packet loss, so chances
of congestion occurrence is minimized. Moreover, overall
the end-to-end delay is increases with increasing number of
data source nodes which means that now relay nodes may
need to forward more data in the network. So, chances of
congestion occurrence increases with increased number of
data source nodes, hence end-to-end delay increases.
4.4 Packet loss
Impact of increasing data rate on packet loss is shown in
Fig. 4a. It is obvious from the figure that IR-DRA scheme
has lower packet loss rate as compared to the other schemes.
The reason is the efficient in-route storage mechanism and
run-time data rate adaptation based on the different network
factors. Overall, as shown in Fig. 4a, packet loss due to
interface queue overflow increases for increasing data rate
because queues of forwarding nodes overflow due to high
data rate. Moreover, Fig. 4a shows that packet loss in WRA
is higher as these schemes do not consider any rate adapta-
tion to avoid packet loss. Similarly, SRA scheme adapts the
data rate statically without considering the run-time network
conditions. On the other hand, DRA scheme adapts data rate
efficiently by considering mobility factor and number of
source nodes along with queue length which causes smaller
packet loss as compared to the WRA and SRA scheme but
has higher packet loss rate as compared to the IR-DRA.
The reason is that the DRA scheme lacks the in-route stor-
age mechanism and data rate is adapted by the source node
directly which takes time to take actions based on the con-
gestion notification, hence cause more packet loss unless
and until data rate is adapted.
0
200
400
600
800
1000
1200
1400
2 4 6 8 10
Average End-to-End Delay (ms)
Data rate (packet/sec)
WRA-scheme
SRA-scheme
DRA-scheme
IR-DRA-scheme
(a)
0
100
200
300
400
500
600
700
800
900
2 4 6 8 10
Average end-to-end delay (ms)
Maximum node speed (m/sec)
WRA-scheme
SRA-scheme
DRA-scheme
IR-DRA-scheme
(b)
0
200
400
600
800
1000
1200
1400
10 20 30 40 50
Average End-to-End Delay (ms)
No. of Sources (%)
WRA-scheme
SRA-scheme
DRA-scheme
IR-DRA-scheme
(c)
Fig. 3 Effect of data rate, node speed, and No. of sources on end-to-end delay
Mobile Netw Appl
0
5
10
15
20
25
30
2 4 6 8 10
Packet Loss (%)
Data rate (packet/sec)
WRA-scheme
SRA-scheme
DRA-scheme
IR-DRA-scheme
(a)
2
4
6
8
10
12
14
16
18
20
22
2 4 6 8 10
Packet loss (%)
Maximum node speed (m/sec)
WRA-scheme
SRA-scheme
DRA-scheme
IR-DRA-scheme
(b)
2
4
6
8
10
12
14
16
18
20
22
10 20 30 40 50
Packet loss (%)
No. of sources (%)
WRA-scheme
SRA-scheme
DRA-scheme
IR-DRA-scheme
(c)
Fig. 4 Effect of data rate, node speed, and No. of sources on packet loss
To analyze the impact of node mobility on packet loss,
Fig. 4b shows the packet loss for increasing node mobility.
Due to node mobility, forwarding node queues are overflow
as nodes change their position frequently and forwarding
nodes need to store packets for longer in their queues.
Figure 4b shows that packet loss in IR-DRA scheme is
lower in comparison to all the other schemes under the all
node speed values because of the efficient data rate adapta-
tion mechanism. Pakcet loss in WRA scheme is higher as
this scheme does not consider any strategy to avoid packet
loss. Similarly, the DRA scheme adapts data rate efficiently
based on multiple network factors to avoid packet loss but
still higher packet loss rate than IR-DRA scheme.
Figure 4c shows the effect of increasing number of nodes
on packet loss rate due to interface queue overflow. It is
obvious from the figure that IR-DRA scheme has lower
packet loss rate in comparison to the other schemes. The
reason is the efficient data rate adaptation mechanism as
discussed previously. As WRA scheme is not using any
strategy to avoid packet loss so packet loss using WRA
scheme is maximum as shown in Fig. 4c. Moreover, increas-
ing number of data source nodes results in high packet loss
ratio because queues of forwarding node may overflow fre-
quently as more number of data sources are sending data to
the forwarding nodes.
5 Conclusion and future work
In this paper, we proposed a technique to adapt the data
rate at sender nodes and intermediate nodes based on the
run-time network conditions around forwarding nodes. Pro-
posed scheme adapts the data rate at intermediate nodes
based on the mentioned parameters by buffering the data
packets in queues up to bearable threshold, and then grad-
ually shift the effect of rate adaptation to the source nodes.
Proposed technique is based on the analysis of queue length
of the forwarding nodes. The queue length of forwarding
nodes is communicated periodically to the neighbor nodes.
Based on the current status of the queue length of for-
warding node, the sending node adapt its sending data rate
to avoid congestion and to ensure reliable communication
among nodes. We have simulated our proposed technique in
NS-2 and achieved better results in terms of packet deliv-
ery ratio and average end-to-end delay in comparison to the
static rate adaptation technique.
In future, we also plan to evaluate the proposed scheme
in VANETs scenario under different performance metrics to
test the compatibility and scalability of the proposed scheme
in high dynamic environment.
Acknowledgments The work reported in this paper has been par-
tially supported by Higher Education Commission (HEC), Pakistan,
and the University of Malaya under UMRG grant RG325-15AFR.
References
1. Chen K, Nahrstedt K, Vaidya N (2004) The utility of explicit
rate-based flow control in mobile ad hoc networks. In: IEEE
Wireless Communications and Networking Conference, vol 3,
pp 1921–1926
2. Deshmukh VS et al (2014) A survey of congestion control in
proactive source routing protocol in mobile ad hoc net-works.
Compusoft 3(12):1350
3. Soleymani SA, Abdullah AH, Anisi MH, Hassan WH, Goudarzi
S, Mandala S, Razak ZB, Mohamed N (2016) BRAIN-F: Bea-
con Rate Adaption based on Fuzzy Logic in Vehicular Ad Hoc
Network
4. Soleymani SA, Abdullah AH, Hassan WH, Anisi MH, Goudarzi S,
Baee MAR, Mandala S (2015) Trust Management in Vehicular Ad
Hoc Network: A Systematic Review. EURASIP J Wirel Commun
Netw 2015(1):1–22
5. Yaghmaee MH, Adjeroh DA (2009) Priority-based rate control
for service differentiation and congestion control in wireless
multimedia sensor networks. Comput Netw 53(11):1798–1811
6. Hussain A, Zia K, Farooq U, Ahmed S (2004) PLP–A Packet Loss
Prevention Technique for Partitioned MANETs using Location-
Based Multicast (LBM) Algorithms. In: National Conference on
Emerging Technologies
7. Senthil Kumaran T, Sankaranarayanan V (2011) Early conges-
tion detection and adaptive routing in manet. Egyptian Inf J
12(3):165–175
Mobile Netw Appl
8. Adem O, Kang SJ, Ko YB (2013) Packet loss avoidance in content
centric mobile adhoc networks. In: 15th International Confer-
ence on Advanced Communication Technology (ICACT), pp 245–
250
9. Puri S, Devene SR (2009) Congestion avoidance and load bal-
ancing in aodv-multipath using queue length . In: 2nd IEEE
International Conference on Emerging Trends in Engineering and
Technology (ICETET), pp 1138–1142
10. Karthikeyan N, Palanisamy V, Duraiswamy K (2009) Perfor-
mance comparison of broadcasting methods in MANETs. Interna-
tional Journal of Future Generation Communication and Networ-
king 2(2):47–58
11. Ee CT, Bajcsy R (2004) Congestion control and fairness for many-
to-one routing in sensor networks. In: Proceedings of the 2nd
ACM international conference on Embedded networked sensor
systems, pp 148–161
12. Camp J, Knightly E (2010) Modulation rate adaptation in urban
and vehicular environments: crosslayer implementation and exper-
imental evaluation. IEEE/ACM Trans Networking 18(6):1949–
1962
13. Chevillat P, Jelitto J, Barreto AN, Truong HL (2003) A dynamic
link adaptation algorithm for ieee 802.11 a wireless lans. In: IEEE
International Conference on Communications (ICC’03), vol 2,
pp 1141–1145
14. Xi Y, Kim BS, Wei JB, Huang QY (2006) Adaptive multirate auto
rate fallback protocol for ieee 802.11 wlans. In: IEEE Military
Communications Conference (MILCOM06), pp 1–7
15. Senthilkumaran T, Sankaranarayanan V (2013) Dynamic conges-
tion detection and control routing in ad hoc networks. Journal
of King Saud University Computer and Information Sciences
25(1):25–34
16. Zhai H, Chen X, Fang Y (2005) Rate-based transport control for
mobile ad hoc networks. In: IEEE Wireless Communications and
Networking Conference, vol 4, pp 2264–2269
17. Manikandan K, Durai MAS (2014) Active queue manage-
ment based congestion control protocol for wireless networks.
International Journal of Enterprise Network Management 6
(1):30–41
18. Tiwari SK, Rana YK, Jain A (2014) An ECN approach to conges-
tion control mechanisms in MANETs. Parameters 5(7)
19. Mishra TK, Tripathi S (2015) Explicit throughput and buffer
notification based congestion control: A cross layer approach.
In: Eighth International Conference on Contemporary Computing
(IC3), pp 493–497
20. Thakur S, Gupta M (2014) Mitigating congestion using data rate
control for MANETs
21. Rahman KC, Hasan SF (2010) Explicit rate-based congestion con-
trol for multimedia streaming over mobile ad hoc networks. Inter-
national Journal of Electrical & Computer Sciences IJECS-IJENS
10(04):28–40
22. Benslimane A, Rachedi A (2014) Rate adaptation scheme for
IEEE 802.11-based MANETs. J Netw Comput Appl 39:126–
139
23. Al-Saadi A, Setchi R, Hicks Y, Allen SM (2014) Multi-rate
medium access protocol based on reinforcement learning. In:
IEEE International Conference on Systems, Man and Cybernetics
(SMC), pp 2875–2880
24. Khan MS, Anisi MH, Waris S, Ali I, Khan MI (2016) Data
Rate Adaptation Strategy to Avoid Packet Loss in MANETs. In:
1st EAI International Conference on Future Intelligent Vehicular
Technologies
25. Khan MS, Midi D, Khan MI, Bertino E (2015) Adaptive trust
threshold strategy for misbehaving node detection and iso-
lation. In: IEEE Trustcom/BigDataSE/ISPA, vol 1, pp 718–
725
26. Khan MS, Midi D, Khan MI, Bertino E (2015) Adaptive trust
update frequency in MANETs. In: 21st IEEE Intl. Conf. on
Parallel and Distributed Systems (ICPADS), pp 132–139
27. Qin L, Kunz T (2006) Mobility metrics to enable adaptive routing
in MANETs. In: IEEE International Conference on Wireless and
Mobile Computing (WiMob’2006), Networking and Communica-
tions, pp 1–8
28. Samar P, Wicker SB (2004) On the behavior of communication
links of a node in a multi-hop mobile environment. In: Proceed-
ings of the 5th ACM international symposium on Mobile ad hoc
networking and computing, pp 145–156
A preview of this full-text is provided by Springer Nature.
Content available from Mobile Networks and Applications
This content is subject to copyright. Terms and conditions apply.
