Efficient and Fair Bandwidth Allocation AQM Scheme for Wireless Networks

Rafe Alasem
International Journal of Computer Networks (IJCN), Volume (2): Issue (2) 132
Efficient and Fair Bandwidth Allocation AQM Scheme for
Wireless Networks
Rafe Alasem rafe.alasem@gmail.com
Computer Science College
Computer Science Department
Imam Muhammed Ibn Saud Islamic University
PO Box 5701, Riyadh 11432, KSA
Abstract
Heterogeneous Wireless Networks are considered nowadays as one of the
potential areas in research and development. The traffic management’s schemes
that have been used at the fusion points between the different wireless networks
are classical and conventional. This paper is focused on developing a novel
scheme to overcome the problem of traffic congestion in the fusion point router
interconnected the heterogeneous wireless networks. The paper proposed an
EF-AQM algorithm which provides an efficient and fair allocation of bandwidth
among different established flows.
Finally, the proposed scheme developed, tested and validated through a set of
experiments to demonstrate the relative merits and capabilities of a proposed
scheme
Keywords: Wireless Network, Congestion Control, Active Queue Management, Random Early Detection.
1. INTRODUCTION
Wireless communication technology is playing an increasingly important role in data networks.
Wireless networks are usually connected to the internet via backbone gateway routers. The
packet loss may occur at fusion points that connect the backbone network to the wireless
networks.
However, in a wireless heterogeneous network, the loss occurs due to the channel nature and
characteristics. For example, in IEEE 802.11 wireless networks, congestion may be defined as a
state where the shared wireless medium is completely occupied by the nodes because of given
channel characteristics in addition to external interference. The shared environment of the
wireless medium causes a node to share the transmission channel not just with other nodes in
the network, but also with external interference resources [1].
In reality that a wireless channel is shared by challenging neighbor nodes and the number of
nodes sharing this channel may change all the time [2]. An additional reason is that the wireless
link bandwidth is affected by many changing physical conditions, such as signal strength,
propagation distance, and transmitter power. For example, an IEEE 802.11 node can modify its
MAC-layer data rate dynamically according to different situations, which means the output
bandwidth of this node and other neighbor nodes may also change .

Rafe Alasem
In wireless networks throughput degradation can occur due to the sharing the lossy channel and
packets collision. Slotted CSMA/CD is used to overcome a bit the collision occurrence at the
lossy channel. Sending the frames at the slot start reduces the chance of frames' collision.
In addition, the using of a conventional mechanism in managing the traffic in inside the wireless
gateway node causes an additional loss to truly arrived frames. Throughput degradation can
occur also due to improper use of traffic management schemes at the fusion points of
heterogeneous wireless networks [6]. For these reasons, an efficient mechanism for congestion
control should be applied at the bottleneck nodes to overcome the additional loss in an accurate
data.
In an IEEE 802.11 wireless network, the occurrence of high density nodes in a single collision
domain can cause congestion, in consequence causing a substantial bottleneck gateway router
[5]. Packet dropping, packet delays and session disruptions are a consequence of congested
network.
Congestion control problem occurs when the demand on the network resources is greater than
the available resources and due to increasing mismatch in link speeds caused by intermixing of
heterogeneous network technologies. This congestion problem cannot be solved with a large
buffer space. Clearly too much traffic will lead to a buffer overflow, high packet loss and large
queuing delay. Furthermore, congestion problem cannot be solved by high-speed links or with
high-speed processor, because the high-speed link connected via the high-speed switch with the
low-speed links will cause congestion at the wireless fusion point of interconnection.
Drop Tail has been proposed in [4]. The most operational routers currently use Drop Tail coupled
with FIFO (First in first out) scheduling scheme. In Drop Tail, all packets are accepted until the
maximum length of the queue is reached and then dropping subsequent incoming packets until
space becomes available in the queue.
Drop Tail is not appropriate as a feedback control system for high-speed networks because it
sustains full queues and this may increase the average queuing delay in the network. More
importantly, Drop Tail can cause a lockout due to traffic phase effects and the global
synchronization, and thus results in low throughput. The lost packet from a Drop Tail queue will
usually be retransmitted by TCP protocol via its retransmission timer. No congestion is detected
until the buffer becomes full and the maximum congestion indicator is generated because all
arriving packets are dropped. Then each source detects lost packets it will slow down the arrival
rate of the sending packets until the queue will be less than the capacity of the link. No
congestion indicator will be generated when the queue is not full, each source will increase until
overflow happens again
In the recent years, Active queue management (AQM) mechanisms have been proposed to
provide an efficient queue management by selectively dropping/marking packets when
congestion is anticipated so that TCP senders can reduce their transmission rate before an
overflow occurs. AQM mechanisms are employed in the Internet by the routers to provide better
stability, fairness, and responsiveness to dynamic variations in computer networks. Using queue
management mechanisms in an efficient way will avoid the congestion collapse and lead to high
link utilization.
In this paper, we present a novel buffer management approach for congestion control in a
wireless network. This approach achieves both efficient and fair allocation of bandwidth among
flows by randomly dropping frames and increases data throughput to the next hop.
The rest of this paper is organized as follows: In section II, the network model is presented. The
proposed scheme is developed in section III. Extensive simulations and results are investigated in
section IV. Section V concludes this paper..

Rafe Alasem
2. NETWORK MODEL
The network model considered in this paper will be explained in this section. IEEE 802.11 based
wireless LAN networks have been chosen as a reference model of this investigation. A
bottleneck wireless gateway router has been taking into account in this topology( Figure 1). A
wireless channel L is shared by N challenging neighbor nodes and the number of nodes sharing
this channel may change all the time. We suppose that all nodes use the same power and
modulation methods. Figure 1 illustrates the wireless network model. The traffic model is shaped
as follows. We assume that each node is transmitting HTTP packets of size S bits, and they are
generated according to Markovian-Modulated Poisson Process MMPP with arrival rate .
FIGURE 1: IEEE 802.11 Wireless network
The fusion point that connects the wireless network with wired network has a finite amount of
buffer space, and this buffer is managed via an adaptive queue management scheme. In
addition, the output link has a fixed bandwidth and connects the wireless network with Internet.
But for a wireless network, for example in 802.11b, the node can dynamically change its MAC-
layer data rate to 1, 2, 5.5, or 11Mbps. Consequently, when congestion is occurred the TCP is
unable to maintain fairness and stability with improper estimation of the link capacity parameter.
For this reason, the need for an adaptive and intelligent AQM algorithm to be implemented at the
fusion point of a wireless network is critical and crucial.
The congestion sliding widow w is increased by one every round trip time if no congestion is
detected, and it is reduced by half if a congestion is detected. This is called an additive-increase
multiplicative decrease (AIMD) mechanism that represents the behavior of TCP flows, and it is
described by the following nonlinear differential equation [9]:
1)p(t
1)2r(t
1)w(t
r(t)
1
w 




(1)

Rafe Alasem
Where R corresponds to the round-trip time (seconds) and p is representing the probability of
packet drop/mark. In addition, the instantaneous queue can be expressed in the following
equation:
CN
tr
tw
q 
)(
)(

(2)
Where C is the link speed (packets/sec) and N load factor (number of TCP connections).
3. EF-AQM SCHEME
EF-AQM scheme has been designed and analyzed in terms of feedback control theory (Figure 2).
The advantages of using control theory are to increase the speed of response and to bring further
improvement to the system robustness and stability. These advantages can be achieved by
regulating the output queue length around a target value Qref. An important goal of the AQM
design is to stabilize the queue length
)(tq at a given target refQ
, so that the magnitude of the
error signal.
)()( tqrefQte 
(3)
is kept as small as possible.
The output of the EF-AQM controller represents the dropping probability and is simplified as:
)1())1()(()1()( 10  kTeKkekeKkyky (4)
where T is the sampling period time, K1 and K2 represent the tuning parameters of the
controller.
FIGURE 2: EF-AQM as a feedback control system

Rafe Alasem
The dropping probability is calculated according to the intelligent controller, and it is considered
as a function of the difference between the current value of the queue length and the reference
queue length. Our aim is to compute this dropping probability
)(tpd such that it will keep the
instantaneous queue length close to the target queue. Therefore, the dropping probability
)(kpd
can be achieved as:










1)(1
1)(0)(
0)(0
)(
ky
kyky
ky
k
d
p
(5)
The dropping probability is calculated at every packet arrival. The reason for randomizing the
packet drops is the hypothesis that users generating more traffic would have a greater number of
packets dropped. The same as in Random Early Detection (RED) [7], if the average queue length
exceeds a minimum threshold Lmin, incoming packets are dropped/marked with a probability that it
is a linearly increasing function of the average queue length. When the average queue size
exceeds a maximum threshold Lmax, the router is likely to incur congestion, and all incoming
packets are dropped/marked. When it is between, a packet is dropped with a probability p which
represents the output of EF-AQM controller.
4. PERFORMANCE EVALUATION
To validate the performance and the robustness of the proposed EF-AQM algorithm for wireless
network, we simulated it using OMNET++ platform [10] with highly bursty traffic. Different
scenarios have been chosen to validate the proposed algorithm with different number of flows.
The parameters used in EF-AQM simulation are: N = 10,20,30 and 40 flows, Lmax = 500 packets,
Lmin = 100 packets, packet length = 100 bytes, IEEE 802.11 propagation delay = 10 ms.
Probability based dropping pmax =0.2 and the target queue length is (Lmax + Lmin)/2.
4.1. Throughput
As shown in Figure 3, EF-AQM has offered higher throughput as compared to the classical
algorithm Drop Tail and RED for the 10, 20, 30 and 40 flows respectively. It is observed that
although the number of TCP flows has increased, EF-AQM has offered higher throughput for
different control approaches and reach (100%) for the smaller flows. This due to the stability of
EF-AQM in maintaining the queue length which makes it more stable around the target queue
and shrunk in width; thereby packet dropping is less despite of increasing the number of TCP
flows.

Rafe Alasem
0
20
40
60
80
100
120
10 20 30 40
No. of flows
Normalizedthroughput(%)
DropTail
AQM-IS
RED
FIGURE 3: Throughput comparison for different flows
4.2. Queuing Delay
The queuing delay is considered as one of the important metrics in performance evaluation of any
AQM controller. Figure 4 shows that the queuing delay of the proposed EF-AQM is very small
compared other approaches for a variety of flows. It is noted that the queuing delay becomes
constant for EF-AQM controller despite increasing traffic load or changing the type of data traffic.
For other schemes, the delay is continued to increase when changes occur in any of the network
parameters.
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
10 20 30 40
No. of flows
Delay(ms)
DropTail
AQM-IS
RED
FIGURE 4: Delay comparison for different flows
4.3 Queue length
Figures 5 Shows the instantaneous queue length evolution comparison under the EF-AQM
approach compared with RED for the number of TCP flows equal to 20, 30 and 40 respectively. It
can be seen that the instantaneous queue length of the EF-AQM is stable and oscillate around
the target queue length. While the instantaneous queue length of the RED algorithm is still
fluctuated away from the target queue length as the number of TCP connection increased due to
the sensitivity of RED to any change in its parameters. It is worth noting that the EF-AQM scheme
with adaptive tuning parameters is the most stable control scheme as compared to the others;
demonstrates steadiness mode despite higher number of TCP flows.

Rafe Alasem
FIGURE 5: Instantaneous queue comparison for different flows
5. PERFORMANCE EVALUATION
This section demonstrates merits and capabilities of the proposed scheme. Some of the critical
issues are further discussed below:
Convergence: It is noted that the speed of convergence of the proposed congestion control
mechanisms to a stable operating point is independent of the number of TCP sources and
connections. The output queue length converges rapidly to the reference queue as the sources
start transmission their data. It is also observed that the EF-AQM scheme has the optimum
convergence time rather than RED.
Fairness: One of the important goals of proposed congestion control mechanisms is the
contiguity between the link utilization and the queuing delay in the bottleneck router. Fair dropping
of packets and fair sharing of bandwidth for all connections is achieved. Also it is worth noting
that EF-AQM offers an optimum round trip time fairness metric in comparison to RED scheme
due to its minimum queuing delay.
Robustness and Stability: One goal in this investigation is to explore the robustness and
stability, in terms of minimizing oscillations of queuing delay or of throughput. In practice, stability
is frequently associated with rate fluctuations or variance. It is observed that EF-AQM scheme
has offered an optimum queuing length and queuing delay oscillations. It is also noted that RED
has offered higher queuing delay and queue length oscillations.
Scalability: It is worth mentioning that global implementation of the proposed EF-AQM
congestion control mechanism in a decentralized form is expected to offer high network
utilization. Scaling the local stability and low queuing delay for individual gateways in a large
network will sufficiently present a global implementation of a congestion control mechanism over
a scalable network.
Efficiency: One of the key concerns in the design of congestion control mechanisms has been
the CPU usage time and maximizing bandwidth utilization. It is observed that the proposed
mechanism provide high throughput in comparison to the standard RED. It is worth noting that
EF-AQM has a moderate CPU usage compared to RED algorithm.
6. CONCLUSION
This paper presented an efficient and fair bandwidth allocation AQM algorithm to overcome the
problem of congestion control in heterogeneous wireless network. It has been demonstrated that
the new EF-AQM has achieved desirable properties such as robustness and fast system
response, as compared to the traditional DropTail and RED. Finally, a set of experiments has
been provided to demonstrate the efficiency of the proposed design approach. It is noted that the

Rafe Alasem
proposed EF-AQM design approach performs significantly better than many well-known
schemes, and guarantees the robustness of the controller.
7. REFERENCES
[1] Prashanth A., Ashish S., Elizabeth B., Kevin A. and Konstantina P. “ Congestion-Aware Rate
Adaptation in Wireless Networks: A Measurement-Driven Approach”, IEEE SECON, San Francisco, CA,
June 2008.
[2] Jian Pu; Hamdi, M., "Enhancements on Router-Assisted Congestion Control for Wireless Networks,"
Wireless Communications, IEEE Transactions on , vol.7, no.6, pp.2253-2260, June 2008.
[3] Acharya, P.A.K.; Sharma, A.; Belding, E.M.; Almeroth, K.C.; Papagiannaki, K., "Congestion-Aware Rate
Adaptation in Wireless Networks: A Measurement-Driven Approach," Sensor, Mesh and Ad Hoc
Communications and Networks, 2008. SECON '08. 5th Annual IEEE Communications Society
Conference on , vol., no., pp.1-9, 16-20 June 2008.
[4] Jacobson V., (1988), “Congestion avoidance and control”, ACM SIGCOMM Computer Communication
Review, Vol. 18(4), p.314-329.
[5] Dong, Y.; Makrakis, D.; Sullivan, T., "Network congestion control in ad hoc IEEE 802.11 wireless LAN,"
Electrical and Computer Engineering, 2003. IEEE CCECE 2003. Canadian Conference on , vol.3, no.,
pp. 1667-1670 vol.3, 4-7 May 2003.
[6] Xiao Laisheng; Peng Xiaohong; Wang Zhengxia; Xu Bing; Hong Pengzhi, "Research on Traffic
Monitoring Network and Its Traffic Flow Forecast and Congestion Control Model Based on Wireless
Sensor Networks," Measuring Technology and Mechatronics Automation, 2009. ICMTMA '09.
International Conference on , vol.1, no., pp.142-147, 11-12 April 2009.
[7] Floyd, S, Jacobson, V., (1993), “Random early detection gateways for congestion avoidance”. IEEE/ACM
Trans. on Networking, Vol. 1(4), pp.397–413.
[8] Alasem, R, Hossain, M. A. and Awan, I., (2007), "Intelligent Active Queue Management Predictive
Controller using Neural Networks", IEEE Computer Society (The 2007 International Conference on Next
Generation Mobile Applications, Services and Technologies), ISBN 0769528783, Cardiff, UK, pp. 205-
210.
[9] C. Hollot, V. Misra, D. Towsley ,W. Gong. “A Control Theoretic Analysis of RED”. Proc. INFOCOM Vol. 3,
April 2001,pp. 1510 – 1519.
[10] Alasem, R., Hossain, M. A., Awan, I., (2007), "Active Queue Management Controller using Smith
Predictor for Time Delay Networks," IEEE International Conference on Networking, Sensing and
Control, London, UK , Vol.1, pp.568-573.
[11] Varga, A. (2010), Omnet++: user manual, [Online : http://www.omnetpp.org ]

Efficient and Fair Bandwidth Allocation AQM Scheme for Wireless Networks

More Related Content

What's hot (19)

Viewers also liked (6)

Similar to Efficient and Fair Bandwidth Allocation AQM Scheme for Wireless Networks (20)

Recently uploaded (20)

Efficient and Fair Bandwidth Allocation AQM Scheme for Wireless Networks