Improving Performance of Ieee 802.11 by a Dynamic Control Backoff Algorithm Under Unsaturated Traffic Loads

International Journal of Wireless & Mobile Networks (IJWMN) Vol. 7, No. 6, December 2015
DOI : 10.5121/ijwmn.2015.7605 45
IMPROVING PERFORMANCE OF IEEE 802.11 BY A
DYNAMIC CONTROL BACKOFF ALGORITHM UNDER
UNSATURATED TRAFFIC LOADS
Hatm Alkadeki, Xingang Wang and Michael Odetayo
Department of Computing, Coventry University, Coventry, UK
ABSTRACT
The IEEE 802.11 backoff algorithm is very important for controlling system throughput over contention-
based wireless networks. For this reason, there are many studies on wireless network performance focus on
developing backoff algorithms. However, most existing models are based on saturated traffic loads, which
are not a real representation of actual network conditions. In this paper, a dynamic control backoff time
algorithm is proposed to enhance both delay and throughput performance of the IEEE 802.11 distributed
coordination function. This algorithm considers the distinction between high and low traffic loads in order
to deal with unsaturated traffic load conditions. In particular, the equilibrium point analysis model is used
to represent the algorithm under various traffic load conditions. Results of extensive simulation
experiments illustrate that the proposed algorithm yields better performance throughput and a better
average transmission packet delay than related algorithms.
KEYWORDS
IEEE 802.11, Backoff algorithm, Contention window threshold, Equilibrium point analysis, Dynamic
control backoff time algorithm
1. INTRODUCTION
Use of wireless local area networks (WLANs) is ubiquitous. IEEE 802.11 is the most important
standard in WLANs; however, achieving acceptable quality of service (QoS) over the standard is
still a challenging task. The IEEE 802.11 standard provides a basic medium access control
(MAC) mechanism called the distributed coordination function (DCF) [1]. The DCF is based on
the carrier sense multiple access with collision avoidance (CSMA/CA) mechanism with binary
exponential backoff (BEB) algorithm to reduce the probability of collisions. The BEB algorithm
is implemented by doubling the backoff time after every unsuccessful transmission. The backoff
time is called the contention window (CW), which is bounded by CWmax. However, the CW is
reset to zero after every successful transmission with the backoff counter in the interval (0, CWi
−1) [2]. Furthermore, the collision probability eventually leads to an unsuccessful transmission,
which decreases throughput. Therefore, improving the backoff algorithm will help to enhance
throughput performance and reduce the transmission delay. In this paper, we propose a new
backoff algorithm that applies the distinction between low and high traffic loads over non-
saturated traffic load conditions. We also evaluate the performance of the proposed algorithm in
terms of throughput and delay.
The remainder of this paper is organized as follows. Section 2 discusses and evaluates related
work, while Section 3 presents the proposed algorithm. Section 4 evaluates the performance of
the proposed algorithm and compares the results with those of existing works. Section 5 presents
our conclusion.

46
2. RELATED WORK
As mentioned above, the backoff algorithm for IEEE 802.11 is very important for controlling
channel access to maximize throughput and fairness [3]. There are several methods for extending
or proposing backoff algorithms. Most of these are based on modifying the backoff parameters
such as CW size and backoff stage (m), which is why much research has focused on modifying
the CW size during execution of the backoff algorithm to improve the performance of the IEEE
802.11 DCF. Therefore, an appropriate CW size leads to an improvement in the system
throughput by reducing the probability of collisions. However, some of the methods do not
account for dynamic traffic loads. For example, the authors in [4], proposed a new backoff
algorithm, called the multiplicative increase and linear decrease (MILD) algorithm. Their work
focused on modifying the CW size to CW×1.5 rather than doubling it after every unsuccessful
transmission. Moreover, CW size is decremented by 1 after every successful transmission rather
than reset to 0. However, decreasing the CW size gradually helps avoid any degradation in
performance. Therefore, the MILD algorithm is better than the BEB algorithm over large
networks. The authors in [5], extended the MILD algorithm by creating a new algorithm called
the linear increase linear decrease (LILD) algorithm. However, the authors applied CW+CWmin as
the size of increasing CW rather than multiplying by 1.5 to avoid the problem of slow linear
change over unsuccessful transmission. Therefore, the LILD algorithm provides good quality
performance over large networks. In another study [6], the authors proposed a new backoff
algorithm, called the exponential increase exponential decrease (EIED) algorithm. This algorithm
is based on increasing and decreasing the CW size exponentially. In [7], the authors proposed a
new algorithm called the double increment double decrement (DIDD) algorithm. This algorithm
is based on doubling the CW size after every unsuccessful transmission, in the same way as the
BEB algorithm, but using CW/2 as the size of decreasing CW after every successful transmission.
The DIDD algorithm generates a better result than the other algorithms mentioned above. In
addition, improving the BEB algorithm is still an active research topic. Therefore, [8] recently
evaluated the performance of BEB as a poor algorithm due to a number of collisions and CW
restoration after every successful transmission. This study is devoted to improve collision
avoidance under saturated traffic loads.
However, the above algorithms do not consider dynamic traffic loads. There are other interesting
directions that can be taken. For example, according to the research in [9], the authors focused on
channel traffic loads, and proposed a new algorithm called the exponential linear backoff
algorithm (ELBA). ELBA combines both exponential and linear algorithms depending on traffic
loads and provides better system throughput than the BEB, EIED, and LILD algorithms. In [10],
the authors used pause count backoff for monitoring channel traffic loads. This algorithm aims to
set an appropriate CW size based on estimation results. The authors in [11], proposed an adaptive
backoff algorithm based on the trade-off of efficiency and fairness for ad hoc networks. This
work is based on a fair schedule to control the increase and decrease in CW size depending on the
channel situation (idle or busy). In [12], the authors considered dynamic traffic loads by
proposing an algorithm based on monitoring the channel before data transmission. In this
algorithm, each station can record the number of busy slots by opening an observation window.
Thus, the sender can calculate a dynamic priority and CW size according to the number of
successful transmissions. In [13], the authors monitored the channel traffic loads by using a
channel state (CS) vector, and proposed a new algorithm called the dynamic deterministic
contention window control algorithm (DDCWC). This algorithm is based on monitoring the
channel traffic load conditions by checking the CS. However, selecting the optimum CW size
based on different traffic load conditions using the CS vector is difficult.
Overall, the majority of research work has paid great attention to improving the performance of a
saturated system without accounting for non-saturated traffic load conditions. Therefore, creating
a new backoff algorithm under non-saturated traffic load conditions is the objective of this paper.

47
3. BACKOFF STRATEGY
In this section, the proposed algorithm is discussed in detail. The discussion starts by describing
the principle behind the proposed algorithm in terms of mechanism and traffic load conditions.
3.1 Principle of the Proposed Algorithm
As mentioned in Section 2, most existing algorithms do not consider traffic loads under non-
saturated conditions, and thus do not take into account practical network operation. In this
section, a new backoff algorithm is proposed, called the dynamic control backoff time algorithm
(DCBTA). The DCBTA is implemented under non-saturated traffic loads using the equilibrium
point analysis (EPA) model [14]. The EPA model provides a very convenient way of evaluating
system performance under non-saturated traffic loads, thereby enabling the presentation of the
DCBTA under more flexible traffic sources. Furthermore, it is possible to investigate network
traffic load conditions under a different number of stations.
In the DCBTA, channel conditions are checked by a CW threshold (CWThreshold). The CWThreshold
value serves as a reference point for the collision rate. Therefore, CWThreshold plays a major role in
the proposed algorithm as illustrated in Figure 1. The CWThreshold size is dependent on the
maximum contention window size (CWmax), where the value of CWThreshold is equal to half that of
CWmax. For example, the value of CWmax in [14] was selected to be 1024. In this case, the value of
CWThreshold is set to 512. Each state of node “i” is initially set to the minimum CW size (CWmin),
which can be increased up to CWmax.
Figure 1. Underlying mechanism of the proposed algorithm (DCBTA)
Figure 1 shows that the proposed algorithm enables the detection of heavy or light traffic load
using the CWThreshold value. After every unsuccessful transmission, if the CW size is smaller than
the CWThreshold value, that is, a light traffic load, the CW size is doubled as (2×CW) similar to the
BEB algorithm. Conversely, if the CW size is greater than CWThreshold, that is, a heavy traffic load,
the CW size is doubled and incremented by two as (2×CW+2). Adding two to double the CW size
leads to a decrease in the collision probability, thus increasing system throughput. A summary of
this discussion is given below:

48
Light traffic load:
If (CWi ≤ CWThreshold)
Successful transmission: CWi = CWi−1−1;
Else (CWi = CWi−1×2).
Heavy traffic load:
If (CWi > CWThreshold)
Successful transmission: CWi = CWi−1−2;
Else (CWi = CWi−1×2+2).
3.2 DCBTA Algorithm under EPA Model
In order to run the proposed algorithm under non-saturated traffic load conditions, the EPA model
is used. The EPA model provides a very convenient way to evaluate the system performance
under non-saturated traffic load conditions. In the EPA model, the traffic load generated by each
station follows the Poisson distribution with rate time/packets. Thus, the packet transmission
probability (R) plays a pivotal role in the EPA model mechanism. However, the proposed
algorithm adaptively changes the CW size with respect to the collision rate or the transmitting
nodes. Therefore, the proposed algorithm under the EPA model affects the transmission
probability of node “Ri” at any state of node “i” as follows:
In networks with a large number of nodes or a high collision rate, the proposed algorithm results
in a very low probability of transmission. In this case, the CW size increases to more than the
threshold, resulting in a high traffic load. The throughput formula is the same, where Ri is
calculated as follows:
Otherwise, the value of CWi decreases to less than or equal to the threshold value, resulting in a
low traffic load. Then Ri is calculated in the same way as the BEB algorithm under the EPA
model. In the case of successful transmission, the CWi size decreases gradually to avoid
performance degradation. However, if the CWi size is less than or equal to CWThreshold, the CW
size for the next stage CWi+1 is decremented by one:
If CWi is greater than CWThreshold, CWi+1 is decremented by two:
4. PERFORMANCE EVALUATION
In this section, the proposed backoff algorithm is compared with related algorithms in terms of
throughput and average packet transmission delay. The comparative evaluation of backoff
algorithms is carried out using MATLAB simulation experiments.

49
4.1 Simulation Settings
The proposed and related algorithms are implemented based on the EPA model assumption in
[14]. Therefore, there are no hidden terminals and system performance can be investigated under
more flexible traffic sources with fixed packet length. The different system parameters used in the
simulation experiments are summarised in Table 1.
Table 1. System parameter settings
4.2 Comparison of Throughput
System performance of the proposed algorithm (DCBTA) is compared with that of the BEB
algorithm under non-saturated traffic load conditions in the work of [14]. In addition, the
performance of DCBTA is compared with other related algorithms, such as ELBA in the work of
[9]. ELBA combines both exponential and linear algorithms, which is why it was selected for
comparison with the proposed algorithm. The number of nodes is set to 50; the maximum number
of backoff stages equals six. Figure 2 illustrates the throughput performance for DCBTA
compared with the BEB algorithm and ELBA under various traffic load conditions. The results
show that the throughput performance of DCBTA is better than that of the BEB algorithm and
ELBA under various traffic loads.
Figure 2. Non-saturated throughput comparison of the proposed algorithm and various related backoff
algorithms, where (CWmin = 8, m = 6)
Parameter Value
Packet Payload 8184 bits
Data Packet 8200 µs
Channel Bit Rate 1 Mbit/s
Physical Slot Time 50 µs
DIFS 128 µs
SIFS 28 µs
ACK_Timeout 300 µs
RTS 350 µs
CTS 350 µs
CWmin 8
CWmax 1024
Maximum Backoff Stage, m 6
Network Nodes (n) 50-100 nodes
Analytical Tool EPA model

50
To investigate the impact of using different CWmin size, Figure 3 plots the throughput
performance for DCBTA, BEB, and ELBA with a varying size of 8, 16, and 32 CWmin. The
throughput increases when CWmin increases, since increasing CWmin contributes to collision
avoidance. Moreover, system throughput depends on the incoming data [15]. Therefore, the
throughput result is equal to the increase in the incoming traffic data rates if the traffic load is
low. Otherwise, throughput becomes saturated if the amount of data is sufficiently high. Hence,
the system performance strongly depends on system parameters, such as CWmin and m.
Figure 3 clearly shows that DCBTA provides better throughput results than BEB and ELBA with
different CWmin size under various offered loads. The DCBTA allows the stations to adjust CW
value appropriately according to the traffic load variation within the network. This means that the
DCBTA mechanism can reduce the number of collisions, which will lead to increased system
throughput. In addition, the performance results show that DCBTA has lower performance
degradation than BEB and ELBA. The reason for this is that the CW size decreases gradually
after every successful transmission.
Figure 3. Non-saturated throughput comparison of the proposed algorithm and various related backoff
algorithms with varying CWmin (8, 16, 32) and m = 6
4.3 Comparison of Delay
In [14], the EPA model represented the MAC channel in idle, transmission, and collision states
under varying traffic load conditions. The MAC channel was proposed as a multi-dimensional
discrete-time Markov chain analysis model. Therefore, the delay can be represented as a sequence
of discrete time delays as follows:
Average transmission delay = Total delay / Total number of transmissions,
where:
Total delay = Total transmission time + Total time delay in the collision + Backoff time + Empty
slot.
Total transmission time = Transmission time of single packet × Total number of transmissions.
Transmission time of single packet = RTS + SIFS + CTS + SIFS + Data + SIFS + ACK + DIFS.
Total time delay in the collision = Delay time of single collision × Total number of collisions.
Delay of single collision = RTS + DIFS.

51
Average packet transmission delays for the BEB algorithm, ELBA, and DCBTA are calculated
over 100 stationary nodes. For further investigation, the performance of algorithms is also
examined under different CWmin values of 32, 64, and 128. All the assumptions and system
parameters related to this experiment are the same as in the previous section. Figure 4, Figure 5,
and Figure 6 show the delay comparison of the BEB, ELBA, and DCBTA algorithms under the
EPA unsaturated model. The increment in CW size in the BEB and ELBA algorithms results in
greater delay compared to that of the DCBTA algorithm. This means that the DCBTA mechanism
produces a small delay by reducing a collision rate. Actually, when there is a high offered traffic
load, the CW size should be kept large to avoid frequent collision. Moreover, DCBTA reduces
CW size more slowly after successful transmission in order to avoid the collision probability. For
these reasons, it can clearly be seen that the proposed algorithm has a smaller average
transmission delay than that of the BEB and ELBA algorithms, as shown in the figures below.
Figure 4. Average packet transmission delay with CWmin = 32, m = 6

52
5. CONCLUSION
In this paper, a new backoff algorithm under non-saturated traffic loads was proposed to represent
actual network situations. A suitable model was selected to evaluate system performance under
non-saturated traffic loads such as the EPA model.
The motivation for this research was to enhance the system performance of the IEEE 802.11 DCF
under non-saturated traffic load conditions in terms of throughput and time delay. To realize this,
a new backoff algorithm was proposed and then integrated with the EPA model.
The performance results show that the proposed algorithm (DCBTA) presents better system
throughput than the BEB algorithm and ELBA. In addition, calculation of the average packet
transmission delay for each algorithm shows that the DCBTA provides a better time delay than
the BEB algorithm and ELBA. This is because the DCBTA decreases the time delay, which leads
to an increase in system throughput. However, throughput and delay are both relevant for the
performance metrics of QoS. Therefore, the proposed algorithm may help to enhance the
effectiveness of the IEEE 802.11 DCF. A possible further extension of the DCBTA would be to
consider the various data frame sizes.

53
REFERENCES
[1] Bianchi, G. (2000) “Performance analysis of the IEEE 802.11 distributed coordination function”,
IEEE Journal on Selected Areas in Communications, 18(3), 535-547.
[2] Gangrade, K., Patidar, P. & Tiwari, A. (2013). “Performance Evaluation of IEEE 802.11 MAC DCF
Using Various Schemes Towards : Throughput, Delay and Reliability”, International Journal of
Advanced Research in Computer and Communication Engineering 2 (7), 2731–2737.
[3] Cho, J.W. & Jiang, Y. (2015). “Fundamentals of the Backoff Process in 802.11 : Dichotomy of the
Aggregation”, IEEE Transactions on Information Theory 61 (4), 1687–1701.
[4] Bharghavan, V., Demers, A., Shenker, S. & Zhang, L. (1994). “MACAW: a media access protocol for
wireless LAN’s”, SIGCOMM ’94 Proceedings of the Conference on Communications Architectures,
Protocols and Applications, 24(4), 212-225.
[5] Deng, J., Varshney, PK. & Haas, ZJ. (2004). “A new backoff algorithm for the IEEE 802.11
distributed coordination function”, Proceedings of CNDS, 4, 18-21.
[6] Song, N., Kwak, B., Song, J. & Miller, ME. (2003). “Enhancement of IEEE 802.11 distributed
coordination function with exponential increase exponential decrease backoff algorithm”, The 57th
IEEE Semiannual Spring VTC, 4, 2775-2778.
[7] Chatzimisios, P., Boucouvalas, AC., Vitsas, V., Vafiadis, A., Economidis, A. & Huang, P. (2005). “A
simple and effective backoff algorithm for the IEEE 802.11 MAC protocol”, Proceedings of the
International Conference on Cybernetics and Information Technologies, System and 2nd
Applications, 1, 48-53.
[8] Cheng, H., Yan, X., Lian, H., Weng, L., Zhang, Q. & Feng, Z. (2014). “A novel collision avoidance
algorithm for IEEE 802.11 wireless LANs”, IEEE Military Communications Conference (MILCOM),
879–884.
[9] Lin, CH., Shieh, CK., Hwang, WS. & Ke, CH. (2008). “An exponential-linear backoff algorithm for
contention-based wireless networks”, Proceedings of the International Conference on Mobile
Technology, Applications and Systems, 42.
[10] Liang, HM., Zeadally, S., Chilamkurti, N. & Shieh, CK. (2008). “A novel pause count backoff
algorithm for channel access in IEEE 802.11 based wireless LANs”, International Sympsium on
Computer Science and Its Applications, 8, 163-168.
[11] Cui, H.X. & Wei, G. (2009). “A novel backoff algorithm based on the tradeoff of efficiency and
Fairness for ad hoc networks”, International Conference on Communications and Mobile Computing,
02, 82-86.
[12] Jingqi, F., Qiang, Z. & Haikuan, W. (2009). “A new backoff algorithm based on the dynamic
modulating parameters of IEEE 802.11”, WICOM ‘9 Proceedings of the 5th International Conference
on Wireless Communications, Networking and Mobile Computing, 1, 4-7.
[13] Balador, A., Movaghar, A., Jabbehdari, S. & Kanellopoulos, D. (2012). “A novel contention window
control scheme for IEEE 802.11 WLANs”, IETE Technical Review, 29, 202-212.
[14] Wang, X., Min, G. & Guan, L. (2009). “Performance modelling of IEEE 802.11 DCF using
equilibrium point analysis”, International Journal of Wireless and Mobile Computing, 3(3), 201-209.
[15] Alam, M., Mumtaz, S., Saghezchi, F.B., Radwan, A. & Rodriguez, J. (2013). “Energy and throughput
analysis of reservation protocols of Wi media MAC”, Journal of Green Engineering, 3, 363–382.

Improving Performance of Ieee 802.11 by a Dynamic Control Backoff Algorithm Under Unsaturated Traffic Loads

Related slideshows

More Related Content

Improving Performance of Ieee 802.11 by a Dynamic Control Backoff Algorithm Under Unsaturated Traffic Loads