ESTIMATION OF MEDIUM ACCESS CONTROL LAYER PACKET DELAY DISTRIBUTION FOR IEEE 802.11

International Journal of Wireless & Mobile Networks (IJWMN) Vol. 5, No. 6, December 2013
DOI : 10.5121/ijwmn.2013.5608 101
ESTIMATION OF MEDIUM ACCESS CONTROL
LAYER PACKET DELAY DISTRIBUTION FOR IEEE
802.11
Hatm Alkadeki, Xingang Wang and Michael Odetayo
Department of Computing, Coventry University, Coventry City, UK
ABSTRACT
The most important standard in wireless local area networks is IEEE 802.11. This is why much of the
research work for the enhancement of wireless network is usually based on the behavior of IEEE 802.11
protocol. However, some of the ways in which IEEE 802.11 medium access control layer behaves is still
unreliable to guarantee quality of service. For instance, medium access control layer packet delay, jitter
and packet loss rate still remain a challenge. The main objective of this research is to propose an
accurate estimation of the medium access control layer packet delay distribution for IEEE 802.11. This
estimation considers the differences between busy probability and collision probability. These differences
are employed to achieve a more accurate estimation. Finally, the proposed model and simulation are
implemented and validated - using MATLAB program for the purpose of simulation, and Maple program
to undertake the calculation of the equations.
KEYWORDS
IEEE 802.11 DCF Standard, MAC Layer Packet Delay Distribution, Busy probability & Collision
Probability
1. INTRODUCTION
The IEEE 802.11 WLANs standard often called Wireless Fidelity (Wi-Fi) networks provide
flexible wireless access [1]. The standard is built on two specifications layers known as the
Physical (PHY) layer and the Medium Access Control (MAC) layer. The MAC layer is a very
important part for supporting quality of service (QoS), in particular to support the multimedia
applications. Therefore, many researchers pay a great attention to the MAC layer so as to try
and solve many of the problems which restrict QoS. These problems include: delay, packet loss
rate, and jitter, which are still opening research questions. Due to the problems with the MAC
layer the guaranteeing of QoS in IEEE 802.11 is still a very challenging task [2]. This is why
QoS to IEEE 802.11 has become an active research area [2].
Estimating the MAC layer packet delay distribution will provide necessary information for QoS
enhancement [3]. However, the primary mechanism for the MAC layer is based on Distributed
Coordination Function (DCF) [4]. The DCF is based on Listen Before Talk (LBT) mechanism
to detect whether a channel is idle or busy so as to avoid a collision. Moreover, the DCF uses
four-way hand shaking (RTS/CTS-DATA/ACK). This mechanism helps the protocol to reduce
number of collisions. These events will cause delay during a transmission process, which are
considered in this paper. This paper aims to improve the work of [3] by considering the
differences between the busy probability and the collision probability. Then, the accuracy of our
new mathematical model is evaluated by carrying out simulation using the MATLAB program.

102
The rest of the paper is organized as follows. Section 2 summarizes and evaluates number of
popular related work. Section 3 presents the MAC layer delay as a terminating renewal process.
Section 4 presents the numerical calculations for our new mathematical model. Section 5
compares the results obtained from our model with the simulation results for wireless node
behavior model based on the Bianchi's model. Moreover, the results will also be compared with
the previous work. Section 6 concludes our study.
2. RELATED WORK
Most of the popular work for studying the behavior of a single node and performance for
wireless network is based on the Markov chain probability model. The back-off mechanism for
IEEE 802.11 DCF is represented by a bi-dimensional discrete time Markov chain [4].
Therefore, Bianchi in [4] proposed a good evaluation for IEEE 802.11 DCF performance but it
has some limitations which need to be investigated. This is why a number of researchers are
working on extending that model so as to improve the performance of wireless networks
(WLANs). For instance, in [5], the authors improved [4] by using the same assumption to
consider the effect of packet retransmission limit during their analysis but using a new scheme
called DCF+ to improve the performance of TCP. This means that the authors worked on the
MAC layer to improve the performance analysis and the transport layer to support the
transmission of packets over WLANs.
On the other hand, many other researchers focused on the MAC layer delay rather than the
transport layer. In [6], the authors developed the performance analysis model in [5] to calculate
the packet delay. The authors proved that the result from their model is better than the result
from the model without considering packet retry limits. However, the authors in [7], improved
[4] from the bi-dimensional Markov chain to a single dimensional to compute throughput. On
the other hand, they calculated the average packet delay by reducing the model in [5] from two
to one dimension. However, one-dimensional Markov chain is a good idea for a simple
calculation but it is not suitable for large network. Therefore, in [8], the authors proposed a new
analytical model to calculate the average packet delay and improved the model in [6].
Furthermore, in [9], the authors evaluated that [7] is an inexact model for a small network which
has less than 20 nodes. The authors proposed a new delay model by extending their previous
work in [8]. Their study proposed most delay events as packet delay average, distribution, and
jitter of IEEE 802.11 DCF. In [10], the authors estimated a delay, jitter, and throughput by using
a multiplayer in a wireless network game model. On the other hand, in [11], the authors improve
the work in [6] to study delay distribution of DCF. However, the study assumed the channel is
idle, which means collision probability and busy probability are not included. Thus in [3], the
authors proposed estimation for the MAC layer delay by using a mathematical model based on
successful transmission. On the other hand, the model does not propose busy probability. Some
of the researchers proposed something different to increase throughput and decrease mean
packet delay for IEEE 802.11 DCF protocol. Therefore, in [12], the authors proposed a new
back-off algorithm (EBA) to improve throughput by decreasing the delay packet.
As we have seen most existing models do not consider the differences between the busy
probability and the collision probability. Therefore, in [13], the authors argued that the
assumption that the busy probability and the collision probability are similar is not valid.
Therefore, the delay occurred from frozen period in the busy state must be considered in
mathematical models. Furthermore, most of all the above research work considered the
estimations of the average MAC layer packet delay but the packet delay distribution is still
unsolved [3].

103
3. ESTIMATION OF MAC LAYER PACKET DELAY DISTRIBUTION
In this paper, we consider all possible events during the back-off transmission mechanism for
IEEE 802.11 DCF as shown in figure1. Therefore, the total time is regarded as a sequence of
intervals of Empty (DEmp), Successful (DSuc), Busy (DBus) and Collision (DCol) delay time.
Figure1. Time Events during the back-off mechanism in IEEE 802.11 DCF represented by bi-
dimensional Markov chain
Therefore, the time delays are calculated using the following equations:
DEmp = 50 µs (1)
DSuc = RTS + SIFS + CTS + SIFS + H + E P + SIFS + ACK + DIFS (2)
DCol = RTS + DIFS (3)
DBus = DIFS + SIFS + ACK (4)
The network in this paper has 20-30 nodes; each node is equally likely to transmit. This uniform
distribution considered because the time was very short. Therefore, this proposed model is
based on two probabilities to represent the behavior model for each state. We supply τtr as the
probability of the sender station attempting a transmission, τnb probability of one neighbor
station attempting a transmission. In this case, we have the following possible different
probability events, which we are able to calculate from (5), (6), (7), (8) and (9):
PEmp → No transmission attempts idle
PSuc → One of neighbor ′s attempting to transmit
POwn → The sender attempting to transmit
PCol → Transmissionattempts at the same time
PBus → Channel busy by transmission or collision

104
PEmp = 1 − τtr . 1 − τnb
n−1 (5)
PSuc = n − 1 . τnb . 1 − τtr . 1 − τnb
n−2 (6)
POwn = τtr . 1 − τnb
n−1 (7)
PCol = τtr . n − 1 . τnb . ( 1 − τnb )n−2 (8)
PBus = 1 − PEmp − POwn − PSuc − PCol (9)
The time taken by a packet from its source to its destination is called delay [1], which means:
Total Delay= Delay on upper layer + Delay on MAC layer
In this study, we work only on the delay on the MAC layer. Therefore, we represent the delay
on the MAC layer by the terminating the renewal process model to develop the work in [3]. In
this case, the collision or busy will cause the time delay of Ti seconds. The delay is represented
as sequence of discrete random variables and terminated by each successful transmission is as
shown in (10).
Sn = T1 + T2 + ⋯ + Tn + DSuc (10)
Ti: Represents discrete random variable for time delay in seconds when a station will face in
case of a collision or frozen period. All Ti have the same improper probability distribution
function (F) and probability density function (f).
In this paper the DEmp, DSuc, DCol, DBus are the random variables whose corresponding
probability density functions are PEmp, PSuc, PCol, PBus which are obtained from (5), (6), (8)
and (9). Therefore,
PEmp = 𝑓 DEmp
PSuc = 𝑓 DSuc
PCol = 𝑓 DCol
PBus = 𝑓 DBus
However, PEmp, PSuc, PCol, PBus present the probabilities of the slots or transmission attempts in
which a station will not transmit. Therefore, probability distribution function (F) will be equal to
1 if POwn is added to it.
F defines in the paper as: F ∞ = 1 − POwn
From theory of probability and stochastic we know that the f can be obtained by taking the
derivation of F. On the other hand, F can be obtained by integrating the f. Therefore, from the
renewal process theory [14], if we compare the basic renewal equation with (11):
ex.y
∞
0
. F dy = 1 (11)

105
We realize x is the transform variable or the Laplace transform variable. Therefore, the (12)
could be derived by obtaining the value of x. The process terminates after a time value t:
P M > 𝑡 ≈
1 − F ∞
X. µ
. e−x.t (12)
Where
µ = y. ex.y
∞
0
. F dy (13)
We can consider (11) and (13) as the following:
PEmp . ex.DEmp + Psuc . ex.Dsuc +Pcol . ex.Dcol + PBus . ex.DBus = 1 (14)
DEmp . PEmp . ex.DEmp + DSuc . Psuc . ex.Dsuc +DCol . Pcol . ex.Dcol + DBus . PBus . ex.DBus = µ (15)
As a result of obtaining x value from (14) and μ value from (15), we can estimate (12).
However, estimating P {M>t} allows us to estimate the MAC layer packet delay distribution for
IEEE 802.11 as follows:
P d ∈ a; b = P M > 𝑎 − P M > 𝑏 (16)
P d ∈ 0; c = 1 − P M > 𝑐 (17)
Equations (16) and (17) represent the MAC layer delay distribution (d) as a histogram, which
are based on estimating (12) after time value t.
4. NUMERICAL RESULT
The time delays were calculated with the help of (1), (2), (3), and (4) and the values given in
Table1.
Table 1. System parameters.
Parameter Value
SIFS 28μs
DIFS 128μs
EIFS 456μs
PHYSICAL SLOT 50μs
RTS 350μs
CTS 350μs
ACK 300μs
DATA PACKET 8200μs
NETWORK NODES (n) 20-30 nodes
Service Time (ms) 0-200 ms

106
Therefore, the time delay values are as follows:
DEmp = 50μs
DSuc = 9412 μs
DCol = 478 μs
DBus = 456 μs
The probabilities are calculated with the help of (6), (7), (8), and (9) where n=20. Therefore, the
probabilities values are as follows:
PEmp = 0.3585
PSuc = 0.3585
PCol = 0.0189
POwn = 0.0189
PBus = 0.2453
Now, to obtain the real value of x we need to calculate the roots of (14). It will be easily done
by converting it into polynomial equation and then we obtain a real root as shown in (18) and
(19).Therefore, by substituting eX.50.10−6
= t and approximating (14) we obtain:
0.3585t + 0.2453t9
+ 0.0.0189t10
+ 0.3585t188
= 1 (18)
0.3585t + 0.2453t9
+ 0.0.0189t10
+ 0.3585t188
− 1 = 0 (19)
This equation has 188 possible solutions so we will use the Maple program for numerical
calculation to find the roots of the above equation. The real root for t is 1.000261721.
However, t = eX.50.10−6
then,
1.000261721 = eX.50.10−6
ln 1.000261721 = X. 50. 10−6
X = 5.234
Therefore, by obtaining x value then we can obtain the value for μ from (15). Finally, we use
the values for x and μ to solve (12) during the interval time between 0 and 200ms. The result
produces a histogram between probability and time interval as shown in Figure2.

107
Figure2. Terminating renewal process model (n=20)
5. MODEL VALIDATION
In this section, we compare the result from our mathematical model with the wireless network
behavior based on the Bianchi's model in [4]. Furthermore, the result will also be compared with
the previous work in [3]. In this case, we use the same assumption as in the previous work for τtr
and τnb both are considered to be the same as τ value in [4] as follows:
P = 1 − ( 1 − τ)n−1
(20)
τ =
2
1 + 𝑊 + P. 𝑊. 2xPi
m−1
i=0
(21)
The system parameters used for both mathematical model and simulation are in Table1. In
addition to the system parameters, W denotes initial contention window size and m denotes
maximum back-off stage. Figures3 and 4 provide us knowledge to possible estimation for packet
delay distribution by using (16) and (17). As it can be seen from the figures, the results of the
proposed model and the simulation are very close at each discrete time value. These results
present for Packet Delay Right Tail Distribution Function (RTDF), where (RTDF(x) = P(X>x)
for x∈ℜ probability that packet delay exceeds x). In these experiments the errors do not exceed
0.0082 for networks of 20 nodes and 0.0025 for networks of 30 nodes. Moreover, these
experiments show a good accuracy compared with previous work in [3], where the errors were
0.0332 for networks of 20 nodes and 0.0235 for networks of 30 nodes. On the other hand, we
agree with the previous work in [3], where the effects for DEmp, DCol, DBus are minor in
comparison to that of DSuc. Furthermore, our new mathematical model provides accurate way
for estimating the MAC layer packet delay distribution.

108
Figure3. Terminating renewal process model comparing with simulation (n=20).
Figure4. Terminating renewal process model comparing with simulation (n=30).
6. CONCLUSIONS
Packet delay distribution depends on the MAC layer. The MAC layer provides a way for
channel access. However, several events can happen during the channel access. Therefore, these
events cause delay during transmission. The proposal uses the terminating renewal process
theory for modeling MAC layer delay. The proposed solution considering the difference
between the busy probability and the collision probability, which will lead to improve the
accuracy for estimating the MAC layer packet delay distribution for single-hop wireless
network.

109
This paper provides a good agreement between our mathematical model and wireless network
behavior simulation. Therefore, the model provides prediction of high quality as expected.
In our future work, we will consider to use new system parameters where timeslots are five
times shorter than Bianchi's model parameters. These will enable us to further develop our
model with more realistic parameters.
ACKNOWLEDGEMENTS
The authors wish to thank the anonymous reviewers for their useful comments that have
significantly improved the quality of the presentation. This work has received support from the
Ministry of Higher Education in Libya and Coventry University in UK.
REFERENCES
[1] G.Ming,Y.Pan,andP.Fan,AdvancedinWirelessNetwork:PerformanceModelling,Analysis and
Enhancement,vol.8.NewYork:NovaScience, 2008,pp.103–187.
[2] J.Villalon,P.Cuenca,and L.Barbosa, “Limitations and Capabilities of QoS Support in IEEE
802.11 WLANs,” Communications ,Computers and Signal processing, pp.633-636, August
2005.
[3] S.Ivanov,D.Botvich,and S.Balasubramaniam,“On Delay Distribution in IEEE 802.11 Wireless
Networks,” Computer and Communication (ISCC), pp.254-256,July 2011.
[4] G.Bianchi, “Performance Analysis of the IEEE 802.11 Distributed Coordination Function,”
IEEE Journal on Selected Area in Communications, vol.18, pp.535-547, March 2000.
[5] H.Wu,Y.Peng,K.Long, S.Cheng, and J.Ma,“Performance of Reliable Transport Protocol over
IEEE 802.11 Wireless LAN:Analysis and Enhancement,” Infocom,vol.2,pp.599-607,2002.
[6] P. Chatzimisios,A.C.Boucouvalas,andV.Vitsas, “IEEE 802.11 Packet Delay- A Finite Retry
Limit Analyses,” IEEE Infocom,vol.2,pp. 950-954, December 2003.
[7] P.Chatzimisios, A.C.Boucouvalas,and V.Vitsas, “Delay Analysis of Different Backoff in IEEE
802.11,” IEEE Vehicular Technology Conference (VTC),vol.6, pp.4553-4557,September 2004.
[8] P.Raptis,V.Vitsas, K.Paparrizos,P.Chatzimisios,A.C.Boucouvalas,and P. Adamidis, “Packet
Delay Modeling of IEEE 802.11 Wireless LANs,” In Proceeding of 2nd International
Conference on Cybemetics Information and Technologies, System and Applications (CITSA),
pp.71-76,2005.
[9] P.Raptis,V.Vitsas,and K.Paparrizos,“Packet Delay Metrics for IEEE Distribution Coordination
Function,” Mobile Networks and Applications, vol.14, pp.772-781, December 2009.
[10] H.Qi, D.Malone,and D.Botvich,“Performance of a burst frame based CSMA/CA protocol:
Analysis and enhancement,” Springer NetherlandsWireless Networks, vol.15,pp.87-98,January
2009.
[11] P.Raptis,V.Vitsas, K.Paparrizos, P.Chatzimisios,and A.C. Boucouvalas, “Packet Delay
Distribution of the IEEE 802.11 Distribution Coordination Function,” World of Wireless Mobile
and Multimedia Networks,pp.299-304, June 2005.
[12] S.W.Kang, J.R.Cha, J.H. Kim,W.Dong,and Y.Gu, “A Novel Estimation- Based Backoff
Algorithm in the IEEE 802.11 Based Wireless Network,” Consumer Communications and
Networking Conference (CCNC), pp.01-05,January 2010.
[13] N.Taher,Y.Doudance, B. El Hassan,and N.Agoulmine, “An accurate analytical model for
802.11e EDCA under different conditions with and Communications,pp.01-24,March 2011.
[14] W.Feller “An Introduction to probability Theory and its Applications ”, second ed.,Volume
II,Chapter 9, Paragraph 6 New York: Wiley, 1971.

110
Authors
Mr. Hatm Alkadeki is currently doing PhD research on Performance Enhancement for
Wireless Network Protocol in Department of Computing and Digital Environment, Coventry
University, UK. He graduated from Computer Engineering Department of Engineering
Academy, Tajoura, Libya. He obtained MSc degree in Computer and Control Engineering
from Budapest University of Technology and Economics, Budapest, Hungary. He has worked
in Academy Tajoura as teaching assistant and lecturer. He also worked as lecturer at other universities in
Tripoli. Hatm is also member in Engineering Association in Libya, Global Leaders Programme (GLP)
and IEEE.
Dr. Xingang Wang is currently a senior lecturer in Department of Computing and Digital
Environment, Coventry University, UK. He has worked in Centre for Security,
Communications and Networks Research, University of Plymouth for 5 years as lecturer
before joining the department. He received his PhD in performance modelling design and
analysis of multiple access protocols and Quality of Service schemes for computer networks
from University of Bradford, UK in 2005. He is current research interests include designing
MAC protocols, integration of networks and enhancing performance of networks. He has held a number of
grants and published over 50 publications in these areas. He is currently co-supervising a number of PhD
students on various topics in these areas and is currently involved in one industrial collaboration project on
Quality of Service management in packet switched networks. He has actively served the academic
community and cofounded International Workshop on Performance Modeling and Evaluation in Computer
and Telecommunication Networks. He has also acted as a co-chair for a number of international conferences
and served on program committees for over 10 international conferences and served on program committee
for over 10 international conference/workshops. He has edited a number of journal issues as a guest editor
and been a referee for many international journals including IEEE Wireless Communications Magazine,
Elsevier Journal on Computer Networks, Elsevier Journal on Wireless Communications and ACM/Kluwer
Mobile Networks and Applications Journal. Dr Wang also holds the Microsoft Certified System Engineer
(MCSE) professional qualification and is a member of IEEE and IET.
Dr. Michael Odetayo is a principal lecturer in the Department of Computing at Coventry
University, UK. He obtained his MSc degree from the Department of Computer Science,
Imperial College, London, UK and his PhD Degree in ComputerScience at the University of
Strathclyde, Glasgow, UK. Before completing his PhD,he worked at the Computer Centre of
Ahmadu Bello University, Zaria, Nigeria,where he led many project teams that designed and developed
computer basedsystems for the University and other organisations outside the University. Hejoined De
Montfort University, Leicester, UK, after completing his PhD where he was a senior lecturer in the
Department of Computer Science for many years before moving to Coventry University. His research
areas include Genetic Algorithms, Hybrid Systems, Biomedical Intelligent Systems, Intelligent
Knowledge-Based Systems, e-Commerce, Neural Networks, Image enhancement for Biometric security
systems, Data Mining and Machine Learning.

ESTIMATION OF MEDIUM ACCESS CONTROL LAYER PACKET DELAY DISTRIBUTION FOR IEEE 802.11

More Related Content

What's hot (17)

Similar to ESTIMATION OF MEDIUM ACCESS CONTROL LAYER PACKET DELAY DISTRIBUTION FOR IEEE 802.11 (20)

Recently uploaded (20)

ESTIMATION OF MEDIUM ACCESS CONTROL LAYER PACKET DELAY DISTRIBUTION FOR IEEE 802.11