You are on page 1of 5

2009 Fifth International Conference on Mobile Ad-hoc and Sensor Networks

QoS Performance Analysis of IEEE 802.15.4 MAC


in LR-WPAN with Bursty Error Channels
Jianliang Gao1 , Jia Hu2 , Geyong Min2 , Li Xu1
1. School of Mathematics and Computer Science, Fujian Normal University, China
Email: { jlgao, xuli}@fjnu.edu.cn
2. Department of Computing, School of Informatics, University of Bradford, Bradford, U.K.
E-mail: { j.hu2, g.min }@brad.ac.uk

AbstractThe IEEE 802.15.4 standard denes physical layer


and Medium Access Control (MAC) layer protocols for the Low
Rate Wireless Personal Areas Network (LR-WPAN). The analytical models of 802.15.4 MAC have been primarily developed under
the assumptions of the ideal channels or uniform error channels
which fail to capture the characteristics of bursty and correlated
channel errors in the practical wireless network environment. In
this paper, we propose an analytical model for 802.15.4 MAC
in LR-WPAN in the presence of bursty error channels. This
model can be adopted to obtain the Quality-of-Service (QoS)
performance metrics in terms of throughput, service time, and
total delay. Utilizing the analytical model, we investigate the QoS
performance of 802.15.4 MAC under various trafc loads, backoff
parameters, numbers of stations, and channel conditions.

1. I NTRODUCTION
Low Rate Wireless Personal Areas Network (LR-WPAN), a
short-range and low data rate wireless network introduced for
low cost and low power communications, [15] has found a
variety of applications such as health care, home automation,
environmental monitoring, and industrial control. The IEEE
802.15.4 standard [5] is uniquely designed to meet the
requirements of the low bandwidth, low energy consumption
applications in the LR-WPAN. In the IEEE 802.15.4 Medium
Access Control (MAC) specication, the Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) protocol
is employed to coordinate the distributed channel access of
wireless nodes.
Recently, there have been many studies focusing on the
performance analysis of IEEE 802.15.4 MAC protocol [1],
[4], [6], [7], [9], [11], [13], [14], [17], [18]. For instance, Lee,
Lee and Chung [6] employed an embedded Markov chain
model to analyze the saturation throughput performance in
the IEEE 802.15.4 LR-WPAN. Ramachandran, Das, and Roy
[14] proposed a Markov model to evaluate the throughput and
energy consumption of 802.15.4 MAC. Park, Kim et al. [11]
and Pollin, Ergen et al. [13] presented analytical models
for the 802.15.4 MAC based on the bi-dimensional Markov
chain proposed in [2], which has been extensively adopted for
modelling the IEEE 802.11 MAC protocol. Gao, Hu and Min
analyzed the unsaturated throughput performance based on an
average backoff period model [4]. Ling, Cheng, Mark et al. [7]
978-0-7695-3935-5/09 $26.00 2009 IEEE
DOI 10.1109/MSN.2009.25

evaluated the throughput and service time of the 802.15.4


MAC using the renewal process. Bhatti, Mehta, Sahinoglu
et al. [1] proposed an enhanced 802.15.4 MAC protocol to
improve the performance of beacon-enabled transmission.
To the best of our knowledge, existing analytical models
on 802.15.4 MAC have been primarily developed based on
the assumptions of ideal channels or uniform error channels.
However, in the practical wireless network environment, the
channel errors always exhibit bursty and correlated nature [10],
[12]. In LR-WPAN that is usually deployed for health monitoring and environment surveillance, any movement of people,
change of natural conditions or other radio interferences can
affect the quality of wireless channels, generally, leads to some
degree of channel errors. Since LR-WPAN and its applications
have gained rapid growth over recent years, it is critical to take
the bursty characteristics of channel errors into consideration
for the analysis of 802.15.4 MAC. As a step towards this end,
this paper evaluates the QoS performance of 802.15.4 MAC
under the bursty channel errors. The major contributions of
this paper include:
1) We model the backoff process of 802.15.4 MAC based on
the mean value analysis method [16], which has an advantage
of lower complexity over the conventional Markovian approach. A two-state Markov model is adopted to characterize
the bursty error channels.
2) We study the QoS performance of 802.15.4 MAC in
terms of throughput, service time, and total delay under bursty
error channels and Bernoulli arrival trafc. We investigate
the QoS performance of 802.15.4 MAC under various trafc
loads, backoff parameters, numbers of stations, and channel
conditions. This model is demonstrated to be a cost-effective
tool for evaluating the QoS performance of 802.15.4 MAC.
The rest of this paper is organized as follows. In section
2 we introduce the two-state Markov model for wireless
channels. Section 3 briey reviews the IEEE 802.15.4 MAC
protocol. Section 4 presents the analytical model in detail.
The performance analysis is conducted in Section 5. Finally,
section 6 concludes the paper.

2. M ARKOV M ODEL FOR W IRELESS C HANNELS

252

NB=0, CCA=2
BE=BEmin
Synchronization
Backoff Period
a random
(2BE-1)
Perform CCA on
backoff period
boundary

Fig. 1.
B BEP

Markov Model for Wireless Channels

CAP
GTS

CFP
GTS

Channel is
idle?

True

False
GTS

NB=NB+1, CCA=0
BE=min(BE+1, BEmax)

CCA=CCA-1

Inactive Period

False
0

10 11 12 13 14 15

NB>
maxBackoff?

Time

True

Superframe Duration =aBaseSuperframeDurationh2SO symbols


Active Period

Failure

CCA=0?

False

True
Success

Beacon Interval =aBaseSuperframeDurationh2BO symbols

Fig. 3.
Fig. 2.

Slotted CSMA/CA protocol

Structure of a MAC Superframe

The bursty characteristics of channel errors can be modelled


by a two-state Markov model [8], [10], [12]. As shown in
Fig. 1, the wireless channel is characterized by a Markov
chain alternating between a Good state and a Bad state. The
transition rate from the good to the bad state is represented by
vg while that from the bad to the good state is denoted as vb .
The durations of the Good state and Bad state are exponential
random variables with means vg1 and vb1 , respectively. In the
Good state, the channel is error free and a frame transmission
would be successful if no collision happens. On the contrary,
the frame transmission fails due to channel errors when the
channel is in the Bad state. The steady-state probabilities of
the channel being in the Good and Bad state, g and b , can
be expressed as
vb
g =
(1)
vg + vb
b =

vg
vg + vb

(2)

3. T HE IEEE 802.15.4 MAC P ROTOCOL


In this section, we briey present the main features of
the IEEE 802.15.4 MAC protocol [5], which offers two
alternative operation modes: beaconless and beacon-enabled.
The beaconless mode employs unslotted CSMA/CA protocol
to control the channel access among the nodes. In this mode,
a node has to wait for a random backoff period before starting
transmission. If the channel is sensed busy after a backoff
period, the node can not retry transmitting until waiting for
another random backoff period. The beaconless mode is also
known as asymmetric channel access mode, which is widely

253

used for control of asymmetric powered devices such as


wireless light switch. While the beacon-enabled mode employs
the slotted CSMA/CA protocol, in which prior to a random
backoff period, each node strictly synchronizes its actions
and transmissions with the others using time slots aligned
with beacon slots broadcasted by a coordinator. As soon as
the backoff period is over, the node starts two consecutive
channel sensing, referred to as the rst Clear Channel Assessment (CCA1 ) and second CCA (CCA2 ). The node can
only start transmission if the channel is sensed idle during
two consecutive CCAs, otherwise, it has to wait for another
random backoff period [5], [7]. Since beacon-enabled mode
enables a majority of IEEE 802.15.4 features such as battery
life extension, sleeping, association, etc which are desirable
for health monitoring and environment surveillance, we focus
on the performance of MAC in the beacon enabled model,
namely the slotted CSMA/CA protocol in this paper.
In the slotted CSMA/CA, superframes are sent periodically
by a PAN coordinator for synchronizing the nodes. Fig. 2
illustrates the structure of a superframe, which is divided
into two periods including the active period and inactive one.
The active period consists of 16 equally-spaced time slots
that include network beacon (B), beacon extension period
(BEP ), contention access period (CAP ), and contention free
period (CF P ). All guaranteed time slots (GT Ss) are allocated
in CF P , which must be a multiple of time slot and be
aligned with the end of the active period. More details for
the superframe can be found in [5].A superframe has the
optional inactive period if the superframe order (SO) is less
than the beacon order (BO). If SO equals to BO, there is
no inactive period in the superframe. As shown in Fig. 3, the
slotted CSMA/CA protocol works as follows.

Step 1: The number of backoff stages (N B), CCA counter,


and current backoff exponent (BE) are initially set to 0, 2,
and BEmin respectively. Next, the time slot boundary of each
node is synchronized with the beacon.
Step 2: Each node selects a random backoff period ranging
from 0 to 2BE 1.
Step 3: Each node performs a CCA after decreasing the
backoff counter to zero.
Step 4: If the channel is sensed busy in a CCA, the CCA
counter is reset to zero, the N B and BE are increased by one,
up to the maximum number of backoff stages (maxBackof f )
and BEmax respectively. If N B is larger than maxBackof f ,
the frame is discarded due to transmission failure.
Step 5: If the channel is sensed idle, the CCA counter is
increased by one. The node starts transmission if the CCA
counter reaches two. Otherwise, the CCA counter is reset to
zero and the node has to perform another CCA.
We assume all nodes work in full duty cycle, SO=BO, that
is, without inactive period. We also do not take GT Ss, ACK,
association and retransmission into account in this paper.

2m W
(4)
2
Since a node can only sense the channel when its backoff
counter decreases to zero, the channel sensing probability,  ,
is given by
+[p1 + (1 p1 )p2 ]m+1

 =

p1 = L[1 (1  )n1 ](1 p1 )(1 p2 )

We consider a single-hop star LR-WPAN with n + 1


nodes working under unsaturated conditions and bursty error
channels. Among the n + 1 nodes, one node is assigned to be
the PAN coordinator and the rest n nodes send frames to the
PAN coordinator simultaneously.
Let denote the probability that a node attempts to sense
the channel given that the node is backlogged. Since a node
only senses the channel when there is at least one frame in its
transmission queue, the channel sensing probability of a node,
 is calculated as


= (1 P0 )

(3)

where P0 is the probability that the transmission queue is


empty.
Let p1 and p2 represent the probabilities that the channel is
sensed busy in the CCA1 and CCA2 of a node, respectively.
Let W = 2BEmin be the minimum contention window size.
At the rst backoff stage, the backoff counter is uniformly
distributed between zero and W 1, therefore the average
backoff period length is W/2 [16]. Similarly, the average
backoff period length at the backoff stage i is 2i W/2,where
i [1, m], m = BEmax BEmin . Let Wbackof f denote the
average backoff period length that a node experiences before
attempting to sense the channel. Conditioning on the fact that
the backoff counter reaches the i th backoff stage with the
probability [p1 + (1 p1 )p2 ]i , we can calculate the average
backoff period length, Wbackof f , as follows
m

2i W
i=0

[p1 + (1 p1 )p2 ]i

(6)

p2 equals the probability that there is at least one remaining


node (e.g., j) transmits in the CCA2 of the tagged node (e.g.,
i). This implies that the node j is already in the CCA2 when
the node i is in the CCA1 . Therefore, p2 equals to the product
of the probability that at least one of the (n 1) nodes (e.g.,
j) starts sensing the channel in CCA1 , and the probability that
the CCA2 slot of node j is idle.
p2 = [1 (1  )n1 ](1 p2 )

4.1 Analysis of the Backoff and Transmission Procedures

(5)

The channel is sensed busy if there is at least one node


transmitting during the CCA1 . Assuming each transmission
lasts L slots, p1 is given by

4. A NALYTICAL M ODEL

Wbackof f = (1 p1 )(1 p2 )

1 P0
Wbackof f

Solving the above equation, we can express p2 as



1
1
p2 = 1 +
1 (1  )n1

(7)

(8)

There are two factors resulting in the failure of transmitting


a frame - channel errors and collisions. Given a channel being
in Good state at a time slot, it will remain in this state
for L time slots with the probability evg L . Therefore, the
transmission failure probability due to channel errors, pe can
be calculated as
pe = 1 g evg L

(9)

A transmitting node encounters a collision when at least


one of the remaining nodes transmits in the given time slot.
Therefore, the collision probability, pc can be given by
pc = 1 n  (1  )n1 (1 p1 )(1 p2 )

(10)

With pc and pe , the probability of successful transmitting a


frame in a slot, Psucc , can be expressed as
Psucc = 1 [pc + (1 pc )pe ]

(11)

Given the frame size of L slots, we can calculate the


throughput, T h, as
T h = LPsucc

(12)

The service time is the time duration from the instant that
a Head-of-Line (HoL) frame starts contending for the channel
to the instant that the frame leaves the transmission queue.
The average service time, E[S], can be given by

254

TABLE I
SYSTEM PARAMETERS
Frame size

8 slots

Data rate

BEmin

3,4,5

BEmax

12

vg

100ms, 50ms

vb

10ms

Slot time

0.00032s

MaxBackoff

E[S] =

5. N UMERICAL R ESULTS

250kbits/s

n(1 P0 )
Psucc

(13)

Then the mean service rate, , is readily given by 1/E[S].


4.2 Queueing Model
The trafc source at every node is assumed to be Bernoulli
arrivals with the rate (frame/slot). Thus the interarrival time
is geometrically distributed and the frame arrives at each slot
with the probability of 1 .
The geometric distribution has a probability generating
function (PGF) as
F (z) = [1 (1 )]z 1

(14)

Then the stationary queue size distribution, i , is given by


[3]

i =

1
(1 )i1

i=0
i1

(15)

where = 1 P0 = / and (0 < < 1) is a unique root


of the equation
z = F (z + (1 ))

(16)

Solving the above equation, we have


==

(17)

Therefore, we can have P0 as


P0 = 1 = 1 = 1

(18)

Queueing delay is the time duration from the moment that a


frame enters the transmission queue of the node to the moment
that the frame leaves the queue. By virtue of Littles theorem,
we can derive the mean queueing delay as
E[W ] =

i=0

ii

1
(1 )

(19)

Finally, the average total delay, E[T ], can be calculated by


E[T ] = E[W ] + E[S]

(20)

In this section, we investigate the QoS performance of


802.15.4 MAC protocol under various working conditions. The
single-hop star LR-WPAN is running at 2.4GHz with a channel
data rate of 250kbits/s. Table I lists the system parameters
used in our analysis. Fig. 4 shows the QoS performance
of 802.15.4 MAC in terms of throughput, service time, and
total delay under various trafc loads, backoff parameters,
numbers of nodes, and channel conditions. The minimum
backoff exponent, BEmin , varies from 3 to 5. The trafc
load at each node increases from 0.001 to 0.012 (frame/slot).
We consider three channel conditions where in Case 1 the
channel is error-free, while the channel exhibits bursty errors
in Cases 2 and 3. The mean Good and Bad state durations
are vg1 = 100ms and vb1 = 10ms, respectively in Case 2,
and are vg1 = 50ms and vb1 = 10ms, respectively in Case
3.
Figs. 4(a) and 4(b) depict the performance metrics in terms
of the service time and total delay under different BEmin and
channel conditions as a function of trafc loads. As expected,
the service time and total delay deteriorate as the channel
condition becomes worse with the smaller vg1 . On the other
hand, we can see that both the service time and total delay
grow as the BEmin increases under unsaturated conditions.
This is because that the larger BEmin implies the longer
backoff time for each node and thus leads to the larger service
time and total delay. Although the larger BEmin can also
reduces the collision probability among nodes, the effects are
negligible when the network is working under unsaturated
conditions and the network size is small.
We next turn to analyze the throughput performance under various channel conditions as a function of the number
of nodes with BEmin = 3 and trafc load being 0.005
(frame/slot). As shown in Fig. 4(c), when the network size
is small (the number of nodes is less than 15), the network
works under light loads and the throughput is almost same
under different channel conditions. However, as the number
of nodes grows, the network loads increase and the impact of
bursty channel errors on the throughput is demonstrated.

6. C ONCLUSIONS
Existing analytical studies on the IEEE 802.15.4 MAC
protocol have primarily focused on the performance under
ideal channels or uniform error channels. These assumptions
fail to capture the characteristics of bursty and correlated
channel errors in the practical wireless network environment..
To address this issue, we have proposed a new analytical model
for 802.15.4 MAC in LR-WPAN with bursty error channels.
The QoS performance metrics in terms of the throughput,
service time, and total delay have been derived with the model.
We have investigated the impacts of trafc loads, backoff
parameters, numbers of stations, and channel conditions over

255

150
V1
=100ms BEmin=3
g

100

V1=100ms BEmin=4
g

V1=100ms BEmin=5
g

50

V1
=50ms BEmin=3
g
V1=50ms BEmin=4

0.6

100

V1=100ms
g
g

V1=100ms BEmin=5
g

50

V1=50ms BEmin=3
g

V1=50ms BEmin=4
g

Fig. 4.

V1
=50ms
g

0.5

Without Channel Error

0.4
0.3
0.2

(a) Service time vs. trafc loads with different


BEmin and channel conditions. (The number
of nodes is 10)

V1
=100ms
g

V1=50ms BEmin=5

V1=50ms BEmin=5

0
2
4
6
8
10
12
Bernoulli traffic load (0.001 to 0.012) x 103

BEmin=3

V1=100ms BEmin=4

Normalized throughput

150
Total Delay (slots)

Service Time (slots)

200

2
4
6
8
10
12
Bernoulli traffic load per node x 103

(b) Total delay vs. trafc loads with different


BEmin and channel conditions. (The number
of nodes is 10)

0.1
0

20
40
Number of nodes

60

(c) Normalized throughput with various network sizes (BEmin=3)

Performance metrics of the IEEE 802.15.4 MAC versus trafc loads per node.

the QoS performance of 802.15.4 MAC. Numerical results


have shown that the throughput, service time, and total delay
are clearly affected by the bursty channel errors.

ACKNOWLEGEMENT
This work is partially supported by Fujian Provincial
Department of Science and Technology under grant NO.
2007F5036, the Education Department of Fujian Province with
grant NO. JA08045 and Natural Science Foundation of Fujian
Province of China (No. 2008J0014).
R EFERENCES
[1] G. Bhatti, A. Mehta, Z. Sahinoglu et al., Modied Beacon-Enabled IEEE
802.15.4 MAC for Lower Latency, GLOBECOM 2008, pp. 1-5, 2008.
[2] G. Bianchi, Performance Analysis of the IEEE 802.11 Distributed Coordination Function, IEEE Journal on Selected Areas in Communications,
vol. 18, no. 3, pp. 535-547, 2000.
[3] I. Elhanany and D. Sadot, Queueing Analysis of Markov Modulated
ON/OFF Arrivals with Geometric Service Times, The 22nd Convention
of Electrical and Electronics Engineers, pp. 189-191, 2002.
[4] J. Gao, J. Hu, and G. Min, A New Analytical Model for Slotted IEEE
802.15.4 Medium Access Control Protocol in Sensor Networks, CMC
2009, pp. 427-431, 2009.
[5] IEEE, Wireless Medium Access Control (MAC) and Physical Layer
(PHY) Specications for Low-Rate Wireless Personal Area Networks
(WPANs), 2006.
[6] T. Lee, H. Lee, and M. Chung, MAC Throughput Limit Analysis of
Slotted CSMA/CA in IEEE 802.15.4 WPAN, IEEE Communications
Letters, vol. 10, no. 7, pp. 561-563, 2006.

[7] X. Ling, Y. Cheng, J. Mark, et al., A Renewal Theory Based Analytical


Model for the Contention Access Period of IEEE 802.15.4 MAC, IEEE
transaction on Wireless Communications, vol. 7, no. 6, pp. 2340-2349,
2008.
[8] G. Min, J. Hu, W. Jia et al., Performance Analysis of the TXOP Scheme
in IEEE 802.11e WLANs with Bursty Error Channels, WCNC 2009,
2009.
[9] A. Mishra, and P. Papadimitratos, A Cross Layer Design of IEEE 802.15.4
MAC Protocol, GLOBECOM 2006, pp. 1-6, 2006.
[10] S. Pack, X. Shen, J. W. Mark et al., A Two-Phase Loss Differentiation
Algorithm for Improving TFRC Performance in IEEE 802.11 WLANs,
IIEEE Transactions on Wireless Communications, vol. 6, no. 11, pp.
4164-4175, 2007.
[11] T. Park, T. Kim, J. Choi, et al., Throughput and energy consumption
analysis of IEEE 802.15.4 slotted CSMA/CA, IEEE Electronics Letters,
vol. 41, no. 18, pp. 1017-1019, 2005.
[12] P. P. Pham, S. Perreau, and A. Jayasuriya, New Cross-Layer Design
Approach to Ad Hoc Networks under Rayleigh Fading, IEEE Journal on
Selected Areas in Communications, vol. 23, no. 1, pp. 28-39, 2005.
[13] S. Pollin, M. Ergen, S. Ergen, and B. Bougard, Performance Analysis
of Slotted Carrier Sense IEEE 802.15.4 Medium Access Layer, Proc.
IEEE GLOBECOM06, pp. 1-6, 2006.
[14] I. Ramachandran, A. Das, and S. Roy, Analysis of the Contention Access
Period of IEEE 802.15.4 MAC, ACM transactions on Sensor Networks,
vol. 3, no. 1, 2007.
[15] K. Sohraby, D. Minoli, and T. Znati, Wireless Sensor Networks:
Technology, Protocols, and Applications, Wiley, 2006.
[16] Y. Tay and K. Chua, A Capacity Analysis for the IEEE 802.11 MAC
Protocol, ACM Wireless Networks, vol. 7, no. 2, pp. 159-172, 2001.
[17] K. Yedavalli and B. Krishnamachari, Enhancement of the IEEE 802.15.4
MAC protocol for scalable data collection in dense sensor networks, Proc.
IEEE WIOPT2008, pp. 152-161, 2008.
[18] J. Zheng and M. Lee, A Comprehensive Performance Study of IEEE
802.15.4, in Sensor Network Operations, IEEE Press Wiley, pp. 218-237,
2006.

256

You might also like