Professional Documents
Culture Documents
( (
( (
(7)
For RTS/CTS MAC scheme:
[ ]
MSDU 8
(MSDU+28) 8 14 8 20 8 14 8
1158 + +
Data Rate Data Rate' Data Rate' Data Rate'
( ( ( (
+ +
( ( ( (
(8)
Where Data Rate is the transmission rate of a MPDU and
Data Rateis the transmission rate of the MAC control frames
(ACK, RTS and CTS).
III. IMPROVED THROUGHPUT PREDICTION MODELS
A. Throughput Model with IP Fragmentation
The prediction model can be applied to analyze the real
WLANs performance with some modifications made to better
characterize the practical scenarios. The first issue is the IP
fragmentation. In the case of Ethernet, the Maximum
Transmission Unit (MTU) in network layer is 1,500 bytes.
Therefore, packets with size larger than its MTU will be
fragmented in Network layer. Though 802.11b accepts much
larger packets (2,346 bytes), the Network layer is not aware of
that and fragments the packets in order to optimize the
performance. A packet can be fragmented into more than one
segment. All the segment transmission time is calculated as
shown.
First n-1 Segment Transmission Time
[ ]
1
T =
(M+8)
1480
(MPDU') 8 14 8
754 + +
Data Rate Data Rate'
| | ( (
| ( (
\ .
(
(
(9)
Where MPDU = 1500 +36, here n is the total number of
segments in a packet; 1500 is the Network layer MTU size and
36 is the sum of 8 bytes of Logical Link Control (LLC) layer
overhead and 28 bytes of MAC header; and M is the
transmitting packet size.
Last Segment Transmission Time,
[ ]
2
(MPDU'') 8 14 8
T = 754
Data Rate Data Rate
+ +
'
( (
( (
(10)
Where, MPDU = Mod[(M+8), 1480] +56, here 8 is the
UDP header and 56 is the total overheads of the rest of layers
(20 bytes of IP header, 8 bytes of LLC header and 28 bytes of
MAC header).
Therefore, the basic throughput model is modified as shown
in equation (11) to predict the performance of the wireless
network more accurately in real scenarios.
1 2
M 8
Throughput =
T T
+
(11)
B. Throughput Model Accounting For Wireless Channel
Condition
In wireless ad hoc network, the link quality between two
nodes degrades as one moves away from the other. The link
quality, which is reflected by Packet Delivery Ratio (PDR),
varies with the wireless channel condition. The channel quality
depends on many factors, including the separation distance and
the obstacles between the two nodes, as well as external
interference in the channel. In this section, the throughput
prediction model is further improved to take into consideration
of channel condition effect.
1) Wireless Channel Condition Description
There are a few factors to consider when modeling the
channel conditions. The first factor is the distance. As the
distance increases, Signal to Noise Ratio (SNR) of the link
decreases and the Frame Error Rate (FER) in the channel
increases. This subsequently affects the throughput
performance.
Another factor is the external interference. The interference
not only increases the noise level of the desired signal, but also
affects the MAC mechanism in wireless link. The effect can be
classified into two models: the protocol model and the physical
model [5]. In the protocol model, concurrent transmissions
from any node within the interference range of a receiver will
cause a collision, which subsequently result in packet loss from
its corresponding sender. In the physical model, a packet from
the sender is lost at the receiver only if the signal-to
interference- plus-noise-ratio (SINR) [6] falls below a given
threshold. In our study, we first consider only the physical
model of the interference by carrying out measurements in a
controlled scenario without the protocol effect. The
interference effect can then be simply treated as another type of
noise in the throughput evaluation.
As a result, the effect of separation distance and external
interference can be combined and the wireless channel
condition can be modeled by SINR.
2) Improved Throughput Prediction Model
After examining the factors that determine the channel
condition, an analytical model on throughput performance with
consideration of channel error is formulated based on IEEE
802.11b MAC mechanism.
SINR computation has a strong correlation with FER on the
channel. It can be related to FER in the following equations.
First, the probability of bit error in Differential Binary
Phase Shift Keying (DBPSK) [7] can be expressed as:
b 0
-E / N
1
BER= exp
2
(12)
Eb/No can be related to SNR as:
b T
0
E B S
=
N N R
(13)
Where B
T
is the signal bandwidth; R is the transmission
data rate.
Obviously, the error performance of a modulation scheme
varies with different SNR values.
The frame error rate in a channel can be expressed in terms
of BER as:
8L
FER= 1-(1-BER)
(14)
Where 8L here is the frame size in bits.
Furthermore, the FER for an L-bytes long data frame taking
into account of MAC mechanism is derived:
m 1 m
FER (data_L) = 1 (1 FER (24) ) (1 FER (28 L) ) +
(15)
Where m = 1, 2, 3 and 4 is the PHY mode representing 1, 2,
5.5 and 11 Mbps transmission rate in IEEE 802.11b,
respectively; FER
1
(24) is the probability of error of the PLCP
preamble/header transmitted using PHY mode 1; FER
m
(28+L)
is the probability of error of the MPDU including the MAC
overhead.
An ACK frame is transmitted at the rate equals to or lower
than the data frame rate, and is 14 bytes long, which is usually
much shorter than the data frame. Therefore, the error
probability of the ACK frame is very low compared to the error
probability of the data frame, and hence ignored in the
calculation here.
The network throughput accounting for the channel errors is
defined as:
actual
No. of frames sent (1-FER)
Throughput =
Total Time Delay
(16)
Where FER is the frame error rate due to the channel error
and the unit for the throughput is frame per second. And Total
Time Delay = No. of frames sent * Delay per second. , hence,
the network throughput can be converted into bits per second
(bps) by multiplying the frame length and the actual throughput
becomes:
actual
max
8L (1-FER)
Throughput =
Delay per MSDU
= Throughput (1-FER)
(17)
Where Throughput
max
is the maximum throughput for ideal
channel condition.
IV. E XPERIMENTS
Experiments were conducted to verify the accuracy of the
formulated analytical models.
The experiment setup is as shown in Fig. 1. An ad-hoc
network was established between node 1 and node 2.
Throughput performance of the network was measured by
NetIQ Chariot [8] with application scripts that emulate the real
network data flow while network traffics were monitored by
observer node using AiroPeek 802.11 Wireless Network
Protocol Analyzer [9].
Figure 1. Experiment setup
Before verifying the accuracy of the analytical model to
predict the throughput performance with channel error, the
channel condition of the experimental environment has to be
investigated and modeled. A methodology was designed to
model the wireless network channel condition by measuring the
FER against Received Signal Strength (RSS) at different
positions of the mobile ad hoc link.
Node 1 and 2 were in an ad-hoc network with presence of
two wireless Access Points (APs) located in vicinity. The
transmitter was kept stationary while the receiver was moved
around in the ad-hoc network to capture the throughput
performance at different position of the network.
Iperf [10] was used as the traffic generator to create traffic
and to investigate the channel FER by reporting the datagram
loss in the wireless link. Constant Bit Rate (CBR) traffic, with
20 bytes payload and Iperf packet generation rate (16kbps)
were set to be lower than the channel transmission rate
(1 Mbps) so that the traffic in the channel is relatively light
compared to saturation throughput experiments. Under this
condition, it is assumed that there is no packet collision due to
beacons sent by the two APs. Traffic is thus reasonably light in
the channel and it can take in more traffic while maintaining
the quality of the transmission.
Adjacent channel interference effect from the two APs can
be deemed as negligible by choosing more than 3 channels
away from the APs transmitting channels. Therefore only the
physical model of the interference was considered in the
measurement.
Also, the traffic was set to operate in multicast mode so that
MAC-level features such as ACKs and RTS/CTS exchanges
were suppressed. It can then be assumed that MAC mechanism
will not have an effect on the test result. Consequently, the
measured RSS should perfectly reflect packet delivery.
The data of channel FER against RSS was collected at
different positions in the ad-hoc network and the interference
plus noise level under our test condition can be calculated
based on the analytical model presented in Section III Part B.
The wireless channel condition (Interference plus Noise) in our
experiments was found to be at around -90dBm.
After obtaining the SINR value under the condition, the
same experiment setup can be utilized to observe the
throughput performance taking into account of channel error.
V. RESULTS AND ANALYSIS
A. Experimental Results for CSMA/CA
The comparisons of theoretical throughputs and
experimental throughputs obtained in the measurement for
CSMA/CA are presented in Fig. 2.
Figure 2. Measured Throughputs vs. MSDU Size for CSMA/CA
The experimental results are comparatively close to the
theoretical ones at low transmission rate of 1 Mbps and
2 Mbps. However the variation increases slightly as the
transmission rate increases (e.g. 5.5 Mbps, 11 Mbps). The
reason is that, as the transmission rate increases, the packet
delivery probability decreases under same channel conditions.
However, in our theoretical calculations, we assume that
channel condition is perfect (SINR is above certain threshold
and FER is 0).
B. Experimental Results for RTS/CTS
The comparisons of theoretical throughputs and
experimental throughputs obtained in the measurements for
RTS/CTS are presented in Fig. 3.
The presented graphs show a definite relationship between
network throughput and the data rate as well as the frame size.
There is also substantial correlation between the analytical
model and measured throughput graphs. Hence the analytical
model formulated in this study is validated and can be used to
predict throughput for RTS/CTS MAC scheme in any 802.11b
WLANs.
0
500 1000 1500
0
1
2
3
4
5
6
MSDU SIZE (Bytes)
NETWORK THROUGHPUT (Mbps)
THROUGHPUT VS MSDU SIZE FOR RTS/CTS
Theoretical at 11 Mbps
Experimental at 11Mbps
Theoretical at 5.5 Mbps
Experimental at 5.5Mbps
Theoretical at 2 Mbps
Experimental at 2 Mbps
Theoretical at 1 Mbps
Experimental at 1Mbps
0
500 1000 1500 2000 2500 3000 3500 4000 4500
0
1
2
3
4
5
6
7
MSDU SIZE (Bytes)
NETWORK THROUGHPUT (Mbps)
THROUGHPUT VS MSDU SIZE FOR CSMA/CA
T for 1Mbps
M for 1Mbps
T for 2 Mbps
M for 2 Mbps
T for 5 Mbps
M for 5 Mbps
T for 11 Mbps
M for 11 Mbps
T: Theoretical Results
M: Measured Results
Figure 3. Measured Throughputs vs. MSDU Size for RTS/CTS
TABLE II. COMPARISON OF THEORETICAL AND MEASURED RESULT FOR
THROUGHPUT VS. RSS
Theoretical Result Measured Result
RSS (dBm) FER Throughput FER Throughput
-66 0 0.674 0 0.668
-76 0 0.674 0 0.663
-80 4.85E-06 0.674 0.004 0.663
-82 0.0077504 0.669 0.025 0.62
-84 0.559326 0.297 0.255 0.422
-86 0.9999998 0 0.388 0.348
-87 1 0 N/A N/A
C. Experimental Result Accounting for Channel FER
The theoretical results and the measured results taking into
account of channel FER are summarized in TABLE II for
comparison.
The theoretical results were calculated based on the
analytical model stated in Section III Part B.
As it can be seen from the tabulated results, the throughput
performance between the two nodes tends to operate in a very
steady manner until the RSS drops below a certain critical level
(-82 dBm). Below this critical level of SNR, the throughput
appears to drop-off quickly.
Moreover, PDR (or 1-FER) is greatly affected by frame
size. For frame size of 500-bytes, PDR drops off quickly from
1 at around -80 dBm to 0 at -86 dBm, for the RSS range of only
6 dBm. However, as for 20-bytes frame (shown in Table 2),
PDR drops off from 0 at around -81 dBm to 0.9 at around -
90 dBm, for a difference of around 9 dBm.
In addition, it is also evident that, the measured throughputs
are quite close to the expected results when the channel
condition is relatively good. However, when the channel
condition becomes poorer (RSS values below -84 dBm), the
measured throughputs are much higher than expected values.
One of the reasons is due to the retransmission effect
implemented by the wireless LAN adaptor. The retry limit for
short packet transmission is 7. A packet will be discarded only
when all 7 retransmissions fail. Most of the packets will get
through after a few retransmissions. This in turns increases the
probability of successful transmission of a packet, hence
improving throughput performance even when the channel
condition becomes poorer.
VI. CONCLUSION & FUTURE WORK
In this paper, a general throughput prediction model was
presented for IEEE 802.11b WLANs. Based on the theoretical
analysis and experimental study, the basic throughput
prediction model was then further improved to take into
consideration of IP fragmentation and channel FER.
Experiments were conducted to verify the accuracy of the
proposed models. Detailed experimental setup was presented
and the obtained experimental results were also analyzed in this
paper.
As for future work, throughput performance for multi-user
multi-hop scenario can be investigated to see how physical
model rather than protocol model is work and if implemented
in MAC together with information from Network layer can
help to improve throughput performance.
REFERENCES
[1] "IEEE Std 802.11-1997 Information Technology- telecommunications
And Information exchange Between Systems-Local And Metropolitan
Area Networks-specific Requirements-part 11: Wireless Lan Medium
Access Control (MAC) And Physical Layer (PHY) Specifications,"
IEEE Std 802.11-1997, pp. i-445, 1997.
[2] "Supplement To IEEE Std 802.11-1997 - Part 11: Wireless LAN
Medium Access Control (MAC) And Physical Layer (PHY)
Specifications: Higher-speed Physical Layer Extension In The 2.4 GHz
Band," IEEE Std 802.11b-1999, pp. i-90, 2000.
[3] J. T. Geier, Wireless LANs 2001.
[4] J. Jangeun, P. Peddabachagari, and M. Sichitiu, "Theoretical maximum
throughput of IEEE 802.11 and its applications," 2003.
[5] P. Gupta and P. Kumar, "The capacity of wireless networks," IEEE
Transactions on information theory, vol. 46, pp. 2.
[6] J. Bardwell, "You believe you understand what you think I said,
http://www.connect802.com/download/techpubs/"
[7] Bernard Sklar., Digital Communications: Fundamentals and
Applications, 2
nd
Edition.
[8] "NetIQ Corporation, http://www.netiq.com."
[9] "Wild Packets AiroPeek,
http://www.wildpackets.com/products/airopeek/overview."
[10] "Iperf, http://dast.nlanr.net/Projects/Iperf/."
[11] A. Di Stefano, G. Terrazzino, L. Scalia, I. Tinnirello, G. Bianchi, and C.
Giaconia, "An experimental testbed and methodology for characterizing
IEEE 802.11 network cards," 2006.
[12] F. Cali, M. Conti, and E. Gregori, "Dynamic tuning of the IEEE 802.11
protocol to achieve a theoretical throughput limit," Networking,
IEEE/ACM Transactions on, vol. 8, pp. 785-799, 2000.
[13] C. Chun-Ting, J. del Prado Pavon, and N. Sai Shankar, "Mobility
support enhancements for the WiMedia UWB MAC protocol," 2005.
[14] Q. Daji, C. Sunghyun, and K. G. Shin, "Goodput analysis and link
adaptation for IEEE 802.11a wireless LANs," Mobile Computing, IEEE
Transactions on, vol. 1, pp. 278-292, 2002.
[15] D. Dhotataut and I. Guerin-Lassous, "Experiments with 802.11b in ad
hoc configurations," 2003.
[16] A. Di Stefano, G. Terrazzino, L. Scalia, I. Tinnirello, G. Bianchi, and C.
Giaconia, "An experimental testbed and methodology for characterizing
IEEE 802.11 network cards," 2006.
[17] F. Eshghi and A. K. Elhakeem, "Performance analysis of ad hoc wireless
LANs for real-time traffic," Selected Areas in Communications, IEEE
Journal on, vol. 21, pp. 204-215, 2003.
[18] V. Gupta and A. Dixit, "The design and deployment of a mobility
supporting network," 1996.
[19] P. Hong Seong and J. Myoung Soon, "Performance analysis of mobility-
supported wireless network system," 2003.
[20] M. J. J. van Nielen, "The impact of mobility on intelligent network
functions," 1992.
[21] [21] M. Bhatt, R. Chokshi, S. Desai, S. Panichpapiboon, N.
Wisitpongphan, and O. K. Tonguz, "Impact of mobility on the
performance of ad hoc wireless networks," 2003.
[22] J. A. Hartwell and A. O. Fapojuwo, "Modeling and characterization of
frame loss process in IEEE 802.11 wireless local area networks," 2004.
[23] M. Yunqian, "Improving wireless link delivery ratio classification with
packet SNR," 2005.