A RELAXED PROBING RATE ADAPTATION IN IEEE802.11 WLANS

ad Biaz and Shaoen Wu

Dept. of Computer Science and Software Engineering, Dustan hall 107, Auburn University, Al 36849, USA

Keywords:

Beacon, Rate Adaption.

Abstract:

The availability of multiple rates in IEEE802.11 WLANs and instability of wireless channel conditions call

for data rate adaption algorithms to optimize network performance. Rate adaption is the process of assessing

instantaneous channel conditions and determining the most appropriate data rate. This paper presents a relaxed

probing rate adaptation scheme to determine the most appropriate instantaneous data rate for both downlink

and uplink channels, especially in the case where control frames such as RTS/CTS are not available. For this

goal, the proposed scheme exploits the mandatory management beacon frame, thus without requiring probing

frames like RTS/CTS necessary for other so far proposed schemes. Ns-2 simulations with IEEE802.11b of the

proposed scheme yield more than 100% throughput improvement in high density networks. Also, simulations

show that the proposed scheme is insensitive to beacon interval.

1 INTRODUCTION

Wireless signals are by nature instable and sensitive to

distance and disturbances (e.g. other wireless source,

temperature and humidity, etc.). Therefore, multiple

rates (modulation schemes) are necessary to enhance

wireless communications. When a wireless channel

is in good condition, it can support high data rates,

otherwise it requires more robust modulation schemes

yielding low data rates. Rate adaptation is the pro-

cess of estimating the instantaneous wireless channel

conditions and then determining the most appropriate

data rate the channel could support.

Generally, rate adaptation strategies can be cate-

gorized into probing or non probing channel schemes.

Non probing rate adaptation algorithms usually are

driven by some metrics such as frame loss rate that

indirectly reﬂects the channel conditions. These adap-

tation schemes do not use control frames such as

RTS/CTS to probe wireless channel conditions. Rep-

resentatives of this category include Auto Rate Fall-

back (Kamerman and Monteban, 1997), and Adaptive

Auto Rate Fallback (M. Lacage, M. Manshaei, and T.

Turletti, 2004). These schemes increase the data rate

if multiple consecutive data frames are successfully

transmitted. If the data frame transmission fails for

several times, then the data rate is decreased. These

schemes are appealing because they do not require

control frames to probe the channel conditions.

For probing rate adaptation schemes, the sender

sends out a probing frame such as RTS, and then

gets a feedback frame from a receiver to complete

the negotiation of an appropriate data rate. Propos-

als Sadeghi, et al (B. Sadeghi, V. Kanodia, A. Sabhar-

wal, and E. Knightly, 2002), Qiao, et al (D. Qiao, and

S. Choi, and K. Shin, 2002), Pavon and Choi (Pavon

and Choi, 2003), and Holland et al (Holland et al.,

2001) are typical probing schemes. Probing data

rate schemes estimate better the channel state than

non probing schemes because non probing schemes

get an implicit information on the channel condition.

Moreover, the implicit information may be mislead-

ing. For example, the schemes that monitor frame loss

rate conclude that channel conditions are poor when

the frame loss rate increases. This may be wrong if

frames are lost due to congestion. The disadvantage

of probing schemes is that probing frames incur some

communications overhead.

This paper proposes a relaxed probing data rate

adaptation scheme. The key difference between

Biaz S. and Wu S. (2007).

A RELAXED PROBING RATE ADAPTATION IN IEEE802.11 WLANS.

In Proceedings of the Second International Conference on Wireless Information Networks and Systems, pages 93-98

DOI: 10.5220/0002150600930098

 SciTePress

the proposed scheme and other probing schemes is

the channel probing mechanism used. The pro-

posed strategy does not use probing frames such as

RTS/CTS that are required by the other schemes,

but rather exploits the mandatory broadcast beacon

frame. Note that RTS/CTS control frames are op-

tional in IEEE 802.11 standard and are not always

used, especially for short data frames. The contri-

butions of this work are: 1) a beacon based probing

data rate scheme. This scheme is characterized as “re-

laxed” because it does not require control frame over-

head such as RTS/CTS; 2) rate estimation for the very

ﬁrst frame of a communication session ; 3:) this work

evaluates the impact of beacon interval on the accu-

racy of the proposed data rate adaptation scheme.

The remaining of this paper is organized as fol-

lows: Section 2 outlines the related work in data rate

adaptation. Section 3 discusses the motivation. Sec-

tion 4 details the proposed scheme. The simulation

results are presented in Section 5. Section 6 concludes

this paper.

2 RELATED WORK

Data rate adaptation has been extensively studied in

last several years in wireless networks from advanced

CDMA network (Bender and et al, 2000) to wireless

local area network like IEEE802.11 (IEEE802.11,

1999). Non probing data rate adaptation schemes esti-

mate the data rate without using probing frames such

as RTS/CTS frames. For example, Auto Rate Fall-

back (ARF) (Kamerman and Monteban, 1997) was

proposed for Lucent WaveLan-II WLAN product. In

this protocol, the data rate adaptation is completed in-

dependently at the sender side without any informa-

tion from the receiver. When the sender fails twice

to transmit a data packet, it automatically decreases

(falls back) its data rate to the next lower level, (for

instance from 5.5 Mbps to 2 Mbps). If the transmis-

sion succeeds for ten consecutive times at the same

data rate, the sender infers that the channel condition

is good enough to support higher data rates and thus

increases its data rate to a higher level. ARF deﬁ-

nitely suffers from data rate ﬂuctuations. For instance,

when the channel condition is good for 5.5 Mbps, but

not good enough for 11 Mbps, the sender will suc-

cessfully transmit ten consecutive times at 5.5 Mbps.

Therefore, the sender will increase the data rate to 11

Mbps and will likely fail transmissions at that rate,

leading to a decrease to 5.5 Mbps. This scenario will

likely repeat itself leading to ﬂuctuations.

In Receiver Based Rate Adaptation (RBRA) (Hol-

land et al., 2001), the authors propose a probing al-

gorithm to adapt the data rate with the cooperation

from the receiver. This protocol requires the exchange

of RTS/CTS control frames between sender and re-

ceiver stations before data/acknowledgement packets

are transmitted. The RTS and CTS control frames are

transmitted at the lowest rate so that they are acces-

sible to all stations in carrier sense range. When the

receiver gets the RTS, it determines the best data rate

that it can support under current wireless channel con-

ditions, based on physical layer measurements. Then

it feeds back the selected data rate embedded in the

CTS frame. This proposal can work in both WLAN

and Ad Hoc networks as long as control (probe)

frames such as RTS/CTS precede the data frames.

A hybrid algorithm named Full Auto Rate

(FAR) (Z. Li, and A. Das, and A.K. Gupta and S.

Nandi, 2005) integrates probing and non probing con-

cepts to achieve full data rate adaptation. The au-

thors of FAR argue that receiver based protocols like

RBRA (Holland et al., 2001). can dramatically be im-

proved if the RTS/CTS can be transmitted at an appro-

priate data rate that the wireless channel can support

rather than at some heuristic or lowest rate. Thus, it

determines the rate for RTS/CTS control frames trans-

mission using a non probing scheme.

In a previous work (Wu and Biaz, 2007), we pro-

pose to estimate the initial rate by snifﬁng the peri-

odic beacon frame for the ﬁrst time. This work dif-

fers from that one in three aspects: ﬁrst, we consider

the data rate estimation for both downlink and uplink

of the wireless channel; Second, we evaluate the im-

pact of beacon period on the rate estimation accuracy;

Third, we compare the role of different coefﬁcients in

the adaptive estimation of the data rate.

3 PROBLEM FORMULATION

Although the rate adaptation is well studied, there still

exists some unaddressed problems.

3.1 Problem #1: How to Probe without

RTS/CTS Control Frames?

In wireless network, when two nodes that can not

sense each other transmit to the same receiver, the

transmissions get collided at the receiving station.

This is called hidden terminal problem. To mitigate

such collision, In IEEE802.11 networks, RTS/CTS is

introduced to clear the channel before the data frame.

However, RTS/CTS control frames are pure overhead

for data frames. These control frames are optional

and are recommended only for large data frames. If a

data frame is small, (which is highly possible in real

WINSYS 2007 - International Conference on Wireless Information Networks and Systems

time applications such as VoIP, video and so on,) the

use of RTS/CTS dramatically increases the overhead.

According to Garg and Kappes (Garg and Kappes,

2003), if a unique 802.11b station transmits VoIP traf-

ﬁc with 160 byte frames, the data packet efﬁciency

drops to about 12% in 802.11b (IEEE802.11b, 1999)

networks at 11 Mbps with RTS/CTS control frames.

Therefore, RTS/CTS frames are recommended only

when the size of data packet exceeds a certain thresh-

old (2347 recommended in IEEE 802.11). For com-

munications not using RTS/CTS frames, most prob-

ing data rate adaptation schemes can not work as they

require RTS/CTS frames. However, probing schemes

can estimate the data rate more promptly and accu-

rately than non probing schemes. Therefore, commu-

nications not using RTS/CTS frames call for a new

probing scheme with minimal overhead.

3.2 Problem #2: Rate Adaptation for

RTS/CTS Frames

In multi-rate IEEE802.11, the recommended rate for

the control frames RTS/CTS is the lowest data rate

so that they can be captured by every station in trans-

mission range. Full Auto Rate (Z. Li, and A. Das,

and A.K. Gupta and S. Nandi, 2005) scheme has an-

alyzed and concluded that throughput improvement

can be achieved if the RTS/CTS frames are transmit-

ted at an appropriate data rate, rather than at the low-

est data rate. Thus, a well-predicted instantaneous

data rate for RTS/CTS beneﬁts the entire network per-

formance.

3.3 Problem #3: Rate Adaptation For

First Transmission

None of the data rate adaptation schemes addresses

the estimation of the data rate at the very beginning

of a communication. Suppose station X wants to ini-

tiate packet transmission to station Y. If station Y is

inactive for a long time (e.g. several minutes), the

sender station X is unaware of the link status to sta-

tion Y. It does not know which data rate is suitable to

restart transmissions. Thus, the initial data rate must

be somehow estimated.

To address the above problems, this work pro-

poses a relaxed probing data rate adaptation scheme

based on the periodic Beacon frame. Note that al-

though the simulation is based on IEEE 802.11 stan-

dard, this strategy is also applicable to all other wire-

less networks with periodic broadcast frames.

4 PROBING RATE ADAPTATION

WITH BEACON

This section details the procedure of the proposed

data rate adaptation with beacon frame. The adap-

tation of initial data rate consists of uplink and down-

link rate adaptation depending on the location where

rate adaptation occurs (at access point or client sta-

tion). Uplink is denoted as the wireless link from a

client station to its access point. And the downlink

refers to the opposite direction. After the initial trans-

mission, either the access point or the client station

can adapt the data rate based on received data or ac-

knowledgement frames.

4.1 Uplink Initial Rate Adaptation

When a client station needs to initiate a commu-

nication to its access point, it has to adopt an ap-

propriate data rate for the initial transmission. In a

802.11 WLAN, the beacon is a mandatory manage-

ment frame. It is broadcast periodically by an ac-

cess point for synchronization and association pur-

poses. Therefore, this periodic frame offers an oppor-

tunity for the wireless client stations to estimate the

uplink data rate. Since the beacon frame is manda-

tory, it does not introduce any overhead for channel

condition probing. When a client station receives the

beacon frame, it retrieves the statistics of the channel

conditions (e.g. signal to noise ratio, signal strength,

and error rate) from the physical adapter. Based on

such wireless channel information, the client station

can determine the best data rate and record it for im-

mediate or future use. The recorded information in-

cludes the modulation level or data rate and the time

when the last beacon was received. When this station

needs to transmit a data frame, it looks up for the data

rate it can use. Then it instructs the physical layer to

transmit the data packet with the appropriate modula-

tion. This scheme is especially accurate for slow fad-

ing channels. Although such estimation might be out

of date for fast fading channels in case of long beacon

interval, it is still better than a random selected initial

rate.

4.2 Downlink Initial Rate Adaptation

While there are beacon frames from an access point

to the client stations there are none towards the access

point. If the access point needs to initiate a transmis-

sion to a particular client station, two options are pos-

sible: the ﬁrst is the lowest data rate, and the second is

the rate most recently used. Furthermore, if the low-

est rate is selected, we propose to fragment the ﬁrst

A RELAXED PROBING RATE ADAPTATION IN IEEE802.11 WLANS

data frame so that the length of the short fragment is

equal to the length of an RTS frame. This short frag-

ment functions as a channel reservation and probing

as RTS does. Since the the frame is short, even it is

transmitted at the lowest rate, the transmission time is

negligible and does not impact much the network per-

formance. Then, the new data rate for the remaining

large fragment can be estimated from the acknowl-

edgment frame.

4.3 Adaptive Rate Adaptation After

Initial Rate

After initial rate is adopted, both the access point and

its client stations can adapt their transmission rate

from the exchanged frames. Also, the client station

still can adapt its data rate from the periodic beacon.

We propose an adaptive algorithm for the ongoing

communications.

For instantaneous data rate adaptation, if the rate

is estimated only from the last received frame, the

data rate will ﬂuctuate due to instantaneous channel

condition variations. The performance suffers espe-

cially when a mobile station is experiencing the ping-

pong effect in handoff, where the data rate might

ﬂuctuate between two levels of rates. This results

in costly (in power) retransmissions. Multiple spuri-

ous retransmissions deﬁnitely hurt the network per-

formance. Therefore, to solve the above data rate

ﬂuctuations problem, the data rate should be adjusted

from the latest frame, but adaptively from multiple

past frames information.

Compared to the available discrete data rate lev-

els(e.g. 2, 5.5, 11.), channel statistics like signal

strength has better granularity for adaptive algorithm.

Thus, signal strength is employed in the following for-

mula. An adaptive low pass ﬁlter coefﬁcient is intro-

duced to smooth signal strength impacted by the wire-

less channel upon history signal strength information.

RSSI

= (1− α) ∗ RSSI

t−1

+ α∗ rssi (1)

where 0 < α < 1.

Here, the RSSI

is the adopted signal strength. It

predicts the channel variation trend. rssi is the in-

stantaneous signal strength obtained from a received

frame.

α is further dynamically updated as follows:

α =

rssi− Threshold

low

Threshold

high

− Threshold

low

(2)

Threshold

low

and Threshold

high

are the signal

strength thresholds for the current rate.

1.1e+07

5.5e+06

2e+06

3.4 3.45 3.5 3.55 3.6 3.65 3.7

Data Rate

Time

ARF Data Rate Fluctuation

Figure 1: Auto Rate Fallback rate ﬂuctuation.

5 SIMULATION RESULTS

Simulation conﬁguration: Simulations are per-

formed on ns-2 (NS2, 2006) version 2.30. Three lev-

els of data rates are used: 11 Mbps, 5.5 Mbps and

2 Mbps. The data rate 2 Mbps is deﬁned as the

basic rate. Channel fading is conﬁgured as Ricean

model. These simulations compare the proposed

scheme against the ARF proposal and evaluate the

impact of the beacon interval and the low pass ﬁlter

parameters used in formula 1. The IEEE 802.11 cell

is a 500mx500m square with the access point located

at the center. The radio range is 250m. In all ex-

periments, each client sends UDP CBR trafﬁc to the

access point at 4 Mbps.

Data Rate Smoothing: Figure 1 and Figure 2

show the data rate adaptation resulted from ARF and

our scheme respectively. The channel condition is

stable for 5.5 Mbps in the experiment with ARF.

From Figure 1, data rate in ARF ﬂuctuates heavily be-

cause it adapts the data rate based on frame loss that

is vaguely reﬂect channel conditions. With dynamic

channel conditions, the proposed scheme is tested and

depicted in Figure 2. As expected, its adaptive ability

reduces the sensitivity to channel variations and over-

comes rapid ﬂuctuations.

Throughput performance: Figure 3 illustrates the

improvement of network upon network density (dif-

ferent number of client stations). The X-axis repre-

sents the network density and Y-axis stands for im-

provement over ARF. In this scenario, all nodes are

stationary at their speciﬁed positions, which are uni-

formly distributed along the four diagonal lines to the

center of square. One example of the layout with

8 client stations is shown in Figure 4 (access point

is at the center). It can be observed from Figure 3

that as the density increases, our scheme can improve

WINSYS 2007 - International Conference on Wireless Information Networks and Systems

1.1e+07

5.5e+06

2e+06

30 32 34 36 38 40

Data Rate (bps)

Time (s)

Data Rate in Sender Adaptive Auto Rate

Figure 2: Data Rate Adaptation.

120%

100%

80%

60%

40%

20%

25 16 9 4

Improvement

The Number of Nodes

Figure 3: throughput improvement with stationary location.

throughput sharply and eventually achieve more than

100% improvement over ARF in such nodes layout.

This can be explained by the robustness of our scheme

to frame loss. When network density increases, the

frame loss rate will increase due to more collisions.

ARF will decrease its rate in case of frame loss re-

gardless of the cause of the losses. Our scheme is in-

sensitive to frame loss rate, and reacts only to received

signal strength variations. Thus, more throughput im-

provement can be achieved in denser scenarios.

In the second throughput experiment, all client

stations are roaming in the square ground with speed

from 2 m/s to 20 m/s randomly. We still test the

improvement upon network density. The result is

demonstrated in Figure 5. Generally, our strategy

can achieve more than 50% improvement over ARF.

In some density, the improvement is even more than

100%. As might be observed, when the network den-

sity becomes too high, the improvement decreases

dramatically. This falling is caused by the increased

frame loss from the collision, which can not be dis-

criminated from channel fading by this strategy.

Figure 4: network layout with 8 client nodes.

150%

125%

100%

75%

50%

4 8 12 16 20 24 28 32 36 40

Improvement

The Number of Nodes

Figure 5: throughput improvement with roaming nodes.

Then, we evaluate the beacon interval impact

on the effect of the performance of our adaptation

scheme. Figure 6 depicts the result. In this ﬁgure,

the X-axis denotes throughput and the Y-axis repre-

sents the beacon interval varying from 20 ms to 200

ms. The result are promising and surprising: the per-

formance is insensitive to the beacon interval. This

observation suggests that even if the rate estimation

interval is not short, the estimation is still pretty up-

to-date on slow fading channel.

Figure 7 depicts the throughput improvement with

the adaptive low pass ﬁlter and different constant low

pass ﬁlters with formula 1. The X-axis denotes the

number of nodes in the network area of 500x500 and

the Y-axis represents the throughput improvement.

The different lines in Figure 7 show the improvement

results for different coefﬁcients used. From this ﬁg-

ure, we can observe that, the adaptation scheme using

a constant low pass ﬁlter is outperformed by that of

the adaptive ﬁlter in formula 2. Also, note that differ-

ent ﬁxed coefﬁcients do impact on the improvement.

A RELAXED PROBING RATE ADAPTATION IN IEEE802.11 WLANS

1.1

1.2

1.3

1.4

1.5

20 40 60 80 100 120 140 160 180 200

Throughput (KBps)

Beacon Interval (ms)

Figure 6: throughput of different beacon intervals.

80%

60%

40%

20%

4 8 12 16 20 24 28

Throughput Improvement(%)

Nodes

coef=0.1

coef=0.3

coef=0.5

coef=0.7

coef=0.9

Adaptive Coef

Figure 7: Impact on improvement by coefﬁcients.

6 CONCLUSION

This work proposes a relaxed probing rate adaptation

to exploit the broadcast management beacon frame in

802.11 networks. The scheme includes also an adap-

tive algorithm designed to smooth out the data rate

during temporary variations of channel conditions.

The simulations yield promising throughput improve-

ment over ARF and reveal that the beacon interval has

little impact on the performance improvement on slow

fading channel.

REFERENCES

B. Sadeghi, V. Kanodia, A. Sabharwal, and E. Knightly

(2002). Opportunistic Media Access for Multirate Ad

Hoc Networks. In MOBICOM02.

Bender, P. and et al (2000). High-Speed Wireless Data Ser-

vice for Nomadic Users. Multiple Access for Broad-

band Wireless Networks, pages 70–77.

D. Qiao, and S. Choi, and K. Shin (2002). Goodput Anal-

ysis and Link Adaptation for IEEE 802.11a Wireless

LANs. IEEE TRANSACTIONS ON MOBILE COM-

PUTING, 1:278–292.

Garg, S. and Kappes, M. (2003). An Experimental Study

of Throughput for UDP and VoIP Trafﬁc in IEEE

802.11b Networks. In wcnc2003.

Holland, G., Vaidya, N., and Bahl, P. (2001). A rate-

adaptive MAC protocol for multi-hop wireless net-

works. In In Proceedings of ACM MOBICOM’01.

IEEE802.11 (1999). http://standards.ieee.org/getieee802/

download/802.11-1999.pdf.

IEEE802.11b (1999). http://standards.ieee.org/getieee802/

download/802.11b-1999.pdf.

Kamerman, A. and Monteban, L. (1997). WaveLAN II:

A high-performance wireless LAN for the unlicensed

band. Bell Labs Technical Journal, pages 118–133.

M. Lacage, M. Manshaei, and T. Turletti (2004). IEEE

802.11 Rate Adaptation: A Practical Approach. In

MSWiM04.

NS2 (2006). http://www.isi.edu/nsnam/ns/.

Pavon, J. and Choi, S. (2003). Link Adaptation Strategy

for IEEE 802.11 WLAN via Received Signal Strength

Measurement. In ICC.

Wu, S. and Biaz, S. (2007). Loss Differentiated Rate Adap-

tation in Wireless Networks. Technical report, Auburn

University.

Z. Li, and A. Das, and A.K. Gupta and S. Nandi (2005). Full

auto rate MAC protocol for wireless ad hoc networks.

IEE Procedings on Communication, 3:311–319.

WINSYS 2007 - International Conference on Wireless Information Networks and Systems