AN EVALUATION OF A CONSERVATIVE TRANSMIT
POWER CONTROL MECHANISM ON AN INDOOR 802.11
WIRELESS MESH TESTBED
Karol Kowalik, Marek Bykowski, Brian Keegan and Mark Davis
Communications Network Research Institute, Dublin Institute of Technology, Ireland
Keywords:
Transmit Power Control, Wireless Mesh, 802.11.
Abstract:
Power control techniques for IEEE 802.11 wireless networks have already gained considerable attention. Such
techniques are particularly attractive because they can improve various aspects of wireless network operation
such as interference mitigation, spatial reuse in dense wireless deployments, topology control, and link quality
enhancement. In this paper we propose a novel delivery ratio based Conservative Transmit Power Control
(ConTPC) mechanism. Our implementation is conservative when it comes to deciding if the transmit power
should be reduced for a given link. This is because we do not want poor quality wireless links to further
reduce their quality and be overwhelmed by other links transmitting at maximum power. We have experi-
mentally evaluated the benefit of the proposed power control scheme when compared with fixed power level
systems. We show that our ConTPC mechanism can increase the throughput, however the magnitude of this
enhancement largely depends on the topology of the wireless network.
1 INTRODUCTION
Interference has been identified as a key cause of
performance degradation in Wireless Mesh Networks
(WMNs) (Jain et al., 2003; Padhye et al., 2005).
Transmit power control is one of the methods which
allows for interference mitigation and spatial reuse
through per-link power control.
Per-link power control allows a network node to
transmit packets to its neighbours at different power
levels. Data transfers to close neighbours may be
performed at low transmission powers, thereby min-
imising the interference with remote nodes. Con-
versely, the communication with remote neighbours
may be improved by using a higher transmission
power, i.e. by providing a stronger signal at the re-
ceiver. It has been shown by Muqattash et al (Muqat-
tash and Krunz, 2005) that a power controlled MAC
(POWMAC) protocol can significantly improve net-
work throughput. Furthermore, power control be-
comes a necessity in dense deployments which are
now becoming common as demonstrated by Akella
et al. (Akella et al., 2005).
However, increasing the transmission power for
weak links also has a negative effect of producing in-
creased interference. There are other factors which
can impact on the benefit of employing transmit
power control. For example, Broustis et al. (Broustis
et al., 2007) have observed that when power control is
combined with virtual carrier sensing (RTS/CTS mes-
sages) the performance is often degraded.
Transmit Power Control (TPC) mechanisms when
implemented in WMNs can be used to:
minimise interference with other nodes (and thus
increasing spatial reuse) as implemented in POW-
MAC (Muqattash and Krunz, 2005) or in (Navda
et al., 2007);
improve the quality of wireless links (as imple-
mented in (Son et al., 2004));
reduce the energy consumption (Gomez et al.,
2003);
control the network topology (Ramanathan and
Hain, 2000);
reduce interference with satellites and radar oper-
ating in the 5 GHz frequency band (as required by
the IEEE 802.11h Standard (802.11h IEEE Stan-
dard for Information technology - Telecommuni-
cations and information exchange between sys-
tems Local and metropolitan area networks -
Specific requirements, ));
5
Kowalik K., Bykowski M., Keegan B. and Davis M. (2008).
AN EVALUATION OF A CONSERVATIVE TRANSMIT POWER CONTROL MECHANISM ON AN INDOOR 802.11 WIRELESS MESH TESTBED.
In Proceedings of the International Conference on Wireless Information Networks and Systems, pages 5-14
DOI: 10.5220/0002022400050014
Copyright
c
SciTePress
ensure good coverage (as implemented by some
access point manufacturers).
Some of these objectives can be realised by mod-
ifying a single layer, such as the POWMAC (Muqat-
tash and Krunz, 2005). However, in general power
control needs to be aware of the operation at multiple
layers and is more accurately a cross-layer optimisa-
tion problem. Kawadia and Kumar present an com-
prehensive discussion about the principles of such a
cross layer design of power control in (Kawadia and
Kumar, 2005) .
The majority of the power control techniques pro-
posed require power control with per-packet level
granularity and with low latency. In (Kowalik et al.,
2008) we have shown that WLAN cards with Atheros
chipsets allow for such a fine power control.
In this paper we present a novel delivery ratio
based Conservative Transmit Power Control (Con-
TPC) mechanism. We evaluate it experimentally on
an eleven node static indoor 802.11 wireless mesh
testbed as well as on a four node topology which pro-
vides an easy to analyse example of a disjoint com-
munication case.
2 RELATED WORK
Reports in (Monks et al., 2001; Muqattash and Krunz,
2003) present 802.11 MAC modifications which aim
to utilise power control. However, these solutions re-
quire the use of a separate control channel for the
exchange of collision avoidance information. Such
schemes would require two wireless interfaces at each
mesh node. Thus this approach is not feasible for the
single radio solution considered in this work.
The POWMAC (Muqattash and Krunz, 2005) pro-
posed by Muqattash and Krunz is an enhancement
of the 802.11 MAC enabling power control and re-
sembles the ConTPC scheme proposed in this paper.
However, POWMAC requires detailed information
about the power of a receivedcontrol signal, as well as
the interference power. This may not be a problem if
one uses simulation tools to evaluate the approach as
the authors did. However, this is unrealistic because
we and other researchers have observed that SNR
measurements exhibit a high variability. For exam-
ple Akella et al. (Akella et al., 2005) have shown that
4 dB of variance in RSSI values and noise estimates
is typical. Moreover, under the POWMAC scheme
the transmission of each data packet is preceded by
an access window (AW) during which several pairs
of neighboring terminals exchange their RTS/CTS
control messages. Consequently this can generate a
significant overhead. There is a similarity between
POWMAC and LPERF (Akella et al., 2005) (Load-
sensitive, Power-controlled Estimated Rate Fallback)
because both combine the load information with sig-
nal strength measurements. However, it was stated
by Akella et al. (Akella et al., 2005) that in practice
they found that “achieving good performance and in-
terference reduction using the LPERF technique can
be challenging” owing to the poor accuracy of signal
strength measurements and the difficulties in estimat-
ing the load.
Broustis et al. (Broustis et al., 2007) performed
physical experiments and have analysed three scenar-
ios in which power control may improve network per-
formance:
“overlapping case” where power control can-
not improve the performance of two overlapping
links;
“hidden terminal case” where power control can
improve fairness;
“potentially disjoint case” where power control
can improve throughput significantly.
Their experiments show that power control may im-
prove overall throughput, however virtual carrier
sensing (RTS/CTS messages) needs to be disabled.
Also, their results demonstrate that the benefit from
using power control is topology dependent. In Sec-
tion 6.2 we also present similar findings.
Akella et al. (Akella et al., 2005) have pro-
posed a number of combined power and rate con-
trol mechanisms. The one which resulted in the
best performance was PERF (Power-controlled Esti-
mated Rate Fallback) (Akella et al., 2005) which im-
plemented conservative power control. PERF per-
forms its decisions regarding the rate and transmit
power based on the delivery rate of previous pack-
ets and SNR estimates obtained from the WLAN
adapter. Moreover, each packet is extended to in-
clude information about the transmit power, path
loss, noise estimate of the last packet, and deliv-
ery rate information. ConTPC adopts a similar
approach. However, PERF reduces the transmit
power until estimatedSNR = decisionThreshold +
powerMargin. The powerMargin is used to con-
trol aggressiveness of the algorithm. Our ConTPC
scheme is also conservative, but it takes a different
approach in trying to reduce power provided that it
does not result in an increased packet loss.
WINSYS 2008 - International Conference on Wireless Information Networks and Systems
6
3 CONSERVATIVE TRANSMIT
POWER CONTROL (CONTPC)
We believe that many of the proposed power control
schemes are too opportunistic. The rationale behind
these schemes is that by tolerating a degree of inter-
ference at each node, one can trade-off an increased
BER (Bit Error Rate) for multiple simultaneous trans-
missions. For example, POWMAC (Muqattash and
Krunz, 2005) follows this approach.
We have discovered that in real indoor deploy-
ments this approach may be overly optimistic and
can degrade the network performance. In our wire-
less eleven node static indoor testbed wireless links
often exhibit high loss rates. Moreover, their deliv-
ery rate vs TX power characteristics often resembles
that shown in Figure 1. This sort of characteristic
has the distinct feature that even a small reduction in
the transmit power from its maximum value can re-
sult in a significant drop in the delivery rate. More-
over, when multiple simultaneous transmissions oc-
cur, the noise floor increases. Thus for a link with
such a characteristic its delivery rate would drop even
further. Experimentally we have found that for links
with such a characteristic even small decreases in the
transmit power can significantly degrade their perfor-
mance.
0
0.2
0.4
0.6
0.8
1
0 2 4 6 8 10 12 14 16 18
Delivery Rate
TX Power [dBm]
Figure 1: Delivery ratio vs. TX power characteristics for a
low quality link.
In our ConTPC mechanism we do not allow
the power control mechanism to reduce the transmit
power for all the network links. Instead the Con-
TPC mechanism first checks if the delivery rate vs
TX power curve is flat around its maximum power
level. For example, the characteristic could look like
the one shown in Figure 2. On such a link any reduc-
tion in the TX power does not adversely impact on its
delivery rate. Therefore, reducing the transmit power
will typically not degrade the performance of the link,
but can potentially reduce the interference on nearby
links.
0
0.2
0.4
0.6
0.8
1
0 2 4 6 8 10 12 14 16 18
Delivery Rate
TX Power [dBm]
Selected power level = 13 dBm
Figure 2: Delivery ratio vs. TX power characteristics for a
high quality link.
However, even for high quality links we do not
advocate large reductions in the transmit power as
this may expose the node to the risk of being over-
whelmed by neighboring nodes transmitting at maxi-
mum power. Thus, our powercontrol scheme adopts a
quite conservative approach when it comes to reduc-
ing the transmit power on any given link. ConTPC
will only reduce the transmit power while it does not
adversely impact of the delivery rate. The exact de-
tails of our scheme are provided in the following sec-
tion.
4 CONTPC IMPLEMENTATION
DETAILS
Our ConTPC mechanism controls the output power
on a per-link basis. Therefore, each node periodi-
cally broadcasts probe frames at all allowed power
levels which includes information about the power
level used to sent a given frame. This information
allows each node to create delivery rate vs TX power
characteristics for all its incoming links. In our ex-
periments we have used custom probe messages, but
HELLO messages which are usually exchanged be-
tween routers could also serve this purpose.
Our conservative approach requires the delivery
rate vs TX power characteristic to be almost flat
AN EVALUATION OF A CONSERVATIVE TRANSMIT POWER CONTROL MECHANISM ON AN INDOOR 802.11
WIRELESS MESH TESTBED
7
around the maximum power level. In order to de-
scribe the details of ConTPC power selection algo-
rithm we denote:
P
max
to be the maximum TX power level (for our
WLAN adapters it was 18dBm);
P
selected
to be the power level selected by our Con-
TPC mechanism;
P
step
to be the granularity of our power mecha-
nism;
P
safety
to be the safety margin, which defines the
minimum range of the flat area around the maxi-
mum power;
thr to be the threshold value which defines the end
of the flat area around the maximum power;
dlr to be the delivery rate vs TX power character-
istic.
In Figure 3 we present a diagram showing the de-
tails of ConTPC. Essentially, it selects the power level
at the leftmost part of the flat region around the maxi-
mum TX power, provided that the size of this region is
larger than the safety margin. ConTPC uses three con-
figurable parameters: P
step
, P
safety
and thr. The P
step
together with the frequency of the power broadcasts
can be used to tune the overhead and the accuracy of
the estimation of the delivery rate characteristics. The
P
safety
and thr parameters are used to control the ag-
gressiveness of ConTPC. An evaluation of the relative
importance of these parameters will be given in Sec-
tions 6.5 and 6.6.
if dlr(P) < thr * dlr(P_max) ?
P = P_max
P = P - P_step
if P_max - P > P_safety ?
NO
YES
P_selected = P + P_step
P_selected = P_max
NO YES
Figure 3: ConTPC – power selection details.
The ConTPC mechanism allows a node to deter-
mine the transmit power levels for its incoming links.
Once this has been done, each node needs to inform
its neighbours of the transmit power level that they
need to use when communicating with the node. In
static wireless mesh networks the communication pat-
tern usually resembles a tree whose main root is lo-
cated at the gateway node. To reduce the overhead in-
volved in informing neighbours about the power lev-
els that they need to use, we propose to distribute this
information along this tree-like communication pat-
tern only. Consequently, power control is only per-
formed on links within this communication pattern.
5 EXPERIMENTAL SETUP
We have implemented ConTPC as a Click (Kohler
et al., 2000) module. This module allows us to broad-
cast custom made power probes, generate delivery
rate vs TX power characteristics and inform neigh-
bouring nodes about the power levels that they need
to use.
To perform the experimental tests we used eleven
WLAN nodes distributed across two floors of our re-
search institute building as shown in Figure 4. The
nodes were based around the Soekris net4521 plat-
form. The experiments were performed indoors, and
unfortunately we could not completely isolate the
tested network from external sources of interference.
All nodes were running Linux operating systems and
were operating in the monitor mode to allow for
packet injection by our Click module.
In order to enable the TPC feature of Mad-
Wifi drivers one needs to compile it with
COPTS+= -DATH_CAP_TPC=1
and pass the
tpc=1
parameter to
insmod ath_pci
.
Moreover, the transmission rate was fixed for all
the experiments at 11Mbps and the antenna diversity
was also switched off.
6 RESULTS
6.1 Delivery Ratio Vs SNR
In the ConTPC scheme each node measures the deliv-
ery ratio vs TX power characteristics of its incoming
wireless links. In Figure 5 we have presented these
characteristics for all links of our test bed. One might
think that instead of the delivery ratio it would be eas-
ier to use Signal–to–Noise Ratio (SNR), since it di-
rectly relates to the power level used for the transmis-
sion. Moreover, SNR value can be easily obtained
from the wireless card, for example Atheros chipsets
report RSSI value which can by simply converted to
SNR by subtracting 95dB.
However, we do not advocate the use of SNR be-
cause we have observed a poor accuracy on the RSSI
measurements especially for weak signals as illus-
trated in Figure 6. Akella et al. (Akella et al., 2005)
WINSYS 2008 - International Conference on Wireless Information Networks and Systems
8
(a) Floor 1 (b) Floor 2
Figure 4: Locations of wireless nodes.
have shown that that the variance in RSSI estimates is
high and is typically around 4dB. Moreover, in Fig-
ure 6 one may observe that the SNR vs TX power
characteristics are much flatter than the delivery rate
vs TX power curves.
0
0.2
0.4
0.6
0.8
1
0 2 4 6 8 10 12 14 16 18
Delivery Rate
TX Power [dBm]
A-D
A-O
A-Q
A-B
A-M
A-C
D-A
D-K
D-B
D-C
D-R
O-A
O-D
O-B
K-A
K-D
K-Q
K-B
K-C
K-R
Q-A
Q-D
Q-K
Q-B
Q-C
B-A
B-D
B-O
B-K
B-Q
B-M
B-C
M-A
M-B
C-A
C-D
C-K
C-Q
C-B
C-M
C-R
R-D
R-K
R-C
Figure 5: Delivery rate vs TX power.
In Figure 7 we show how the delivery rate is re-
lated to the RSSI values. Although, it can seen that
there is a correlation between the SNR and the deliv-
ery rate, the large variance associated with the SNR
measurements make it difficult to use. Consequently,
in ConTPC we have decided to use the delivery rate
instead of the SNR.
0
5
10
15
20
25
30
35
40
45
0 2 4 6 8 10 12 14 16 18
RSSI
TX Power [dBm]
A-D
A-O
A-Q
A-B
A-M
A-C
D-A
D-K
D-B
D-C
D-R
O-A
O-D
O-B
K-A
K-D
K-Q
K-B
K-C
K-R
Q-A
Q-D
Q-K
Q-B
Q-C
B-A
B-D
B-O
B-K
B-Q
B-M
B-C
M-A
M-B
C-A
C-D
C-K
C-Q
C-B
C-M
C-R
R-D
R-K
R-C
Figure 6: SNR vs TX power.
6.2 Disjoint Case
In order to validate the operation of the ConTPC
scheme, we first decided to investigate how it per-
forms in the disjoint case where the expected benefit
is the greatest according to Broustis et al. (Broustis
et al., 2007). In this case transmit power control can
allow for the simultaneous use of two links which re-
sults in a twofold increase in throughput compared to
the case where the default maximum transmit power
is used. In general, if we were to have n such node
pairs all interfering with each other under maximum
transmission power conditions, then we could expect
a n-fold increase in throughput through the use of
power control. However, such topologies based on
pairs of nodes located close to each other and traf-
AN EVALUATION OF A CONSERVATIVE TRANSMIT POWER CONTROL MECHANISM ON AN INDOOR 802.11
WIRELESS MESH TESTBED
9
0
0.2
0.4
0.6
0.8
1
0 10 20 30 40 50 60
Delivery rate
RSSI
A-D
D-A
O-A
O-D
K-A
Q-A
Q-D
B-A
B-D
M-A
C-A
C-D
R-D
Figure 7: Delivery rate vs TX power.
fic patterns based on communication between pairs of
nodes only are quite rare.
In our testbed shown in Figure 4 we have moved
node B towards A and D towards B, to create a sce-
nario with two disjoint pairs A–B and C–D. Then we
performed an experiment in which each node sent a
512kbps UDP flow to its neighbouring node. Thus we
had four flows in the network and the experiment du-
ration was 6 hours. During this period we have com-
pared the performance of ConTPC with fixed power
levels at 0dBm and at 18dBm. To ensure that each
of these power schemes is tested under similar con-
ditions, they were dealt with in a round robin fashion
over a duration of 1 min each.
As can be observed in Figure 8 the delivery rate
characteristics were excellent for all power levels be-
tween pairs of nodes. Even though the node pairs
were located on different floors the delivery rate on
link B–D was still good, and also on links A–D and
C–A. Therefore, these pairs were not completely dis-
joint, so contention on the link B–D was still high
even though the minimum power of 0dBm was used.
Thus we did not observe doubling of the throughput,
as one might expect.
In Figure 9 we show that ConTPC resulted in an
increase of 35% when compared with transmission at
the maximum power of 18dBm. What is also interest-
ing here is that ConTPC slightly outperformed trans-
mission at maximum power of 0dBm. This is because
it emerged that for links B–A and C–D it was more
benefitial to use a transmit power of around2 or 3dBm
instead of the minimum value.
The nodes in our test bed are usually not placed in
pairs so close to each other, as presented in the above
experiment. Therefore, for the next step in the inves-
tigation we moved the nodes back to their normal po-
sitions and repeated the tests. Thus, it was no longer
0
0.2
0.4
0.6
0.8
1
0 2 4 6 8 10 12 14 16 18
Delivery Rate
TX Power [dBm]
A-D
A-B
A-C
D-A
D-B
D-C
B-A
B-D
B-C
C-A
C-D
C-B
Figure 8: Delivery rate characteristics observed in disjoint
case.
0
20
40
60
80
100
120
140
160
180
A-B B-A C-D D-C
Throughput [kbps]
Nodes pair
ConTPC
18dBm
0dBm
avg ConTPC
avg 18dBm
avg 0dBm
Figure 9: Throughput observed in disjoint case.
disjoint case. As it can be seen in Figure 10 only link
B–A continued to exhibit an excellent delivery rate
characteristic. Thus, it achieved the highest through-
put when a fixed power of 0dBm was used (as it is
shown in Figure 11).
In Figure 11 it may be observed that in this more
realistic 4 nodes case our ConTPC also outperforms
fixed power levels at 18dBm and at 0dBm. We also
demonstrate that in such a case where nodes are dis-
tributed evenly around the area (usually to maximise
coverage) the use of maximum power is close to op-
timal. Thus if the nodes were to be spread more uni-
formly to maximise coverage then we would observe
that ConTPC would provide the same overallthrough-
put as when the maximum power of 18dBm is used.
This experiment demonstrates that the benefit of
using transmit power control is topology dependent.
It also shows that our ConTPC outperforms fixed
WINSYS 2008 - International Conference on Wireless Information Networks and Systems
10
0
0.2
0.4
0.6
0.8
1
0 2 4 6 8 10 12 14 16 18
Delivery Rate
TX Power [dBm]
A-B
D-C
B-A
C-A
C-D
C-B
Figure 10: Delivery rate characteristics observed in 4 nodes
case.
0
50
100
150
200
A-B B-A C-D D-C
Throughput [kbps]
Nodes pair
ConTPC
18dBm
0dBm
avg ConTPC
avg 18dBm
avg 0dBm
Figure 11: Throughput observed in 4 nodes case.
power levels if there is some nonuniformity in the
node distribution. Thus it can reduce the transmit
power for strong links, thereby reducing the interfer-
ence.
6.3 Eleven Nodes Testbed Case
To evaluate ConTPC over our eleven node test bed we
have used a similar procedure. However, this time the
communication pattern was not performed in disjoint
pairs. Instead each node was sending a 512kbps UDP
flow to the node which was the next hop towards the
gateway node. Such communication pattern is similar
to the typical case when nodes want to connect to the
Internet through a gateway node. We have decided not
to forward the data packets beyond the gateway node
as we wished to observe the effect of transmit power
control on individual links instead of observing the
effect of hop penalty. The duration of this experiment
was 24 hours and all the power schemes were tested
in a round robin fashion.
In Figure 12 we have compared the throughput
achieved by the individual links on our eleven node
testbed. It can be observed that the ConTPC mech-
anism resulted in approximately a 50% increase in
throughputwhen compared with a fixed power levelat
0dBm and a 15% increase in throughput when com-
pared with a maximum power of 18dBm. One can
0
50
100
150
200
250
A-BA-DA-OB-AB-DC-AC-BC-DD-CK-CK-DN-JN-RO-AO-BQ-KR-CR-N
Throughput [kbps]
Nodes pair
ConTPC
18dBm
0dBm
avg ConTPC
avg 18dBm
avg 0dBm
Figure 12: Throughput observed on 11 node testbed.
also observe in Figure 12 that ConTPC offered an im-
provement mostly for links which achieved low or
average throughput when the maximum power was
used. This is because when the transmission was oc-
curring at maximum power, these links were suffering
due to increased contention and noise level. Reduc-
ing the transmit power of high quality links resulted
in improved performance of weak links. So ConTPC
has not only increased the average throughput, but has
also increased the fairness.
6.4 Stability of ConTPC
In Figure 13 we present the probability distribution
functions (PDFs) of the power level selection per-
formed by ConTPC during the experiment performed
in Section 6.3. This diagram allows one to determine
how often ConTPC changes the power levels or if
the power level remains fixed for extended periods of
time. As can be observed on link A–B, ConTPC se-
lected a wide range of power levels. This is because
AN EVALUATION OF A CONSERVATIVE TRANSMIT POWER CONTROL MECHANISM ON AN INDOOR 802.11
WIRELESS MESH TESTBED
11
0
0.2
0.4
0.6
0.8
1
0 5 10 15 20 25 30 35 40
Probability
Power [dBm]
A-B
A-D
A-O
B-A
B-D
C-A
C-B
C-D
D-C
K-C
K-D
N-J
N-R
O-A
O-B
Q-K
R-C
R-N
Figure 13: Power level selection PDF for links of 11 node
testbed.
node A was located in a corridor where there was a
high variability in the propagation conditions due to
the movement of people. However, most of the time
this link exhibited a high delivery rate, thus the se-
lected power tended to switch between 8 and 9dBm.
For the majority of the links, the transmit power
level was selected around a small range of values or
remained at one value for the duration of the experi-
ment.
6.5 Overhead of ConTPC
ConTPC obtains the delivery rate vs TX power char-
acteristics by broadcasting power probes at all speci-
fied power levels. In this section we consider the di-
rect overhead which is the amount of time consumed
by the transmission of these broadcast probe packets.
The overhead can be controlled by the frequency
of these broadcasts. In our experiment we have mea-
sured the overhead on node D which was the gateway
node and featured the highest number of neighbours,
and thus was influenced by the highest overhead. In
Figure 14 we have presented the overhead as a per-
centage for three different intervals between consecu-
tive broadcasts, i.e. every 50ms, 100ms, and 200ms.
As expected reducing the frequency of these broad-
casts resulted in a drop in the overhead from 2.5% to
1.3%, and finally to 0.6%. Although the 200ms in-
terval resulted in the smallest overhead we have ob-
served that ConTPC has significantly slowed the pro-
cess of generation of the delivery rate vs TX power
characteristics. Therefore, if one would wished to
keep the overhead low at around 0.6%, we would rec-
ommend to reduce the number of power levels being
tested. In (Kowalik et al., 2008) we have shown that
WLAN cards with Atheros chipsets allow for the con-
0
0.5
1
1.5
2
2.5
3
0 50 100 150 200 250 300 350
Overhead
Time [sec]
50ms
100ms
200ms
Figure 14: ConTPC overhead.
trol of the transmit power with 0.5dBm granularity.
Not all the power levels need to be used by the power
control mechanism, especially if we wish to limit the
probing overhead. Thus, for a 200ms broadcast inter-
val we have increased the value of P
step
form 0.5dBm
to 1dBm. The increase of P
step
with the broadcast in-
terval set at 200ms allowed ConTPC to continue to
outperform tansmissions at maximum power for our
eleven node test bed.
6.6 ConTPC Parameters Evaluation
ConTPC uses two configurable parameters P
safety
and
thr to control its aggressiveness. The P
safety
is a
safety margin, which defines the minimum range of
the flat region around the maximum power. The thr
is the threshold value which defines if the region is
still flat around the the maximum power. Thus, if
one would set P
safety
to be 0dBm, and the thr to
be equal to 0.9, then all network links would use a
transmit power which results in the delivery rate of
0.9 dlr(18dBm). These settings would effectively
reduce the transmit power on all the network links.
In Figure 15 we have provide evidence that the
opportunistic approach where (P
safety
= 0dBm) re-
sults in the worst performance. The conservative ap-
proach ( where P
safety
>= 1dBm) is much better. It
can be observed that the highest average throughput
is achieved if we select a low threshold thr = 0.8 and
use a wide power safety margin as P
safety
= 4dBm.
Setting a too high a threshold of around thr = 0.95
for a link with a high variance in the delivery rate vs
power curve could cause ConTPC to decide that the
WINSYS 2008 - International Conference on Wireless Information Networks and Systems
12
0
20
40
60
80
100
120
140
160
0 1 2 3 4 5
Avg. Throughput [kbps]
TX Power safety margin [dBm]
thr 0.8
thr 0.85
thr 0.90
thr 0.95
Figure 15: Evaluation of importance of P
safety
and thr pa-
rameters.
flat area is quite narrow. This could effectivelydisable
ConTPC on links with high variance of the delivery
rate vs power characteristics. However, threshold val-
ues of around 0.8 ensure that ConTPC correctly finds
the boundary of the flat area despite some variability.
While a large P
safety
(above 3dBm) ensures that only
good quality links reduce their power. Consequently,
by being conservative this allows for the efficient op-
eration of ConTPC.
7 CONCLUSIONS
In this paper we have presented the ConTPC mecha-
nism which is a conservative transmit power control
scheme. It does not seek to reduce the transmit power
for all the networks links because such an approach
usually penalises low quality links. Instead, it only
reduces the transmit power for links on which such a
reduction does not affect the delivery rate.
We have demonstrated that the performance of
ConTPC is topology dependent. The highest through-
put gain can be observed in nonuniform topologies
such as the disjoint node pairs case, when multiple
transmissions may occur simultaneously. However, in
a typical indoor deployment, as for example in case
of our test bed, such a scheme continues to outper-
form the fixed power levels case, resulting in a 15%
throughput increase when compared with fixed power
operating at its maximum value.
We have also shown that ConTPC not only in-
creases the overall throughput but also increases the
fairness by reducing contention. Moreover, it is pos-
sible to control the overhead of ConTPC while main-
taining its high throughput gain.
ACKNOWLEDGEMENTS
The authors acknowledge the support of Enterprise
Ireland (under the Commercialisation Fund 2007)
and Science Foundation Ireland (under the grant
03/IN3/I396).
REFERENCES
802.11h IEEE Standard for Information technology -
Telecommunications and information exchange be-
tween systems Local and metropolitan area net-
works - Specific requirements. Part 11: Wireless LAN
Medium Access Control (MAC) and Physical Layer
(PHY) specifications, Amendment 5: Spectrum and
Transmit Power Management Extensions in the 5 GHz
band in Europe.
Akella, A., Judd, G., Seshan, S., and Steenkiste, P. (2005).
Self-management in chaotic wireless deployments. In
MobiCom ’05: Proceedings of the 11th annual in-
ternational conference on Mobile computing and net-
working, pages 185–199.
Broustis, I., Eriksson, J., Krishnamurthy, S. V., and Falout-
sos, M. (2007). Implications of power control in wire-
less networks: A quantitative study. In Proceedings of
Passive and Active Measurement Conference (PAM).
Gomez, J., Campbell, A., Naghshineh, M., and Bis-
dikian, C. (2003). PARO: Supporting Dynamic Power
Controlled Routing in Wireless Ad Hoc Networks.
ACM/Kluwer Journal on Wireless Networks (WINET),
9(5):443–460.
Jain, K., Padhye, J., Padmanabhan, V. N., , and Qiu, L.
(2003). Impact of interference on multi-hop wireless
network performance. ACM MobiCom.
Kawadia, V. and Kumar, P. (2005). Principles and proto-
cols for power control in wireless ad hoc networks.
IEEE Journal on Selected Areas in Communications,
23(1):76–88.
Kohler, E., Morris, R., Chen, B., Jannotti, J., and Kaashoek,
M. F. (2000). The Click modular router. ACM Trans-
actions on Computer Systems, 4(18).
Kowalik, K., Bykowski, M., Keegan, B., and Davis,
M. (2008). Practical Issues of Power Control in
IEEE 802.11 Wireless Devices. In Proceedings of
15th International Conference on Telecommunica-
tions ICT’08, St. Petersburg, Russia.
Monks, J., Bharghavan, V., and Hwu, W.-M. (2001). A
power controlled multiple access protocol for wireless
packet networks. INFOCOM 2001. Twentieth Annual
Joint Conference of the IEEE Computer and Commu-
nications Societies, 1(4):219 – 228.
AN EVALUATION OF A CONSERVATIVE TRANSMIT POWER CONTROL MECHANISM ON AN INDOOR 802.11
WIRELESS MESH TESTBED
13
Muqattash, A. and Krunz, M. (2003). Power controlled dual
channel (PCDC) medium access protocol for wireless
ad hoc networks. INFOCOM 2003, pages 470–480.
Muqattash, A. and Krunz, M. (2005). POWMAC: A Single-
Channel Power-Control Protocol for Throughput En-
hancement in Wireless Ad Hoc Networks. IEEE Jour-
nal on Selected Areas in Communications (JSAC) -
Special Issue on Advances in Military Wireless Com-
munications, 23(5):1067–1084.
Navda, V., Kokku, R., Ganguly, S., and Das, S. (2007).
Per-client Power Management for Spatial Reuse in
WLANs. Technical Report, Computer Science De-
partment, Stony Brook University, NY.
Padhye, J., Agarwal, S., Padmanabhan, V., Qiu, L., Rao,
A., and Zill, B. (2005). Estimation of Link Interfer-
ence in Static Multi-hop Wireless Networks. Internet
Measurements Conference.
Ramanathan, R. and Hain, R. (2000). Topology control of
multihop wireless networks using transmit power ad-
justment. In INFOCOM, pages 404–413.
Son, D., Krishnamachari, B., and Heidemann, J. (2004).
Experimental study of the effects of transmission
power control and blacklisting in wireless sensor net-
works. In Proceedings of the First IEEE Conference
on Sensor and Adhoc Communication and Networks,
pages 289–298.
WINSYS 2008 - International Conference on Wireless Information Networks and Systems
14