Experiences with a Metropolitan Multiradio Wireless Mesh Network: Design, Performance, and Application

Abstract

Wireless mesh networks comprise nodes with multiple radio interfaces, and can provide low-cost high-speed Internet access or connectivity for data transfer. In this article we report our experiences and investigations with an experimental metropolitan multiradio mesh network that covers an area of approximately 60 km2 in the city of Heraklion, Crete. We present the design and deployment of the network, experiments to quantify the network's performance, and an application that runs on top of it and exploits it's low-cost wide-area connectivity. The metropolitan network consists of 16 nodes, among which six are core nodes, each with up to four 802.11a wireless interfaces and an additional wireless interface for management and monitoring. The distance between core mesh nodes varies from 1.6 to 5 km, and the mesh network contains two gateways that connect it to a wired network. Our performance experiments involve rate, power, and channel control for long-distance metropolitan links, and include investigations of the timescales for the operation for these mechanisms. Finally, we present a system for continuous online electromagnetic field monitoring and spectrum sensing, which utilizes the metropolitan mesh network for collecting wide-area measurements from low-cost EMF measurement devices.

Full text

1
Experiences with a Metropolitan Multi-Radio
Wireless Mesh Network: Design, Performance,
and Application
Vasilios A. Siris, Elias Tragos, and Nikolaos Petroulakis
Abstract—Wireless mesh networks (WMNs) are comprised of nodes with multiple radio interfaces, and can provide low-cost high-
speed Internet access or connectivity for data transfer. In this paper we report our experiences and investigations with an experimental
metropolitan multi-radio mesh network that covers an area of approximately 60 Km2in the city of Heraklion, Crete. We present the
design and deployment of the network, experiments to quantify the network’s performance, and an application that runs on top of it and
exploits it’s low-cost wide-area connectivity. The metropolitan network consists of 14 nodes, among which six are core nodes, each with
up to four 802.11a wireless interfaces and an additional wireless interface for management and monitoring. The distance between core
mesh nodes varies from 1.6 to 5 Km, and the mesh network contains two gateways that connect it to a wired network. Our performance
experiments involve rate, power, and channel control for long-distance metropolitan links, and include investigations of the timescales
for the operation for these mechanisms. Finally, we present a system for continuous online Electromagnetic Field (EMF) monitoring,
which utilizes the metropolitan mesh network for collecting wide-area measurements from low-cost EMF measurement devices.
Index Terms—metropolitan mesh networks, interference, channel assignment, power-rate control, EMF measurements.
F
1 INTRODUCTION
WIRELESS multi-radio multi-channel mesh networks
have the potential to provide ubiquitous and
high-speed broadband access in urban and rural areas,
to both fixed and mobile users, with low operation and
management costs. To investigate issues related to the
management and performance of a multi-radio mesh
network in an actual metropolitan environment, we have
deployed an experimental multi-radio mesh network
that covers an area of approximately 60 Km2in the city
of Heraklion, Crete, Greece. Our objective is to use the
network as a metropolitan scale test-bed to
•investigate the performance of a multi-radio mesh
network, built from commodity components and
containing 1 to 5 Km links with directional antennas,
•evaluate channel assignment procedures for efficient
wireless spectrum utilization,
•investigate MAC/network layer mechanisms and
routing metrics for supporting performance guaran-
tees in multi-radio, multi-channel, multi-rate mesh
networks, and
•investigate innovative applications that require per-
vasive, wide-area, and high-speed data transfer.
The metropolitan test-bed is built from commodity
IEEE 802.11 components, which leads to significantly
lower costs compared to other technologies, such as
802.16. Other mesh and/or long-distance 802.11 net-
works include the 802.11b-based Digital Gangetic Plains
The authors are with the Institute of Computer Science of the Foundation for
Research and Technology Hellas (ICS-FORTH), P.O. Box 1385, Heraklion
71110, Crete, Greece. E-mail: {vsiris, etragos, npetro}@ics.forth.gr
V. A. Siris is also with the Department of Informatics, Athens University of
Economics and Business, Greece.
rural area test-bed with 1 - 23 Km links [1], the WiLDNet
network with 50-100 Km links [2], the Roofnet network
which considers single-radio mesh nodes [3], and the
Quail Ridge wireless mesh network [4] containing 34
mesh nodes, with the distance of most links being
smaller than 1 Km. Rice University has deployed a
metropolitan mesh network in East-End Houston with
21 mesh nodes in an area of of approximately 3 Km2[5].
Another mesh network deployed by the Waikato Uni-
versity in New Zealand has 17 links with distances from
300 m to 17 Km1.
An important objective in the design of a multi-radio
wireless mesh network is the efficient utilization of the
limited wireless resources and the radio network infras-
tructure, by controlling key parameters of a wireless
communications system, such as the assigned channels,
and the transmission rate and power. Interference is a
key factor that can lead to reduced capacity and per-
formance of wireless networks operating in unlicensed
spectrum bands. Indeed, interference can exist between
links belonging to the same network, or can originate
from external sources employing the same of different
wireless technologies.
One method for reducing the interference is to appro-
priately select the channels of wireless interfaces, which
also affects the connectivity of wireless mesh networks
since two interfaces with omnidirectional antennas that
are within the transmission range of each other can com-
municate only if they operate on the same channel. The
objective of rate control is to adapt the transmission rate
to the channel characteristics in order to improve perfor-
1. http://www.crc.net.nz/crcnet.php

2
mance, in terms of throughput and packet transmission
delay. Throughput depends on both transmission rate
and packet loss ratio, and the maximum throughput
is not necessarily achieved by the maximum transmis-
sion rate or by the transmission rate with the lowest
packet loss ratio. The power control problem in wireless
networks is that of selecting the transmission power at
each radio interface in the network, in order to balance
consumption and performance. We have investigated the
operation and performance of the above mechanisms in
the metropolitan test-bed, in addition to the time-scales
in which they should operate. A key question we address
is whether the aforementioned mechanisms should be
performed in a small time-scale (in the order of packets
or hours) or in a much larger time-scale (in the order of
days or weeks).
In this paper, we also present an innovative system
that utilizes the low-cost wide-area connectivity pro-
vided by the metropolitan wireless mesh network for
the collection, processing, and presentation of electro-
magnetic field (EMF) measurements. The rapid growth
of wireless technology has brought to the forefront of
public interest and concern the issue of increasing EMF
radiation, especially from mobile telephony systems;
thus, monitoring of EMF radiation is becoming increas-
ingly important, as is the necessity to verify conformance
to national and international thresholds. Even when the
assigned legal limits are not exceeded, the evolution of
EMF levels can provide important information, which
can be later used by the scientific community or public
bodies and regulators.
The rest of the paper is organized as follows: Sec-
tion 2 describes the design and the deployment of
the metropolitan mesh network, and the interference
and performance monitoring. Section 3 discusses exper-
iments on channel assignment, rate and power adapta-
tion, and the investigation of the time-scales of operation
for these mechanisms. Section 5 describes a system for
real-time collection of electromagnetic field (EMF) mea-
surements that utilizes the low-cost wide-area connectiv-
ity provided by the metropolitan mesh network. Finally,
Section 6 concludes the paper.
2 METROPOLITAN MULTI-RADIO MESH
NETWORK DESIGN
2.1 Topology
The metropolitan mesh network covers an area of ap-
proximately 60 Km2and currently contains 14 nodes,
Figure 1, among which six are core mesh nodes. The
distance and antennas used for the links between core
mesh nodes are shown in Table 1. Each wireless interface
is assigned a static IP address, and the OLSR protocol
is used for routing traffic in the network. The mesh
network is connected to a fixed network via two nodes
(M1 - FORTH and K4 - University of Crete/UoC).
Fig. 1. Topology of the metropolitan mesh network.
TABLE 1
Links between core mesh nodes. *This is a
point-to-multi-point link. The parenthesis contain antenna
configurations used in some of the experiments.
Link Distance (Km) Antennas
K1 - K2 5.1 36 dBi - 36 dBi dish
(29 dBi grid-21 dBi panel)
K1 - K3 4.9 26 dBi - 21 dBi panel
(29 dBi grid-21 dBi panel)
K2 - K3 2.0 21 dBi - 19 dBi panel
K4 - K2 1.6 36 dBi - 36 dBi dish
(21 dBi-21 dBi panel)
K4 - K5 3.3 26 dBi - 19 dBi panel
K4 - K6 2.8 19 dBi - 19 dBi panel
K5 - K2∗2.0 26 dBi - 19 dBi panel
K5 - K6∗0.4 26 dBi - 19 dBi panel
K6 - K3 0.8 19 dBi - 19 dBi panel
2.2 Multi-radio mesh node
Each multi-radio mesh node consists of a mini-ITX board
(EPIA SP 13000, 1.3 GHz C3 CPU, 512 MB DDR400
memory) and a 40 GB 2.5 HDD. A four slot mini PCI
to PCI adapter (MikroTik RouterBOARD 14) holds four
802.11a/g mini PCI adapters (NL-5354 MP PLUS Aries
2, Atheros based High Power Super A/G dual Band
802.11a/b/g). The mini-ITX runs Gentoo 2006 i686 Linux
(2.6.18 kernel) with the MadWiFi driver version 0.9.2.
Finally, the nodes run OLSR daemon version 0.4.10 (by
olsr.org), which implements the Optimized Link State
Routing (OLSR) protocol.
One of our design requirements was to allow remote
management, monitoring, and recovery of the mesh
nodes, even in situations when a node’s mini-ITX board
crashes or its wireless interfaces are down. To address
this requirement we added to each mesh node an ad-
ditional 802.11a client, which connects to a manage-
ment and monitoring network that operates in parallel
to the experimental mesh network. The independent
management and monitoring network has proved to
be indispensable, allowing us to continuously monitor

3
the performance of the network, and conduct remote
experiments with channel, power, and rate control mech-
anisms without loosing connectivity to the mesh nodes.
Additionally, each mesh node contains an intelligent
remote power switch (Dataprobe iBoot), which supports
on/off power switching through a web interface, and
timed power reboots based on the results when the
power switch pings other devices, such as the mini-
ITX board or some remote device to verify the wireless
connectivity. Together with the independent manage-
ment and monitoring network, the remote power switch
helped to provide continuous remote connectivity to all
mesh nodes, enabling remote recover from crashes and
deadlocks, which are very common when experimenting
with wireless devices.
2.3 Interference and performance monitoring
2.3.1 Interference measurements
Most prior measurement studies of outdoor 802.11 links
focused on measuring the path loss and the time corre-
lation of losses, and how loss is affected by factors such
as received signal strength, link distance, interference,
weather conditions, and technology (802.11b/g), e.g. [1],
[6]. Work on the impact of adjacent channel interference
for 802.11b/g is contained in [1], [7] and for 802.11a
in [8]; the latter focuses on measuring the impact on
the signal-to-noise ratio, and considers link distances
of 60 meters. On the other hand, the results presented
in this section consider metropolitan 802.11a links with
longer distances (1.6 - 5 Km), and focuses on the impact
that interference has on the throughput above the MAC
layer.
Next we investigate the interference between
metropolitan distance links, when one of the two
interfaces of each link under investigation is located in
the same mesh node (node K2 in our case). In particular,
we consider the link pair K2 - K3 and K2 - K4, and the
link pair K2 - K3 and K2 - K1. For the first pair, the two
interfaces in node K2 are connected to two 21 dBi panel
antennas (parenthesis in Table 1), which are both on the
same mast with a distance of approximately 0.75 meters,
and have a relative angle of approximately 150 degrees.
For the second pair, the two interfaces in node K2 are
again connected to two 21 dBi panel antennas, which
are however on a different mast with a distance of
approximately 2.5 meters, and have a relative angle of
approximately 90 degrees. Each experiment we present
below shows the average from 10 runs, each run lasting
for 100 seconds. For all results, the confidence interval
is less than 6%. Finally, the experiments involved UDP
traffic with rate 3 Mbps, generated using the iperf tool.
Tx/Rx in the same node: We first consider the in-
terference between links when a receive and transmit
interface exists in the same mesh node. Two 3 Mbps
UDP streams are transmitted over the links K2 →K3
(2 Km distance) and K4 →K2 (1.6 Km distance). Note
that the two streams are independent, and the iperf
sender for the first stream is located in a workstation
connected to our internal laboratory network. Table
2 - Experiment 1 shows the throughput achieved by
each UDP flow, for three different channel assignments.
Observe that when both links are assigned the same
channel (Channel distance=0, channels 36 - 36 in Table
2), the transmitter significantly affects the receiver (see
throughput for Rx K2 →K4), as they are both located
in the same mesh node (K2). When the two links are
assigned neighboring channels (40-36), the interference
is significantly reduced, but still appears to exist. On
the other hand, when there is a one channel separation
(44-36), there is no interference and the throughput is
essentially equal to the UDP sending rate.
We performed the same experiment as the one de-
scribed above, but with two 3 Mbps UDP stream trans-
mitted over the links K2 →K1 (5.1 Km distance) and K3
→K2 (2 Km distance). Table 2 - Experiment 2 shows
that the interference in this case is lower than in the
previous experiment. This is due to the larger distance
(approximately 2.5 meters) between the antennas for
node K2 that correspond to the above two links.
Rx/Rx in the same node: Next we investigate the
interference between links when two receive interfaces
are located in the same mesh node. Two 3 Mbps UDP
streams are transmitted over links K3 →K2 and K4 →
K2. In this experiment the achieved throughput for both
flows is identical, and equal to the UDP sending rate (as
shown in Table 2 - Experiment 3), even when the links
are assigned the same channel. Hence, when the receive
interfaces are located in the same node, the interference
between the two links is not significant.
The above experiments show that, depending on the
distance between antennas, there can be significant inter-
ference between metropolitan links, when the transmit-
ting and receiving ends of the two links are located in the
same node, and even when the two links are assigned
different but adjacent 802.11a channels. The interference
can be avoided if the links are assigned channels with a
one channel separation, or by placing antennas at some
distance, typically more than 2 meters.
2.3.2 Continuous online performance monitoring
Continuous monitoring of core mesh network links al-
lows quick detection and identification of anomalous
link behavior. Indeed, as we will see in the following
sections, for metropolitan links to achieve high per-
formance under normal operation, it is sufficient to
fix the transmission rate, power, and channel; the per-
formance should be continuously monitored, and only
when anomalies are observed the rate, power, channel
needs to be adapted. For this reason, we have developed
a set of perl and shell scripts that continuously monitor
important performance metrics for all links between
core nodes. The metrics include the signal-to-noise ratio
(SNR), transmission rate, MAC and physical layer errors,
two-way delay, and throughput. The scripts are executed
every five minutes, except the scripts for measuring the

4
TABLE 2
Throughput (Mbps) measurements of two flows in 3
different experiments.
Experiment 1 - receiver and transmitter is in same mesh node
(K2) and antennas on same mast at distance ≈0.75 meters
Channel distance TX K2→K3 RX K2←K4
0 (36-36) 2.970 2.358
1 (40-36) 2.995 2.976
2 (44-36) 2.997 2.997
Experiment 2 - receiver and transmitter is in same mesh node
(K2) and on different mast at distance ≈2.5 meters
Channel distance TX K2→K1 RX K2←K3
0 (36-36) 3 2.75
1 (40-36) 3 3
2 (44-36) 3 3
Experiment 3 - two receivers are in same mesh node (K2),
and antennas on same mast at distance ≈0.75 meters
Channel distance RX K2←K4 RX K2←K4
0 (36-36) 2.996 2.996
1 (40-36) 2.996 3
2 (44-36) 3 3
throughput, which are executed every 30 minutes. The
collected data is stored in an Round Robin Database
(RRD), and the corresponding daily and weekly graphs
are made available through an http server2using the
RRD Tool.
3 PE RFORM ANCE OF R ATE,POWER,AND
CH ANNEL C ONTROL OVE R METROPO LITAN
LINKS
Next we investigate the performance of rate, power,
and channel control in the metropolitan mesh network,
including the time-scales for the operation of these mech-
anisms.
3.1 Rate control
We compare the throughput that is achieved using a
fixed transmission rate scheme, with the throughput
achieved with MadWifi’s SampleRate algorithm, which
we will refer to as auto-rate scheme, that adjusts the
transmission rate on a per packet basis. The traffic was
generated using the iperf tool. The graphs in this section
report values averaged over intervals of two minutes.
Figure 2a shows the throughput achieved on a spe-
cific link (K2-K4), with the auto-rate scheme (horizontal
straight lines) and fixed transmission rate for different
transmission powers. Observe that a higher throughput
is achieved by using a fixed rate scheme, if the fixed
transmission rate is appropriately selected. Moreover,
the figure shows that a higher improvement is achieved
in the case of lower SNR values: when the transmission
power is 15 dBm the improvement is approximately 30%,
whereas when the transmission power is 12 dBm the
improvement increases to approximately 42%.
2. Online access to some of the measurements is available at
http://www.ics.forth.gr/HMESH/
(a) Throughput for auto-rate and fixed transmission rate.
Link K2-K4, distance 1.6 Km.
(b) Throughput for auto-rate for different periods of the day.
Link K2-K4, distance 1.6 Km.
Fig. 2. Rate control measurements.
Figure 2b shows the throughput for auto-rate and
fixed rate, at different hours of the day. Observe that im-
provements achieved by fixed rate remain relatively the
same during the course of a day. The conclusions drawn
from the above experimental results are the following:
•For high quality (high SNR) links, fixing the rate
to the highest transmission rate achieves similar
throughput as SampleRate.
•For low quality (low SNR) links, an appropriately
selected fixed transmission rate can achieve signifi-
cantly higher throughput compared to SampleRate.
•Finally, a fixed transmission rate scheme shows sim-
ilar performance over long time-scales (day). Hence,
for metropolitan link with directional antennas, the
transmission rate does not necessarily need to be
adapted over smaller time-scales.
The above suggests that the adaptation of the transmis-
sion rate in small time-scales (e.g., of the order of packet
arrivals), which is typically followed by all auto-rate
algorithms, is not necessary, and can even reduce per-
formance for long distance metropolitan wireless links.
3.2 Joint power and rate control
Next we investigate power control in metropolitan wire-
less links. The advantage of reducing the transmission
power is the consequent reduction of the interference
that is produced. Our objective is to identify the min-
imum transmission power that achieves close to maxi-
mum performance, and investigate how this minimum

5
(a) Throughput for different transmission powers.
Link K2-K3, distance 2 Km.
(b) Transmission power for achieving throughput above 94 % of the
maximum. Link K2-K3, distance 2 Km.
Fig. 3. Joint power and rate control measurements.
transmission power varies with time. Figure 3a shows
the throughput for different transmission powers for
both auto-rate and fixed transmission rates. The key ob-
servation from this figure is that the same high through-
put is achieved with transmission power higher than
8 dBm with the auto-rate scheme and a 36 Mbps fixed
transmission rate, and higher than 11 dBm for a 48 Mbps
fixed transmission rate. These transmission power values
are significantly lower than the maximum transmission
power of 15 dBm and they achieve the same throughput
as seen in the figure. Figure 3b shows, for different
times-of-day, the transmission power required to achieve
throughput above the 94% of the maximum throughput.
This figure shows that the necessary minimum transmis-
sion power to achieve high throughput does not change
significantly throughout the day. The conclusions drawn
from the power control experiments are the following:
•The transmission power can be significantly re-
duced, without a large impact on the achieved
throughput.
•The minimum transmission power to achieve
some minimum performance does not significantly
change throughout the course of a day, hence adjust-
ment of the transmission power can occur on longer
time-scales.
3.3 Channel assignment
Next we present experiments related to channel assign-
ment in a metropolitan mesh network with directional
antennas. A key requirement for channel assignment,
when links operate in an unlicensed band, is to account
for both internal (or intra-network) interference between
links belonging to the mesh network, and external in-
terference from sources external to the network. In this
section we present results of an approach that captures
both types of interference, and show the influence on the
overall performance when internal or external interfer-
ence is not taken into account.
One approach for capturing intra-network interfer-
ence is the Multi-Point Link Conflict Graph (MPLCG)
presented in [9]. A vertex in the MPLCG represents
a multi-point communication link, which is a set of
interfaces that communicate with each other; all inter-
faces belonging to the same multi-point link should
be assigned the same channel. An edge between two
vertices in the MPLCG indicates that the two corre-
sponding links interfere, hence cannot be assigned the
same or neighbouring channels; the latter is enforced
because there can exist interference between adjacent
channels, even in the case of IEEE 802.11a. In addition to
capturing interference, another important component of
channel assignment is the actual selection of channels.
Possible metrics for selecting a channel are the one-
way SNR, two-way SNR (which is the average SNR
on the two interfaces belonging to the same link), or
round-trip delay. These metrics can be measured online,
and can capture the level of interference from external
sources, both 802.11 and non-802.11; other approaches
to channel assignment capture only interference between
internal links, or external interference solely from 802.11
sources. The two SNR metrics capture adjacent channel
interference, but do not capture MAC-layer contention
between interfaces assigned to the same channel. On
the other hand, the round-trip delay metric can capture
interference due to both adjacent and co-channel inter-
ference, since it is influenced by MAC layer contention.
Figure 4a compares the average packet delay achieved
with a channel assignment procedure using the Multi-
Point Link Conflict Graph for capturing intra-network
interference and the packet delay metric for channel se-
lection, with the Measurement-based Directional Chan-
nel Assignment (M-DCA) scheme presented in [10]. The
M-DCA scheme relies on an antenna overhearing trans-
missions from other antennas inside its neighborhood, to
identify which channels are already used by its neigh-
bors, and avoid selecting them. Figure 4a shows that
the M-DCA scheme achieves an average delay which is
approximately 19% higher than average delay achieved
by the proposed scheme (channel assignment using the
multi-link conflict graph for interference modeling, and
the delay metric for channel selection). The above result
shows that considering only interfering 802.11 networks
in the channel assignment process is not enough, and
that it is important to also consider the quality of the
links, in terms of the packet delay or the SNR; this
is especially important in networks with long distance
links (typically, above 1 Km as in our metropolitan test-

6
bed), compared to networks with smaller distances (as in
the network of [10] that contains links up to 60 meters).
Figure 4a also shows the average packet delay when
only the link quality (in terms of packet delay) is used,
without modeling intra-network interference. Observe
that the average delay is 29% higher than the average
delay achieved when the multi-link conflict graph is
used, and 10% higher than the case when interferences
from 802.11 networks are considered. This shows that,
in addition to the external interference and link quality
(captured using the packet delay metric), it is important
to consider the intra-network interference in order to
achieve high performance. Finally, the fourth bar in
Figure 4a shows the average delay when only intra-
network interference is taken into account. Observe that
the average delay is 46% higher than the average delay
when channel selection takes into account both external
interference and link quality, in addition to intra-network
interference; hence, many of the schemes appearing in
the literature, which focus exclusively on capturing in-
terference among links inside the network, would yield
very low performance.
Figure 4b compares, for an interval of 22 days, the
average packet delay when the channels are selected
once at the beginning of the 22 days period, with an
adaptive approach where new channels are selected once
every day. Our proposed procedure was used in both
cases. The results show that the fixed approach achieves
an average packet delay that is within 11,5% of the
adaptive approach. This suggests that, under normal
conditions, there are no significant gains in performing
channel assignment in a time-scale smaller than 1-2
weeks. Of course, with the more widespread use of
802.11a networks, this can change in the future.
4 EMF MONITORING USING A METROPOLITAN
MESH NETWORK
In this section we discuss an application for EMF mon-
itoring, that uses the metropolitan wireless mesh net-
work. The key idea is to use a low-cost EMF mea-
surement device located in a node that is connected to
the metropolitan mesh network. The EMF monitoring
node consists of the following parts: a low-cost EMF
measurement device, a mini PC for controlling the EMF
measurement device, and software modules for collect-
ing, processing, and presenting the EMF measurements.
The EMF measurement device is a low-cost Aaronia
Spectran Analyzer. Our current system uses the HF 6060,
which has a measurement range of 10 MHz to 6000
MHz. Additional monitoring nodes that we are currently
deploying will use the HF 6080, which has a range of 10
MHz to 7000 MHz. The HF60X0 spectrum analyzer can
be connected to two types of antennas: a directional an-
tenna (HyperLOG 6080) and an omni-directional dipole
antenna (BicoLOG 20300). The HyperLOG 6080 direc-
tional antenna has a range of 700 MHz to 7 GHz. The
BicoLOG 20300 omni-directional dipole antenna has a
(a) comparison of channel assignment algorithms
(b) Comparison of performance when channels are assigned once in
the beginning of a 22 day period, with the performance when
channels are adjusted each day.
Fig. 4. Channel assignment measurements.
range of 40 MHz to 3000 MHz. The BicoLOG antenna
is more costly than the HyperLOG antenna, but can
measure lower frequencies, which include the FM and
TV bands. On the other hand, the lower cost HyperLOG
directional antenna can measure higher frequencies, up
to 7 GHz, but requires pointing it towards the area we
are interested in measuring. The currently deployed EMF
monitoring node includes the omni-directional dipole
antenna BicoLOG 20300.
The EMF monitoring node contains a small PC, which
is based on a mini-ITX board (EPIA SP 13000, 1.3 GHz
C3, CPU) with 512 MB DDR400 memory, and an 80 GB
2.5” HDD. The mini-ITX PC runs the Aaronia software
for controlling and collecting measurements from the
HF60X0 analyzer, which also allows external configu-
ration of the spectrum analyzer, for parameters that
include frequency range, sample time, resolution band-
width. An apache server running on the mini-ITX allows
remote access of the measurement graphs through a web
interface. There are four Perl scripts executed every 5
minutes, which take the data from the HF60X0 analyzer
and create graphs presenting the EMF measurements
in a different manner: per-band, per-operator, and time
series, and can also be executed on demand through a
web interface.
4.1 Advantages
The advantages of using an EMF measurement device
connected to a metropolitan mesh network for collecting
EMF measurements include the following:

7
•Higher range: EMF measurement device can be used
to monitor frequencies up to 7 GHz (with the appro-
priate antenna), which is higher than the capabilities
from specialized stand-alone EMF monitors, whose
range is typically limited to 3 GHz.
•Real-time remote measurement collection: EMF
measurement devices with real-time monitoring ca-
pabilities together with a metropolitan coverage
mesh network allow real-time remote collection of
EMF measurement data.
•Low cost and high speed: Small (hand-held) EMF
measurement devices with advanced spectrum ana-
lyzer capabilities are significantly cheaper than stan-
dalone EMF monitoring devices with remote com-
munication (GSM, 3G, etc) capabilities, and wireless
mesh networks provide higher speeds at lower costs
compared to mobile technologies.
•Advanced flexibility: Together with the mini PC, the
EMF monitor can be controlled remotely to collect
measurements in different frequency ranges (bands)
and different time windows.
4.2 Monitoring capabilities
The collected measurement data is in dBm units. In
addition to the default units, the measurements can be
presented in dBµV, mV/m, µA/m2and finally mW/m2.
We note here that electromagnetic radiation limits are
usual represented in mV/m or mW/m2.
4.2.1 Per-Band Monitoring
This option allows the presentation of EMF levels in
frequencies of various well-known bands. Figure 5a
shows an example of per-band monitoring. The dis-
played values are the average of measurements taken in
time intervals of approximately 6 minutes. The graph is
refreshed periodically, based on the last stored values in
the database. Each bar corresponds to a different band.
Note that the values for the GSM and 3G/UMTS bands
are higher than approximately -80 dBm, which is due to
the mobile telephony antennas located opposite of the
EMF monitor node.
4.2.2 Per-Operator Monitoring
This option displays EMF levels of per-operator fre-
quencies, according to the official frequency assignments
made by the Greek government (Hellenic Telecommu-
nications and Post Commission (EETT), National Reg-
ulatory Authority, http://www.eett.gr). The EMF levels
include measurements in all frequencies (GSM 900, GSM
1800, and 3G/UMTS), as they are presented in Figure 5b.
4.2.3 Time Series Measurements
Here we present the variations of measurements in a
larger time-frame. To achieve this, we use the RRD
(Round Robin Database) Tool, which allows the daily,
weekly, monthly, and yearly presentation of measured
values. A snapshot of the time series graphical display is
(a) Per band monitoring
(b) Per operator monitoring.
(c) Time series monitoring
Fig. 5. Presentation of EMF measurements.
shown in Figure 5c. Finally, we can use this monitoring
option to compare the EMF levels in different bands,
in the course of time. This will be especially important
as the percentage of data, which is typically bursty in
nature, carried over wireless networks increases.
5 CONCLUSIONS
We presented some of our experiences and investigations
with an experimental metropolitan multi-radio mesh
network. In particular, we investigated how interference
between metropolitan 802.11a links operating on adja-
cent channels affects throughput. For rate control, our

8
results show that the adaptation of the transmission
rate on very small time-scales (on a per packet basis
as is commonly the case with widely used auto-rate
algorithms) not only does not improve performance,
but can also result in significantly lower performance
compared to an appropriately selected fixed transmis-
sion rate scheme. For channel assignment, we showed
results demonstrating the importance of taking into ac-
count both internal and external interference. Our exper-
iments for the time scales of rate, power, and channel
control in metropolitan links with directional antennas,
showed that there are no significant improvements for
performing adaptation in short timescales, of the order of
minutes or hours. Hence, a more appropriate approach
would be to select a static rate, power, and channel,
and continuously monitor the network to detect any
anomalies or performance drops; once such events are
detected, the rate, power, and channel can be adjusted.
Finally, we presented a system for continuous online
Electromagnetic Field (EMF) monitoring, which utilizes
the low-cost wide-area connectivity provided by the
metropolitan mesh network for collecting, processing,
and presenting data from low-cost EMF measurement
devices.
ACKNOWLEDGMENTS
This work was supported in part by the European
Commission in the 7th Framework Programme through
project EU-MESH (Enhanced, Ubiquitous, and Depend-
able Broadband Access using MESH Networks), ICT-
215320, http://www.eu-mesh.eu
REFERENCES
[1] K. Chebrolu, B. Raman, and S. Sen. Long-Distance 802.11b Links:
Performance Measurements and Experience. In Proc. of ACM MOBI-
COM, 2006.
[2] R. Patra, S. Nedevschi, S. Surama, A. Sheth, L. Subramanian, and
E. Brewer. WiLDNet: Design and Implementation of High Performance
WiFi Based Long Distance Networks. In Proc. of USENIX Symposium
on Networked Systems Design and Implementation (NSDI), 2006.
[3] J. Bicket, D. Aguayo, S. Biswas, and R. Morris. Architecture and
Evaluation of an Unplanned 802.11b Mesh Network. In Proc. of ACM
MOBICOM, 2005
[4] D. Wu, D. Gupta, and P. Mohapatra. Quail Ridge Reserve Wireless
Mesh Network: Experiences, Challenges and Findings. In Proc. IEEE
TRIDENTCOM, 2007.
[5] J. Camp, J. Robinson, C. Steger, and E. Knightly, Measurement
Driven Deployment of a Two-Tier Urban Mesh Access Network. In Proc.
of ACM MobiSys, 2006.
[6] G. Bianchi, F. Formisano, and D. Giustiniano. 802.11b/g Link Level
Measurements for an Outdoor Wireless Campus Network in Proc. of
IEEE WoWMoM, 2006.
[7] T. Ireland, A. Nyzio, M. Zink, and J. Kurose. The Impact of Di-
rectional Antenna Orientation, Spacing, and Channel Separation on
Long-distance Multi-hop 802.11g Networks: A Measurement Study. In
Proc. of 3rd Int’l Workshop on Wireless Network Measurement
(WiNMee), 2007
[8] C.-M. Cheng, P.-H. Hsiao, H. T. Kung, and D. Vlah. Adjacent
Channel Interference in Dual-Radio 802.11a Nodes and Its Impact on
Multi-hop Networking. In Proc. of IEEE Globecom, 2006.
[9] M. Delakis and V. A. Siris, Channel Assignment in a Metropolitan
Wireless Multi-Radio Mesh Network in Proc. of the 5th International
Conference on Broadband Communications, Networks and Sys-
tems (Broadnets), London, September 2008.
[10] S. M. Das, H. Pucha, D. Koutsonikolas, Y. C. Hu, and D. Peroulis.
DMesh: Incorporating Practical Directional Antennas in Multi-
channel Wireless Mesh Networks. IEEE J. Select. Areas Commun.,
24(11):2028–2039, 2006.