ArticlePDF Available

Abstract and Figures

Mobile networks are experiencing a great development in urban areas worldwide, and developing countries are not an exception. However, sparsely populated rural areas in developing regions usually do not have any access to terrestrial communications networks because operators cannot ensure enough revenues to justify the required investments. Therefore, alternative low-cost solutions are needed for both the access network and the backhaul network. In this sense, in order to provide rural 3G coverage in small villages, state-of-the-art approaches propose to use Small Cells in access networks and inexpensive multihop wireless networks based on WiFi for long distances (WiLD) or WiMAX for backhauling them. These technologies provide most of the required capabilities; however, there is no complete knowledge about the performance of WiFi and WiMAX in long-distance links under quality of service constraints. The aim of this work is to provide a detailed overview of the different alternatives for building rural wireless backhaul networks. We compare both IEEE 802.11n and IEEE 802.16 distance-aware analytical models and validate them by extensive simulations and field experiments. Also WiFi-based TDMA proprietary solutions are evaluated experimentally and compared. Finally, results are used to model a real study case in the Peruvian Amazon in order to illustrate that the performance provided by these technologies may be sufficient for the backhaul network of a rural 3G access network based on Small Cells.
This content is subject to copyright. Terms and conditions apply.
Research Article
Assessing IEEE 802.11 and IEEE 802.16 as Backhauling
Technologies for 3G Small Cells in Rural Areas of
Developing Countries
Javier Sim ´
o-Reigadas ,
Carlos Figuera,
Eduardo Morgado,
Esteban Municio ,
and Andr´
es Mart´
Rey Juan Carlos University, Department of Signal eory and Communications, Madrid, Spain
IDLabDepartment of Mathematics and Computer Science, University of AntwerpIMEC, Antwerp, Belgium
Correspondence should be addressed to Javier Sim´o-Reigadas;
Received 27 June 2018; Accepted 9 December 2018; Published 17 January 2019
Academic Editor: Paolo Bellavista
Copyright ©2019 Javier Sim´
o-Reigadas et al. is is an open access article distributed under the Creative Commons Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is
properly cited.
Mobile networks are experiencing a great development in urban areas worldwide, and developing countries are not an exception.
However, sparsely populated rural areas in developing regions usually do not have any access to terrestrial communications networks
because operators cannot ensure enough revenues to justify the required investments. erefore, alternative low-cost solutions are
needed for both the access network and the backhaul network. In this sense, in order to provide rural 3G coverage in small villages,
state-of-the-art approaches propose to use Small Cells in access networks and inexpensive multihop wireless networks based on WiFi
for long distances (WiLD) or WiMAX for backhauling them. ese technologies provide most of the required capabilities; however,
there is no complete knowledge about the performance of WiFi and WiMAX in long-distance links under quality of service
constraints. e aim of this work is to provide a detailed overview of the different alternatives for building rural wireless backhaul
networks. We compare both IEEE 802.11n and IEEE 802.16 distance-aware analytical models and validate them by extensive
simulations and field experiments. Also WiFi-based TDMA proprietary solutions are evaluated experimentally and compared.
Finally, results are used to model a real study case in the Peruvian Amazon in order to illustrate that the performance provided by
these technologies may be sufficient for the backhaul network of a rural 3G access network based on Small Cells.
1. Introduction
Although users would like to have universal and ubiqui-
tous access to 3G services everywhere, operators limit
the deployment of infrastructures to areas where reve-
nues compensate the capital expenditures (CAPEX, mainly
related to the deployment of infrastructures) and the op-
eration expenditures (OPEX, including maintenance, op-
eration, licenses, etc). Consequently, many large rural areas
in developing regions that are sparsely populated lack 3G
coverage, and often they do not even have any telecom-
munication services at all. Hence, in the context of the
current trend towards ubiquitous communications
worldwide, the need for more affordable 3G access and
backhaul solutions becomes apparent.
Small Cells, which were initially conceived as a solution
for coverage holes in urban areas, are now becoming a
promising solution for rural access networks. Femtocells are
generally used inside buildings and are designed to be low-
cost, adaptable, and flexible and to have low power con-
sumption so that a residential Asymmetric Digital Sub-
scriber Line (ADSL) may be used as the backhaul. A 3G
femtocell may be a suitable solution for the access network in
a small village, since its cost and power consumption make it
much more affordable for an operator than a common
eNodeB. erefore, a few small cells may ensure cellular
coverage in most human settlements over a large sparsely
populated region at a very low cost.
However, backhauling these 3G femtocells may become
challenging. Many sparsely populated areas may be too far
Mobile Information Systems
Volume 2019, Article ID 4383945, 15 pages
from any well-connected location for a one-hop terrestrial
link. is implicitly leaves only two alternatives: multihop
broadband wireless networks or satellite communications.
Although Very Small Aperture Terminal (VSAT) systems
have a high OPEX and a round-trip delay over 500 ms (if
geostationary communications satellites are used), they are
commonly used as backhaul solutions for remote cellular
spots; alternative terrestrial multihop wireless networks may
be considered, but the cost is then a critical limitation. is
trade-off attracts the interest on nonlicensed low-cost
technologies compliant with wireless broadband standards
IEEE 802.11 [1] (WiFi) and IEEE 802.16-2009 [2] (com-
monly known as WiMAX (although the WiMAX Forum
does not have a profile for the WirelessHUMAN physical
layer that describes the operation in nonlicensed bands)).
is is precisely the approach adopted by the project
TUCAN3G [3] which proposes technically feasible but
economically sustainable solutions for the progressive in-
troduction of voice and broadband data services in small
communities of rural areas of developing countries, using
conventional 3G cellular terminals. Although 4G small cells
(and 5G spots soon) may be a better solution in terms of the
service level offered, they are also much more demanding in
terms of backhaul capacity, which has a direct impact on cost
effectiveness, so only 3G coverage will be considered in this
Although the alternatives pointed out above are be-
coming more and more consistent and widely accepted [4],
there is still a lack of accurate knowledge about the be-
haviour of alternative terrestrial backhaul technologies
used under carrier-grade QoS (Quality of Service) re-
quirements in long-distance links. For this reason, this
paper analyses and compares the expected performance of
WiFi, WiFi-based Time Division Multiple Access (TDMA)
solutions, and WiMAX in long-distance links and de-
termines for these technologies the conditions to be ade-
quate as backhaul technologies for rural 3G femtocells. is
is carried out by first studying their distance-aware ana-
lytical models and then comparing their performance
through extensive simulations and few field tests in a real
rural network. As a side contribution, we provide with the
simulation platform used for testing these technologies as
rural backhaul networks.
e reminder of this work is organized as follows. First,
Section 2 introduces the problematic of a rural network and
gives a general overview of the traffic requirements of a low-
cost rural backhaul network. Section 3 presents the char-
acteristics of three technologies that are relevant for their use
as backhaul of 3G femtocells in rural areas: WiFi, WiFi-
TDMA, and WiMAX. Section 4 presents the theoretical
models adapted to long-distance links for WiFi and WiMAX
systems, including specific simplifications for point-to-point
(PtP) links. Section 5 summarizes the methods and con-
ditions used to validate the theoretical models and to obtain
the performance results that are presented in Section 6.
Finally, Section 7 uses the performance results for validating
the referenced technologies for their use in real scenarios
such as rural backhaul network for 3G femtocells (the Napo
network). Section 8 concludes this paper.
2. Motivation: Rural Wireless
Backhaul Networks
A rural 3G access network based on femtocells requires of a
backhaul solution with enough capacity and bounded delay,
jitter, and packet-loss for an acceptable communication
quality between the femtocells and the operator’s network in
which the HNB (Home Node B) controller resides.
e traffic supported by the backhaul in both directions
contains three main aggregated traffic classes: high priority
signalling traffic exchanged among the HNB and the con-
troller, voice real-time traffic, and data. Of course, a more
complex classification is possible because ‘data’ may include
traffic of very different natures, but this three classes vision is
general enough for the scope of this paper. e backhaul
must have resources available in excess for high-quality
transport for all the traffic or, alternatively, it must be
able to prioritize signalling and voice traffic over more in-
tensive data traffic, shaping the delay, jitter, and packet-loss
as required for each traffic class.
In order to obtain a first overview of the capacity re-
quired for the backhaul of rural femtocells in developing
regions, a prospective study was done in the frame of the
TUCAN3G project [3] based on usage data obtained in
covered areas in Peru. Although this study and its results are
beyond the scope of this paper, there are some outcomes that
are relevant to modelling the offered load for a backhaul
transport network.
2.1. Voice Traffic. Approximately 50% of the people in rural
areas are potential users of telephone, generating a traffic
intensity of 15 mEr/user in the busy hour. With the blocking
probability imposed to the operator in its regulation domain,
this permits to calculate for a given access network the
number of channels required in a coverage area for si-
multaneous phone calls. e throughput generated depends
on the codec used and on the number of simultaneous calls.
e AMR voice codec at 12.2 kbps is assumed, and circuit
switch multiplexing is applied in the UL. Besides the pay-
load, each packet also contains several headers for all the
layers of the protocol stack: RTP, UDP, IPSec, and IP. Based
on these considerations, the link-layer throughput is around
22 kbps in the UL and 60 kbps in the DL per phone call. For
the case of the Peruvian regulatory context, a small village
with only 500 inhabitants would need 9 voice channels,
generating a throughput of 736 kbps of voice traffic in the
busy hour. A large village with as much as 5000 inhabitants
would require 48 voice channels, generating 3936 kbps in the
busy hour. e calculations for other population figures are
2.2. Data Traffic. For data traffic, the potential market is
approximately 5% of the population. A peak throughput per
user of 3 Mbps must be ensured for a good quality of ex-
perience. e minimum capacity per user is 15 kbps
(DL) + 5 kbps (UL). is means roughly that 3000 kbps must
be available for all cases, and the capacity needs to grow
proportionally to the population beyond 3000 inhabitants.
2Mobile Information Systems
2.3. Signalling Traffic. Signalling between the HNB and the
HNB controller has been carefully studied and is known to
be lower than 1% of the traffic [3].
With the previous assumptions, it can be estimated that
the required backhaul capacity of a 3G femtocell in a small
village with less than 500 inhabitants must be at least
4.5 Mbps, and this requirement grows up slightly over
10 Mbps for a population of 5000 inhabitants. In fact, the
data analysed come from areas where 3G coverage already
exists, in which the average income of the population is
higher than that in remote areas. erefore, the projection
overestimates the expected demand in very isolated areas.
ITU recommendations suggest certain end-to-end
values for performance indicators in telephony [5]. 150 ms
is a maximum limit for the one-way delay, and 2% is the
maximum tolerable packet-loss. However, the backhaul has
a share in these end-to-end values. As the access network
and the core network may also introduce packet-loss and
delay, in this paper, the backhaul contribution to these
maximum values is considered limited to 0.8% for the
packet-loss and 60 ms for the delay, which represent a
contribution of 60% to the total values.
Finally, in this paper, we focused our study in the raw
capacity and delay at the MAC layer. erefore, we consider
upper layers out of the scope of the work. On one hand, QoS
provision has been already studied for rural backhaul net-
works in Simo-Reigadas et al. [6, 7], and hence, we assume
that a number of mechanisms such as DiffServ, MPLS, or
MPLS-TE exist for guaranteeing the requirements of each
traffic type and their different applications. On the other
hand, multihop routing is not addressed since it can be
trivially implemented with standard well-studied proactive
routing protocols such as OSPF or even with static routing.
Finally, although we assume that the MAC links are
encrypted with their correspondent L2 security protocol, we
mainly relay in the security provided by the IPSec tunnels
established in the Iuh 3GPP interface between the core
network and the HNB. Accordingly, the authentication
procedures are also transferred to the HNB and the access
3. Using WiLD and WiMAX as Backhaul
Technologies in Rural Areas
3.1. WiLD (WiFi for Long Distances). e IEEE 802.11
standard [1], popularly known as WiFi (accurately
speaking, WiFi is a commercial denomination for products
certified by the WiFi Alliance as IEEE 802.11-compliant
systems that meet certain requirements), is conceived for
local and metropolitan area networks operating in 2.4 GHz
or 5 GHz nonlicensed bands. e Medium Access Control
(MAC) protocol used is Carrier Sense Multiple Access with
Collision Avoidance (CSMA/CA), which is not well suited
to long-distance communications. Nonetheless, some re-
searchers have demonstrated that WiFi may also be used
for long-distance links in rural areas, as far as line of sight
(LOS) may be assured [8, 9]. Additionally, Reigadas et al.
[9] proposed adjustments to CSMA/CA for optimal per-
formance and showed in what conditions general analytical
models are also valid for long distances. WiFi can be used
normally with 20 MHz channels, but 5, 10, 40, and 80 MHz
channels are also permitted. e standard includes several
alternatives for the physical layer (PHY), being the High
roughput (HT) Orthogonal Frequency-Division Multi-
plexing (OFDM) PHY the one with best performance.
Additionally, spatial diversity with Multiple Input Multiple
Output (MIMO) can be used, and many commercial
outdoor WiFi systems are prepared with dual antennas for
MIMO 2 ×2. Depending on the received signal strength,
different Modulation and Coding Schemes (MCSs) may
also be used permitting different physical bitrates. For
example, MIMO 2 ×2-enabled 802.11n systems can operate
from a 6.5 Mbps bitrate (MCS0, 20 MHz channel, Single
Input Single Output (SISO), 800 ns of guard interval in
OFDM symbols) up to 300 Mbps (MCS15, 40 MHz chan-
nel, MIMO 2 ×2, 400 ns of guard interval). However, the
actual saturation throughput at MAC layer is only a
fraction of that PHY bitrate. Other interesting properties
offered by WiFi are frame aggregation, which permits to
integrate big bundles containing many frames in a single
header, and traffic differentiation, which permits to pri-
oritize some traffic classes over others.
No matter what MAC protocol is used, any commu-
nications solution with WiFi hardware can benefit from the
spatial diversity in the HT PHY, which raises the question
whether these benefits apply to long-distance LOS links or
not. ere is a flexible trade-off between diversity gain and
multiplexing gain [10]. In order to benefit from any of them,
the multiple spatial data streams transmitted in parallel must
be highly independent. It is possible to achieve long-distance
high-rank LOS MIMO channels, at the cost however, of an
excessive size of towers to ensure the necessary separation
between antennas [11, 12]. Nonetheless, cross polarization
makes possible to obtain two orthogonal channels, especially
in LOS scenarios and only slightly worsened with long
distance [13–15].
3.2. 802.11n-Based Proprietary Solutions Using TDMA MAC.
Some manufacturers of broadband wireless systems with
IEEE 802.11 interfaces include as an option an alternative
proprietary MAC protocol in their outdoor products for
improved performance for long-distance links. Normally
none of these proprietary MAC protocols is published with
all the technical details. Only limited information is available
in the datasheets provided and unbiased raw data from the
vendors is usually difficult to obtain, since this may even-
tually lead to the data being used against the issuing com-
pany itself. However, implementations of academic
proposals of this type like [16] permit to guess what these
systems do. At a first glance, these products use common
802.11n cards with modified drivers that disable the CSMA/
CA MAC, thus converting the wireless network interfaces in
raw data transceivers that use the 802.11 PHY. On top of this,
a simple TDMA Time-Division Duplexing (TDD) solution
seems to be programmed. is TDMA protocol avoids
contention, making the use of the wireless channel much
more efficient at long distances.
Mobile Information Systems 3
Hence, these systems show a behaviour similar to
WiMAX (see Section 3.3), having lower spectral efficiency,
but with potentially higher capacity due to the wider
channels they can use. Some researchers such as [16, 17]
have proposed these WiFi-based solutions for outdoor long-
distance links and [18] demonstrated that under similar
conditions, this kind of solutions tends to be better than
common WiFi for very long distances, while WiFi can be
tweaked to be better for short and medium distances.
Among many others, Ubiquiti AirMAX [19] and Mik-
rotik NV2 [20] product families are extremely popular for
long-distance links as low-cost powerful solutions. ese
systems can be experimentally tested, but, due to the pro-
prietary nature of these solutions, it is not possible to either
compare them theoretically or to extend the experimental
results to other comparable solutions.
3.3. WiMAX for Long Distances. e IEEE 802.16 standard
[2], popularly known as WiMAX (as for the case of WiFi,
WiMAX is a certification given by the WiMAX Forum to
products that are IEEE 802.16-compliant, that pass certain
interoperability tests and that fit in certain profiles; the use of
the WiMAX denomination in this paper is formally wrong
because there is not any profile in the WiMAX Forum for
products using the IEEE 802.16 WirelessHUMAN PHY;
however, the term is used in the paper as a synonym of the
standard to make it more readable), is conceived for met-
ropolitan area networks, although the solution is also valid
for rural broadband networks. e WirelessHUMAN PHY
in the standard describes the operation in the 5 GHz non-
licensed band. Other WiMAX profiles allow to operate in
licensed bands, which is very relevant for these countries
where the use of licensed bands in backhaul network is
mandatory. However, although the 5G Hz band is more
prone to interferences than licensed bands, its use for
carried-grade service cannot be neglected due to the high-
gain directive antennas used and the limited radio-frequency
emissions present in unpopulated rural areas.
WiMAX uses OFDM at the PHY layer and TDM/TDMA
with TDD for contention-free operation. Different traffic
classes are recognized and can receive differentiated QoS
guarantees. In WirelessHUMAN, channels can be as wide as
10 MHz. e capacity of a WirelessHUMAN link depends
on the MCS (being BPSK 1/2 the slowest and 64 QAM 3/4
the fastest), the guard interval in OFDM symbols (1/4–1/32),
the frame size (2.5–20 ms), and so on. A stable WiMAX link
may have a raw capacity between 1.6 Mbps and 36 Mbps. e
link distance impacts on the link budget, determining what
MCS can be used, but has little impact on the performance
otherwise. Regarding the use of spatial diversity, the same
considerations previously explained for WiFi can also be
extended to WiMAX.
Although WiLD, WiFi-TDMA, and WiMAX seem
promising solutions for rural backhauls, there are not
previous systematic analysis aiming to know the real MAC
capacity in long-distance links and no works exist that
compare their performance depending on the distance and
on the real impact of the different features mentioned above.
To the best of our knowledge, this work is the first in doing
4. Performance Analysis of IEEE 802.11 and
802.16 WirelessHUMAN Long Links
4.1. Modelling the Performance of 802.11 for Long Distances.
Simo et al. [9] proposed an analytical model for long-
distance WiFi links based on [21]. While the hints given
in [9] for the adaptation of WiFi to long distances are valid
for any standard of the 802.11 family, the performance
model itself is not valid for 802.11n because of certain
limitations inherited from Bianchi’s model [21]. Bianchi
revised his model and proposed a new one in 2010 [22] that
can be applied to the different PHY in IEEE 802.11-2012,
which can be summarized in the following equations:
1+1p/2 1 pR+1
 􏼁 􏼁􏽐R
( ) (1p/2),
where pand τare, respectively, the conditional collision
probability and the probability of a station to transmit in any
slot; Wis a constant, corresponding with the highest possible
value of the contention window for a packet that is being
transmitted for the first time; and Ris the number or
retransmissions (not considering the first transmission)
before dropping a frame and trying with the next; and nis
the number of WiFi stations.
e previous analysis assumes an ideal channel in which
packet-loss only occurs by collision when two or more
stations transmit simultaneously. However, in real condi-
tions, the wireless channel may cause significant packet-loss
due to channel errors and interferences. erefore, it is also
necessary to take in account the Packet Error Rate (PER) in
the model as previously done in other works [23, 24]. When
using 802.11 as a backhaul technology in rural areas, in-
terferences are not common, and packet-losses are also
infrequent because modulation and coding schemes are
chosen for having a large margin between RSSI (Received
Signal Strength Indication) and Sensitivity. However, we can
carry out a simple and practical analysis considering that a
transmitting WiFi station is unable to distinguish whether a
transmitted packet has not reached its destination due to
collision or the packet has arrived but with errors. In both
cases, the transmitting station takes the same actions to
retransmit the packet [25, 26]. us, we can just redefine pas
the probability that a packet should be retransmitted either
by collision or errors in the channel, and (2) becomes
􏼐 􏼑(1PER) + PER.(3)
is set of equations cat be solved numerically. For the
case of ideal channels, Table 1 shows the results for the most
important values of W that can be considered here (the
default values for the different PHYs), for the standard
number of retransmissions (7 for basic mode, 4 for RTS/CTS
mode), and for up to 5 stations. For the case of real channels,
4Mobile Information Systems
Tables 2 and 3 show the corresponding results considering
PER 0.01 and 0.1, respectively. It is important to notice
that PER 0.1 is actually a maximum value because the
standard uses it for the definition of sensitivity and de-
termines that a link should not be established with higher
PER values.
is model is valid for long-distance links, conditioned
to the adjustment of SlotTime, ACKTimeout, and CTSTi-
meout MAC parameters as indicated in [9].
Based on the values of pand τobtained from (1) and (3),
the equations proposed in Section 5 of [22] for the saturation
throughput and for the delay are valid with some remarks:
(i) e SlotTime δmust be replaced in all expressions by
δdfollowing indications in [9].
(ii) e time durations used in the throughput equation
must be redefined considering the propagation time
Tp, calculated consistently with [9].
E Ts
􏼂 􏼃δd+W
􏼐 􏼑,
E Tc
􏼐 􏼑,
for basic mode, or
E Ts
􏼂 􏼃δd+W
E Tc
􏼐 􏼑,
for RTS/CTS mode.
e model in [9] modified by the previous equations
permits to obtain the saturation throughput and delay for
WiFi links at any distance. is result provides the maxi-
mum capacity at link level. e following subsections will
use the saturation point as a reference to find an optimal
unsaturated operation point that keeps the delay under a
threshold that is adequate for a backhaul link. e model will
be used as a reference to validate the NS3 simulator, which is
the tool that allows to accurately study the performance of
long-distance links in a flexible manner.
4.1.1. Simplified WiLD Model for a PtP Setup. e previous
model is valid for nstations, and doing n2, we obtain a
PtP model where (3) becomes
is leads to validate performance results for a PtP
WiLD link under saturation conditions. Although the model
used is based on the hypothesis of nbeing higher, simu-
lations have proved that the incremental error using the
model for n2 (point-to-point case) is negligible.
4.2. Modelling the Performance of WiMAX. In this section,
we develop a theoretical model for computing the perfor-
mance of IEEE 802.16 WirelessHUMAN. First, we present a
general model which accounts for different number of sta-
tions, several strategies for packing the information in the
MAC level, and the possibility of including a MIMO radio
interface. All the computations are based on information
gathered from the IEE802.16-2009 standard [2] and the
analysis in [27]. Table 4 collects the main parameters used in
this analysis.
802.16 WirelessHUMAN profile [2] assumes OFDM
modulation with N256 subcarriers. Only Ndata 192 of
them are used for data. A cyclic prefix of length CP ·Nis used
to avoid intersymbol and intercarrier interference. Given the
sampling frequency FsnsBwhich depends on the channel
bandwidth Band the sampling factor ns, the duration of one
OFDM symbol is TOFDM N(1+CP)/(nsB). Each OFDM
symbol transmits with only one MCS with modulation
order Mand coding rate r. Defining the function β(r, M) �
Ndatarlog2(M)as the bits carried by an OFDM symbol, the
raw bitrate supported by the OFDM modulation is
PHY β(r, M)
Available MCSs are shown in Table 5. If MIMO is used, a
maximum multiplexing gain Gmax min(Mt, Mr)can be
obtained, where Mtand Mrare the number of transmitter
and receiver antennas, respectively. Assuming that more
than one antenna is present both in the transmitter and
Table 1: Values of pand τcalculated from (1) and (2).
nW4W8W16 W32
p τ p τ p τ p τ
2 0.289 0.289 0.188 0.188 0.109 0.109 0.058 0.058
3 0.383 0.214 0.280 0.151 0.183 0.096 0.106 0.054
4 0.438 0.174 0.337 0.128 0.236 0.085 0.146 0.051
5 0.477 0.149 0.377 0.111 0.276 0.077 0.180 0.048
Table 2: Values of pand τcalculated from (1) and (3) with
PER 0.01.
nW4W8W16 W32
p τ p τ p τ p τ
2 0.293 0.286 0.194 0.186 0.116 0.107 0.067 0.057
3 0.386 0.212 0.284 0.150 0.189 0.095 0.114 0.054
4 0.440 0.173 0.340 0.126 0.241 0.084 0.153 0.051
5 0.479 0.148 0.380 0.110 0.280 0.076 0.186 0.048
Table 3: Values of pand τcalculated from (1) and (3) with
PER 0.1.
nW4W8W16 W32
p τ p τ p τ p τ
2 0.330 0.256 0.248 0.165 0.186 0.095 0.146 0.051
3 0.413 0.192 0.324 0.133 0.245 0.084 0.184 0.048
4 0.463 0.158 0.373 0.113 0.288 0.075 0.216 0.045
5 0.500 0.136 0.408 0.099 0.321 0.068 0.243 0.042
Mobile Information Systems 5
receiver, the actual multiplexing gain Gwill depend on the
MIMO configuration [2] and the statistical dependence
among channels, so 1 GGmax. en, raw physical rate
would be given by R(raw)
Raw physical rate is diminished due to frame structure.
First, time gaps RTG (Receive/transmit Transition Gap) and
TTG (Transmit/receive Transition Gap) allow the trans-
ceiver to switch between transmission and reception. If TDD
is used, temporal split between downlink (DL) and uplink
(UL) can be modified. We denote fDL for the fraction of
useful transmission time which is used for DL and Tffor the
frame duration. en, the number of OFDM symbols in the
DL and UL can be computed as
 􏼁fDL
􏼁 1fDL
 􏼁
In the DL, the frame begins with a preamble (2
symbols) and the Frame Control Header (FCH) field (1
symbol). en, a broadcast (BC) burst is transmitted which
contains the UL-MAP, the DL-MAP, and the downlink
and uplink channel descriptors (DCD and UCD, re-
spectively). e length of these fields depends on the
number of modulations and users that are active in a
frame, and the number of optional variable-length mes-
sages which are used. erefore, the most relevant over-
head in the BC burst is due to
(i) UL-MAP: bUM 64 +48(Nss +2)bits (mandatory
every frame), where Nss is the number of active
subscribers in that frame;
(ii) DL-MAP: bDM 64 +32Nss bits (optionally every
(iii) UCD and DCD: bDCD 16 +24NDL
bbits bUCD
16 +24NUL
bwhere NDL
band NUL
bare the numbers
of bursts which are scheduled in the DL and the
UL, respectively, and are usually different.
Note that the transmission frequency for the channel
descriptors fCD depends on the channel coherence time and
the implementation. e number of OFDM symbols used
for the BC burst is
bUM +bDM +bDCD +bUCD
 􏼁fCD
 􏼁
where MBC 2 and rBC 1/2 are the modulation order and
coding rate for the BC burst, respectively. e BC burst and
the users’ bursts are optionally preceded by a short 1-symbol
preamble. If we consider that each user is assigned a single
burst, the number of symbols in the DL that can be used for
MAC data is
raw 2+1+SBC +1Nss +1
 􏼁 􏼁.(10)
In the DL, the following fields reduce the number of
available OFDM symbols:
(i) Ranging (SRNG SRNGOPNRNGOPS ). A variable
number of opportunities NRNGOPS are given for
user ranging. e duration of each one SRNGOP
depends on the maximum user distance allowed in
the cell and the particular implementation.
Table 4: WiMAX model and simulation parameters.
Parameter Description Value (std.) Value (exper.)
TpAir propagation time (D[m]/300)
NSubcarriers 256 —
Ndata Subcarriers used for data 192
BChannel bandwidth 10 MHz
nsSampling factor 144/125
CP Cyclic prefix length 1/32 to 1/4 1/4
TfFrame duration 2.5 to 20 ms 2.5 ms
TTG Transmit/receive transition gap 100 μs
RTG Receive/transmit transition gap 100 μs
fDL Fraction of time for downlink 0.5
(M, r)(Modulation order, coding rate) (2,(1/2))(64,(3/4))
SRNGOP Ranging op. Duration (OFDM symbols) To max(1,2×Tp)
MTU Maximum transmission unit 1400 bytes 1400 bytes
Table 5: Sensitivities from typical equipment used to compare the
different technologies for SISO and MIMO [28, 29].
Sensitivity (dBm)
Wifi and NV2 WiMAX
MCS0 96 BPSK 1/2 92
MCS1 95 QPSK 1/2 91
MCS2 92 QPSK 3/4 89
MCS3 90 16 QAM 1/2 87
MCS4 86 16 QAM 3/4 84
MCS5 83 64 QAM 2/3 80
MCS6 77 64 QAM 3/4 76
MCS7 74 —
MCS8 95 BPSK 1/2 94
MCS9 93 QPSK 1/2 93
MCS10 90 QPSK 3/4 92
MCS11 87 16 QAM 1/2 89
MCS12 84 16 QAM 3/4 85
MCS13 79 64 QAM 2/3 81
MCS14 78 64 QAM 3/4 79
MCS15 75 —
6Mobile Information Systems
(ii) Bandwidth request (SBR). A variable number of 2-
symbol length bandwidth request is scheduled in the
UL to allow users to request bandwidth for each
(iii) A 1-symbol preamble is placed before each UL
If we consider that each user is assigned to a different
burst, the number of symbols in the UL that can be used for
MAC data is
raw SRNG +SBR +Nss
 􏼁.(11)
In a particular frame B(x), bursts are transmitted in the x
direction, with xDL,UL
{ } a variable introduced to sim-
plify the notation. Each burst carries S(x)
bOFDM symbols,
each one with a MCS defined by the coding rate r(x)
band the
modulation order M(x)
b, with b the burst index. e number
of bursts and the MCS for each one depends on the QoS
needs and the channel state for each user, and the following
equalities must hold: 􏽐B(x)
S(x). en, the net rates
provided by the PHY to the MAC level are computed as
with b(x)the number of bits that are available for MAC
PDUs in the xdirection, which is given by
b, M(x)
􏼐 􏼑.(13)
Once the PHY operation has been modelled, the rest of
this subsection is going to focus on the MAC layer. e
subsequent analysis is valid for the DL and UL. e main
overheads introduced in the MAC level are (i) the MAC PDU
header (6 bytes) and CRC (4 bytes), so the overhead is OPDU
(6+4)8bits and (ii) the Packing Subheader (OPSH 2bits)
before each SDU, if packing is enabled. Also, several man-
agement messages are sent as MAC PDUs, including func-
tionalities like ARQ and H-ARQ. Finally, other secondary
subheaders are described in the standard for different sig-
nalling procedures. Since management messages and sec-
ondary subheaders are not very frequent, we only consider the
PDU header and CRC and the Packing subheaders.
In the MAC level, each burst may carry one or more
PDUs, depending on whether the MAC SDUs packing
functionality is enabled. If packing is enabled, several MAC
SDUs are packed in a unique PDU, and usually only one
MAC PDU exists per burst. If not, one MAC SDU is en-
capsulated using one MAC PDU, and several PDUs exist in a
burst. Let us denote Pbas the number of PDUs in the b-th
burst and Db,p as the number of SDUs in the p-th PDU of the
b-th burst. Under these assumptions, the MAC overhead in
the b-th burst in the xdirection is computed as
bOb,PAD +􏽘
OPDU +Db,pOPSH ,(14)
where Ob,PAD are padding bits in order to complete the last
OFDM symbol of the b-th burst.
Finally, the MAC rate for the xdirection is computed as
b, M(x)
􏼐 􏼑O(x)
4.2.1. Simplified WiMAX Model for a PtP Setup. If WiMAX
is used in a PtP link, only one subscriber is connected to the
base station (Nss 1), and several simplifications in the
model may be done. Moreover, in order to compare the
model with our simulations and experiments, we use the
specific configuration used during the tests. First of all, for
only one user, the DL-MAP is optimized by carrying
scheduling information in the Downlink Frame Prefix
(DLFP) and reducing the length of the DL-MAP to only 32
bits. Regarding the DCD and the UCD, a fixed number of
b7 and NUL
b11 bursts, respectively, are considered
by the hardware. However, only one frame, each 1200 ms,
carries these two fields. In the UL, one ranging opportunity
and two bandwidth request opportunities are provided by
default, although these are configurable by software. Since
only one user is active, the number of bursts is one, and the
number of bits available for MAC PDU’s in (13) is simplified
as b(x)
PHY S(x)β(r(x), M(x)), where the number of symbols
S(x)are computed by taking into account the assumptions
made above, and r(x)and M(x)are the coding rate and
modulation used for each direction.
In the MAC level, SDU packing is enabled in our device,
and only one PDU is carried in the only burst in the sub-
frame. Hence, MAC overhead is now computed as O(x)
OPAD +OPDU +D(x)OPSH, where OPAD accounts for the
padding bits, and D(x)is the number of MAC SDU’s that are
packed in a subframe. e latter can be computed as
where the ceil operator is used since we assume that frag-
mentation is enabled and MTU is the maximum trans-
mission unit, which equals the SDU length in our setup.
Finally, the MAC rate is computed as
MAC b(x)
PHY O(x)
􏼐 􏼑 1
5. Measuring the Performance of the
Technologies Being Assessed
Once the WiLD and WiMAX performance models have
been described, we validate them comparing theoretical
results with an external reference. On one hand, the use of
the distance as a fundamental variable, with objective values
in the range 0–60 km, makes almost impossible to run ex-
tensive tests with real equipments at arbitrary distances. On
the other hand, it is necessary to extend the capability of
measuring performance indicators to any distance, with any
configuration and under any traffic load.
For these reasons, a simulator can be used to cross
validate both the models, obtaining results for a more
Mobile Information Systems 7
realistic and wider set of scenarios and configurations. e
event-driven NS-3 network simulator [30] has been chosen
as the most widely used in research on wireless networks (the
code and scripts implemented for this work for testing long-
distance point-to-point links for WiLD and WiMAX in NS-3
are available in
tucan3g). Finally, laboratory and field tests, although limited
to 0 km (5 m) and 30 km (29.8 km), will additionally permit
to cross check these simulation results as well. Note that in
this work, where we use the term 0 km, we always mean
5 meters, and for the experiments at 30km, we mean actually
29.8 km.
For NV2 and AirMAX systems, the lack of information
about the implemented protocols and algorithms in these
proprietary solutions makes impossible both theoretical
analysis and software simulation. A testbed in laboratory has
permitted to measure the performance at short distances,
and the small knowledge about this technology together with
punctual measurements on a few long-distance real links has
been the base for projections that permit to estimate the
expected performance at any distance in the range of in-
terest. Mikrotik RouterBoard R52Hn with MIMO 2 ×2
capabilities has been used for experimenting with the
802.11n PHY layer, both with the standard CSMA/CA MAC
layer and with the proprietary NV2 MAC layer, and Ubiquiti
Rocket M5 systems have been used for testing AirMAX.
We have performed all tests in the 5 GHz band, injecting
traffic with D-ITG [31]. Since NV2 and AirMAX have shown
equivalent performance in the experiments, only NV2 has
been retained for comparison with WiLD and WiMAX. A
few WiMAX experiments have also been run in order to
validate theoretical calculations and simulations at short
distances, using Albentia ARBA PRO 4900–5875 MHz base
station and subscriber stations.
For simulations of WiLD links with NS-3, AckTimeout
and CTSTimeout values have been correctly set as explained
in [9]. Table 5 shows the typical specifications in WiLD,
WiMAX, and NV2 equipment. Table 6 shows the parameters
taken to estimate the link budget. We have used a bi-
directional UDP static flow for both the simulations and the
experimental tests. We focus our performance analysis in
raw capacity as backhaul links. For performance analysis for
different applications and traffic types and under QoS
constraints in rural backhaul networks, we refer to our
previous works [6, 7]. Finally, other performance metrics
can be easily obtained with the same tools as done in
previous works [32, 33].
6. Performance Results
6.1. Cross-Validating the Analytical WiLD Model and NS-3.
In this subsection, we present comparative results between
the model, the NS-3 simulator, and the performed experi-
mental tests. e tests have been done by saturating a point-
to-point link bidirectionally with UDP packets with 1372
bytes of payload, without neither frame aggregation nor
Block-ACK, using 20 MHz of bandwidth and adapting the
SlotTime to the distance as indicated in [9]. e Guard
Interval is set to 800 ns in all cases. Each sample corresponds
to the averaged value of 20 samples (simulation and
Figure 1(a) shows the saturation throughput achieved by
WiLD as the distance is increased. For the sake of clarity, we
only represent the results with the lowest and highest MCS
with SISO and MIMO 2 ×2, i.e., MCS0 and MCS7, and
MCS8 and MCS15, respectively. Continuous and dashed
lines represent the results obtained through the analytical
model and the NS-3 simulator, respectively. It can be seen
that the differences between analytical model and simulation
are negligible, specially for the lower MCSs. Figure 1(a)
shows also the saturation throughput achieved in experi-
mental tests in laboratory conditions at a distance of 0 km.
Although the IEEE 802.11 implementation in NS-3 has been
already widely validated by a number of works [34, 35], we
show that our preliminary results also confirm the match
between the three approaches.
6.2. Cross-Validating the Analytical Model for WiMAX and
NS-3. Tests and calculations have been done saturating a
point-to-point link bidirectionally with UDP packets with
1372 bytes of payload, using the most efficient frame du-
ration and cyclic prefix: frame length of 20 ms and cyclic
prefix of 1/32. e UL/DL ratio is set to 50%. Results are
obtained experimentally for a distance of 0 km (5 m) in
laboratory, and for all distances for the analytical model and
for NS-3. Figure 1(b) represents the aggregated saturation
throughput of UL and DL for MCS from BPSK 1/2 up to 64
QAM 3/4. At a distance of 0 km, the experimental mea-
surements and the simulations match fairly well, and the
results obtained with the analytical model are also very close
to the experimental results (errors under 1% for the worst
case, with frame duration of 20 ms). It is convenient to
indicate that WiMAX systems have many more imple-
mentation specific options that condition the performance
importantly, and only manufacturers know the exact ad-
justments for their products.
6.3. Results Obtained for WiLD. e aim of this subsection is
to determine what is the optimal performance of IEEE
802.11n depending on the distance for long point-to-point
links and how it can be achieved. ere are two main
questions that have to be answered: what is the dependence
of the capacity on the frame aggregation and what is the
expected delay depending on the load level.
Figure 2(a) presents the saturation throughput achieved
for different frame aggregation levels. e Guard Interval is
800 ns, the bandwidth is 20 MHz, and UDP packets with a
payload of 1372 bytes are injected bidirectionally in order to
saturate the link. Nis the number of packets aggregated per
Table 6: Parameters used to estimate link budgets.
Link budget parameters
Transmission power 24 dBm
Directional antenna gain 27 dB
Cable and connectors attenuation 2 dB
Margin 20 dB
8Mobile Information Systems
transmission. Although the standard permits to achieve
higher levels of aggregation, many implementations have a
maximum threshold of 8192 bytes/block, and the results
show that the advantage of N>5 is little in terms of capacity.
From Figure 2(a), the interest of using frame aggregation
with the highest possible value becomes apparent.
In order to represent how the average delay changes with
the frame aggregation level, the distance has been fixed to
30 km, and the results are given in Figure 2(b). e Xaxis
represents the offered load as a percentage of the nominal
bitrate, where 100% represents the PHY throughput of a
given MCS at a given distance. MCS0 is used in this case, and
the rest of parameters are the used ones for the previous
Figure 2(a)).
It can be seen that the advantage of the frame aggre-
gation in terms of capacity is at no cost in terms of latency.
e curves of latency (log scale) are very characteristic and
drive to some clear conclusions. e delay is low (around
3 ms in the simulations and experiments) up to a point
where it grows up to values of more than 1 second. is high
increase in latency is basically due to the queuing delay
originated when CDMA/CA is at its limit and collisions
(although n2, traffic is bidirectional) reduce the channel
efficiency. is reveals that although the throughput can be
increased substantially up to a maximum saturation
throughput value (100% of the link capacity), it is advisable
to keep the load of the link below the saturation point
(i.e., below the 50% of the actual maximum throughput, in
roughput (Mbps)
Distance (km)
0 102030405060
roughput (Mbps)
Distance (km)
0 102030405060
Exp BPSK 1/2
Exp QPSK 1/2
Exp QPSK 1/2
Exp 16QAM1/2
Exp 16QAM3/4
Exp 64QAM3/4
Exp 64QAM2/3
Figure 1: Average saturation throughput values obtained with the model, NS-3 simulator and experimental tests regarding the distance.
(a) WiLD roughput without frame aggregation. (b) WiMAX roughput with 20 ms of frame duration and 1/32 CP.
roughput (Mbps)
0 102030
Distance (km)
40 50 60
Figure 2: Simulation results for WiFi. (a) Average saturation throughput values for different frame aggregation thresholds. (b) Average
delay values for different frame aggregation thresholds regarding the load at a distance of 30 km.
Mobile Information Systems 9
the flat part of Figure 2(b)), so that the delay stays bounded
under low delay figures.
Additionally, the packet-loss is zero up to that satu-
ration point and increases linearly beyond that point. is
is specially crucial if the link is used for backhauling traffic
from 3G cells, where the QoS requirements of telephone
enforce the use of the link only in the low-latency region
(as seen in the previous subsection). For example, the
theoretical bitrate for MCS0 is 5.5 Mbps at MAC level, but
the maximum offered load for a 30 km link that keeps the
average delay under 5 ms is 3.3 Mbps, which is approx-
imately 55% of the nominal bitrate. By limiting the traffic
in the link and keeping it under its saturation point, the
capacity decreases, but permits to obtain a low delay and
negligible packet-loss in exchange. at threshold must
be known for all MCS and all distances. ese values
(shown in the following subsection) can be obtained
with NS-3 and permit to associate any link (given distance
and MCS) in the network to a maximum allowed traffic
load [6, 7].
6.4. Results Obtained for WiMAX. e performance of
WiMAX long-distance links depends on many parameters,
and the adjustments for long point-to-point links are
straight-forward for many of them. For example, there are
different options for the bandwidth, being 10 MHz the
highest in the 5 GHz nonlicensed band. As this is the widest
bandwidth available, it will be the bandwidth used in all
cases. Other parameters such as the cyclic prefix may depend
on the specific context: the lowest cyclic prefix (1/32) permits
to obtain the highest capacity, but it can only be used if
reflections are not significant. e frame duration has to be
studied more carefully. On one hand, a higher frame du-
ration reduces the overhead, increasing the efficiency. On
the other hand, the longer the frames, the higher the average
Figures 3(a) and 3(b) show how the capacity and the
delay evolve for different frame durations as the distance is
increased. Figure 3(a) is obtained with BPSK 1/2, and
represents the aggregated UL + DL saturation throughput
injecting 1372 bytes UDP packets, with a cyclic prefix of 1/4
and a bandwidth of 10 MHz. e saturation throughput is
lower for the shortest frames; however, Figure 3(b) shows
how the latency (log scale) in the nonsaturation region in
unsaturated WiMAX links is penalized as the frame duration
is increased. While a frame duration of 20 ms drives to the
highest capacity, the drawback is a delay as high as 42 ms for
best-effort traffic.
On the contrary, the shortest frame drives to a very short
delay, around 7 ms, but the capacity drops more than 30%.
is trade-off should be solved looking first at how much
delay can be spent in the WiMAX link, and then choosing
the longest frame duration that still keeps the average delay
under that limit. In this paper, a 4 ms frame duration is
chosen for comparisons. Finally, it is worth stressing that
since WiMAX is a TDMA technology, the increase in delay
after the saturation point is practically only due to queuing
delay originated by the PHY bottleneck.
6.5. Results Obtained for NV2 (A WiFi-Based TDMA
Solution). As explained in the previous section, nobody
except the manufacturers (i.e., Mikrotik) can propose an
analytical model for NV2. However, based on the little
existing knowledge about its nature and based also on the
experience of many users [36, 37], it is known that the
capacity of a long NV2 link drops very slowly and linearly
with the distance, just as WiMAX. In absence of theoretical
models and simulators, only experimental measurements
permit to obtain the performance of NV2 for different MCS.
e performance has been measured for 0 km of distance in
laboratory (5 m) and for a 30 km link on the field (actual
distance 29.8 km). e measurements are done with similar
adjustments as in the case of WiLD and a frame duration of
2 ms and due to the restrictions of the link budget, only up to
16 QAM 3/4 could be tested (i.e., MCS4 for SISO and MCS12
for MIMO).
Table 7 shows the saturation throughput for the different
MCSs for 0 and 30 km for NV2. Also values for WiLD are
shown as reference. In this case, a frame aggregation of 8192
bytes is used, which highly boosters the performance in
comparison with the models, specially at the higher MCSs.
As can be seen, WiLD behaves at short distances better than
NV2 for the different MCSs; however, at 30 km, NV2 renders
much higher throughput due to its TDMA MAC layer, as
WiMAX does. is is due to that both NV2 and WiMAX can
adjust the timers for high-propagation times so that the
effects of the distance in the throughput are minimized,
while for NV2, there is a performance drop of around 10%
for 30 km, in WiLD, this drop in performance is more than
50%. is shows that for long distances, the efficiency is
higher for TDMA MAC layers than CSMA MAC layers.
Finally, Figure 4 shows NV2 latency in function of the
offered load at 30 km. It can be seen that the range of offered
load with low delay is around 60%, which is equivalent to the
values in WiLD for frame aggregation N6, as shown in
Figure 2(b). Note that as with WiLD and WiMAX, a traffic
load of 100% is equivalent to the maximum nominal rate
available in the link.
6.6. Comparison between Different Technologies. Once the
three technologies have been characterized in performance
for different distances, we compare them subject to the
conditions that permit to use them as backhaul for 3G rural
For this reason, Figure 5 compares 802.11n (WiLD),
WiMAX, and NV2 in terms of the maximum supported
throughput with a delay lower than 5 ms (i.e., unsaturated
conditions). Since transport networks are likely multihop,
the cumulative nature of the delay enforces us to limit the
one-hop delay strongly. is means that 2.5 ms of frame
duration is chosen for WiMAX and 2 ms for NV2.
For WiLD, the frame aggregation is set to N6 since it
does not affect the latency in the nonsaturated region. Re-
sults are obtained, as before, for bidirectional UDP traffic
with 1372 byte payload size. For WiLD and NV2, we use
GI 800 ns and a cyclic prefix of 1/4 for WiMAX. e
channels are 20 MHz width for WiLD and for NV2 and
10 Mobile Information Systems
10 MHz for WiMAX. Since a linear behaviour is expected in
the performance of NV2 (as occurs with WiMAX) due to its
TDMA-based protocol, we have interpolated NV2 values
between 0 and 30 km and between 30 km and 60 km, so that
all NV2 values are estimated values but the values at 0 and
30 km.
e best performance at short distances is obtained with
WiLD, up to 2-3 km. It is important to notice that, beyond
that distance, NV2 is always better than WiLD as already
roughput (Mbps)
Distance (km)
0 1015202530 4035 45 50 55 60
Frame 2.5ms
Frame 8ms
Frame 20ms
Latency (ms)
Link load (%)
0 1020304050 7060 80 90 100
Frame 2.5ms
Frame 8ms
Frame 20ms
Figure 3: Simulation results for WiMAX. (a) Average saturation throughput values for different frame duration in WiMAX regarding the
distance. (b) Average delay values in WiMAX for different frame duration regarding the load at a distance of 30 km.
Table 7: Average saturation throughput values in Mbps obtained in
real experiments for NV2 and WiLD for 0 and 30 km.
0 km 30 km 0 km 30 km
NV2 MCS0 4.35 4.1 WiLD MCS0 6.1 5.8
NV2 MCS1 9.09 8.4 WiLD MCS1 11.2 5.9
NV2 MCS2 13.82 12.8 WiLD MCS2 16.7 6.8
NV2 MCS3 18.57 18.0 WiLD MCS3 22.7 10
NV2 MCS4 27.9 26.6 WiLD MCS4 34.3 15
NV2 MCS5 37.12 WiLD MCS5 45
NV2 MCS6 41.75 WiLD MCS6 50.8
NV2 MCS7 45.97 WiLD MCS7 55.7
NV2 MCS8 10.4 8.8 WiLD MCS8 12.2 6.3
NV2 MCS9 19.2 17.8 WiLD MCS9 24.1 9.6
NV2 MCS10 27.8 26.4 WiLD MCS10 36.2 15.8
NV2 MCS11 37.0 35.3 WiLD MCS11 48.1 19.6
NV2 MCS12 63.5 52.6 WiLD MCS12 71.6 30.8
NV2 MCS13 77.2 WiLD MCS13 90.2
NV2 MCS14 81.1 WiLD MCS14 96.4
NV2 MCS15 85.2 WiLD MCS15 102.3
0 1020304050 7060 80 90 100
Link load (%)
Latency (ms)
Figure 4: Average delay values obtained for NV2 in experimental
tests at 30 km for different offered load.
roughput (Mbps)
Distance (km)
WiMAX 16QAM 1/2
WiMAX 16QAM 3/4
WiMAX 64QAM 2/4
WiMAX 64QAM 3/4
0 5 10 15 20 25 3530 40 45 50 55 60
Figure 5: 5 ms delay-bounded throughput comparison between
the different technologies obtained from the NS-3 simulator (WiLD
and WiMAX) and from the experimental results (NV2).
Mobile Information Systems 11
shown in Table 7. However, the performance of WiMAX is
always below WiLD and NV2. is is basically due to the fact
that the maximum bandwidth allowed in the IEEE 802.16
standard for WirelessHUMAN is 10 MHz, so that the real
capacity is far below the capacity obtained with IEEE 802.11n
PHY using 20 MHz channels. Although for all distances and
with short frame sizes, WiMAX exhibits the maximum
spectrum efficiency, it is not an interesting alternative for
backhaul networks since its performance can be easily
surpassed by the 20 MHz channels of WiLD and NV2. Only
very particular scenarios where high capacity is not required
but a strong QoS policy is essential, which could benefit of
building WiMAX links. Otherwise, usual QoS requirements
in a backhaul networks can fairly be accomplished using
WiLD or NV2.
In general, links of 45 Mbps can be obtained at distances
up to 20 km with the delay bounded to 5ms with NV2 when
using MIMO and MCS11. is throughput is normally no
longer achievable at longer distances since the link budget
becomes a significant constraint. For 55 km, NV2 can
support up to 20 Mbps. WiLD can only offer 37 Mbps and
10 Mbps at 20 km and 55 km, respectively, with the delay
bounded at 5 ms and conditioned to using frame aggrega-
tion. Finally, WiMAX can only offer 28 Mbps at 20 km and
8 Mbps at 55 km.
7. Example Use Case
7.1. A Backhaul Solution for 3G Femtocells: A Multihop
Network along the Napo River. In order to illustrate that the
technologies assessed can be used in a real context, this
section presents the network dimensioning for the Napo
network use case, one of the scenarios chosen in the
TUCAN3G project as proof of concept [3]. Figure 6 shows
the topology of this network that is currently deployed with
WiLD and NV2 links connecting rural healthcare facilities
with the Iquitos’ hospital. Hence, the network is a chain of
long-distance links in the range of 20–55 km, covering a total
end-to-end distance of 450 km approximately. TUCAN3G is
deploying a parallel demonstration network reusing the
towers from Santa Clotilde to Iquitos, with 3G femtocells in a
few villages. e traffic load potentially generated by the
villages within the first 5 years of use has been calculated and
is shown in Table 8, and compared with the capacity that
WiLD, WiMAX or NV2 could be provided in each hop.
e link budget parameters used in the network are the
ones shown in Table 6, and the sensitivity figures are shown
in Table 5. e Friis propagation model is used for the link
budget, and a fading margin of 20 dB has been used in all
cases for choosing the modulation and coding scheme with
the highest performance in each hop. Also, parabolic dual
polarization antennas with 27 dBi gain were used.
Rural backhaul networks have also different peculiarities
that require special attention. e lack of power grids in
rural areas makes essential the use of alternative power
sources such as solar panels and different mechanisms to
cope with an unreliable power supply. Also, enclosures have
to be minimum 4-4X NEMA type in order to properly
protect the equipment from weather conditions and towers
have to be physically secured. ese issues were solved by the
GTR research group, which have proved experience
deploying and operating rural networks [38].
e results in Table 8 show that both WiLD and NV2
provide largely enough capacity for the backhaul re-
quirements of a 3G provider that aims to provide services in
these villages. On the contrary, WiMAX would be a valid
solution for the four links furthest from the city, but not for
the three closest ones. In linear multihop networks, links
require more capacity as they are closer to the gateway
because they carry the traffic of all the cells that are behind.
e Napo network is currently operating with with NV2
since the end of 2016.
Of course, this is only an illustrative example of the
feasibility of a multihop transport network based on these
technologies, which depends on the specific requirements of
this particular scenario. Together with the Balsapuerto
Network explained in [6, 7], these two proof-of-concept
networks are currently operating with a mix of the refer-
enced technologies supporting 3G femtocells backhaul
traffic from different isolated villages.
7.2. Cost and Operation. Regarding the cost of these tech-
nologies as backhaul, even though 802.11 and 802.16 systems
are inexpensive themselves for operation in nonlicensed
bands, both may require expensive infrastructures for long-
distance links in flat rural areas due to the high towers
required in order to guarantee the LOS. is is particularly
important in nonlicensed bands, where the maximum
transmission power is strictly limited [9]. is may imply a
high CAPEX for some scenarios, while these costs may be
substantially lower in mountainous landscapes in which
towers may be placed in naturally elevated points. On the
other hand, the OPEX is extremely low, which may result in
an overall positive balance for WiFi and WiMAX in most
real scenarios compared to other more expensive alterna-
tives such as VSAT communications. A more detailed
analysis on this can be found in [6].
Regarding the use of nonlicensed bands, their exploi-
tation for a carrier-grade service implies free additional
Figure 6: Schema of the rural Napo network, connecting several
villages to the city of Iquitos using a set of linearly arranged point-
to-point wireless links.
12 Mobile Information Systems
costs, and regulations in many developing countries are
evolving towards permitting operators use them in deprived
areas. Nonlicense bands are more prone to interferences
than licensed bands, even in a rural scenario where RF
emissions are scarce. However, since high-directive anten-
nas are used, the nondesired RF power received in the RX
antenna can be neglected. In case of scenarios located in
countries where licensed bands are mandatory for back-
hauling carried-grade services, WiMAX is the only low-cost
alternative, since it is the only technology of the ones
analysed in this work that has a profile that operates in li-
censed bands. Although the use of licensed bands implies
additional significant costs, an operator can still benefit of
using a low-cost infrastructure based on WiMAX.
8. Conclusions and Future Works
3G providers usually base their infrastructures on carrier-
grade backhaul technologies and macrocells for the access
network, driving to solid professional solutions that result to
be cost effective in densely populated areas. However, re-
mote rural areas require low-cost technologies that permit
the operators to deploy infrastructures in sparsely populated
areas and still ensure economical benefits.
Wireless broadband technologies operating in 5 GHz
nonlicensed bands are undoubtedly cheaper than traditional
carrier-grade backhaul technologies, and the shared spec-
trum may not be an inconvenient in rural regions. However,
there is not common knowledge about the actual perfor-
mance of WiFi, WiMAX, or WiFi-based TDMA solutions in
long-distance links, particularly within backhaul networks,
where the traffic is shaped to keep links working under their
saturation point in order to bound the delay.
is work has shown that these technologies, and spe-
cially 802.11-based TDMA solutions, are a solid solution for
rural 3G backhaul links, providing enough capacity and
acceptable latency as long as external mechanisms limit the
amount of traffic sent through these links [7].
As cited above, standardization bodies such as ETSI
BRAN are following this approach. is paper contributes to
demonstrate its feasibility, specially focusing in finding the
optimum operation point for a link in the curve delay/
offered load. is point is usually not the point where
maximum throughput is achieved but the point that offers
maximum throughput while bounding the delay below 5 ms.
e results presented in this paper can be of interest when
designing such rural backhaul networks and for implement
and operate their traffic control systems.
Data Availability
e source code used to support the findings of this study are
included within the article. e simulation and experiment
data used to support the findings of this study are available
from the corresponding author upon request.
Conflicts of Interest
e authors declare that they have no conflicts of interest.
is work has been performed in the framework of the FP7
project TUCAN3G IST-601102 STP, which is funded by the
European Community. e authors would like to ac-
knowledge the contributions of the colleagues from
TUCAN3G Consortium ( is
research has also been partially supported by the Spanish
MINECO grant no TEC2013-41604-R.
[1] IEEE, “802.11-2012IEEE standard for information tech-
nology exchange between systems local and metropolitan
area networks: Wireless LAN Medium Access Control
(MAC) and Physical Layer (PHY) specifications,” Technical
Report IEEE Std 802.11-2012, IEEE, New York, NY, USA,
[2] IEEE, “802.16-2009—IEEE standard for local and metropol-
itan area networks part 16: air interface for broadband
wireless access systems,” Technical Report, IEEE, New York,
NY, USA, 2009.
[3] A. Martinez-Fernandez, J. Vidal, J. Simo-Reigadas et al., “e
TUCAN3G project: wireless technologies for isolated rural
communities in developing countries based on 3G small cell
deployments,” IEEE Communications Magazine, vol. 54, no. 7,
pp. 36–43, 2016.
[4] ETSI, “Broadband Radio Access Networks (BRAN); broad-
band wireless access and backhauling for remote rural
communities,” Technical Report ETSI TR 103 293, ETSI,
Sophia Antipolis, France, 2015.
[5] T. G. Itu, “1010: end-user multimedia QoS categories,”
Technical Report, ITU, Geneva, Switzerland, 2001.
Table 8: Comparing achievable link capacities with traffic requirements in the Napo network (the acronyms in the links refer to the initials
of the locations (see Figure 6)).
Link Dist. (km) Mbps req. WiMAX WiLD NV2
MCS Mbps MCS Mbps MCS Mbps
SC-TC 39.1 6.4 16 QAM 1/2 20.6 MCS12 17.1 MCS12 52.6
TC-NU 25.5 9.2 16 QAM 3/4 15.5 MCS13 31.2 MCS13 70.6
NU-TP 32.2 12.1 16 QAM 3/4 15.4 MCS12 24.9 MCS12 54,0
TP-HU 26.5 14.9 16 QAM 3/4 15.4 MCS13 30.3 MCS13 70.5
HU-M 22.3 17.7 16 QAM 3/4 15.6 MCS13 33.9 MCS13 70.9
M-PP 19.9 24.5 64 QAM 2/3 21.3 MCS13 38.0 MCS13 71.2
PP-HI 11.7 24.5 64 QAM 2/3 21.7 MCS14 41.3 MCS14 74.9
Mobile Information Systems 13
[6] J. Simo-Reigadas, E. Municio, E. Morgado et al., “Sharing
low-cost wireless infrastructures with telecommunications
operators to bring 3G services to rural communities,”
Computer Networks, vol. 93, pp. 245–259, 2015.
[7] J. Simo-Reigadas, E. Municio, E. Morgado, E. M. Castro,
and A. Martinez, “Sharing low-cost wireless infrastructures
with telecommunications operators for backhauling
3G services in deprived rural areas,” in Proceedings of IEEE
16th International Symposium on a World of Wireless,
Mobile and Multimedia Networks (WoWMoM), pp. 18,
IEEE, Boston, MA, USA, 2015.
[8] K. Chebrolu, B. Raman, and S. Sen, “Long-distance 802.11 b
links: performance measurements and experience,” in Pro-
ceedings of the 12th Annual International Conference on
Mobile Computing and Networking, pp. 74–85, ACM, Los
Angeles, CA, USA, Sepetember 2006.
[9] J. S. Reigadas, A. Martinez-Fernandez, J. Ramos-Lopez, and
J. Seoane-Pascual, “Modeling and optimizing IEEE 802.11
DCF for long-distance links,” IEEE Transactions on Mobile
Computing, vol. 9, no. 6, 2010.
[10] A. Lozano and N. Jindal, “Transmit diversity vs. spatial
multiplexing in modern MIMO systems,” IEEE Trans-
actions on Wireless Communications, vol. 9, no. 1,
pp. 186197, 2010.
[11] F. Bohagen, P. Orten, and G. E. Oien, “Design of optimal
high-rank line-of-sight MIMO channels,” IEEE Transactions
on Wireless Communications, vol. 6, no. 4, pp. 1420–1425,
[12] I. Sarris and A. R. Nix, “Design and performance assessment
of high-capacity MIMO architectures in the presence of a line-
of-sight component,” IEEE Transactions on Vehicular Tech-
nology, vol. 56, no. 4, pp. 2194–2202, 2007.
[13] C. Oestges, V. Erceg, and A. J. Paulraj, “Propagation modeling
of MIMO multipolarized fixed wireless channels,” IEEE
Transactions on Vehicular Technology, vol. 53, no. 3,
pp. 644–654, 2004.
[14] L. Jiang, L. iele, A. Brylka, S. Jaeckel, and V. Jungnickel,
“Polarization characteristics of multiple-input multiple-
output channels,” in Proceedings of IEEE 19th International
Symposium on Personal, Indoor and Mobile Radio Commu-
nications, PIMRC 2008, IEEE, Cannes, France, September
[15] J. M. Vella and S. Zammit, “Performance improvement of
long distance MIMO links using cross polarized antennas,” in
Proceedings of 15th IEEE Mediterranean Electrotechnical
Conference (MELECON 2010), pp. 1287–1292, Valletta, Malta,
April 2010.
[16] B. Raman and K. Chebrolu, “Design and evaluation of a new
MAC protocol for long-distance 802.11 mesh networks,” in
Proceedings of the 11th Annual International Conference on
Mobile Computing and Networking, pp. 156–169, ACM,
Cologne, Germany, August-September 2005.
[17] R. K. Patra, S. Nedevschi, S. Surana, A. Sheth, L. Subramanian,
and E. A. Brewer, “WiLDNet: design and implementation of
high performance WiFi based long distance networks,” NSDI,
vol. 1, p. 1, 2007.
[18] S. Salmer´
on-Ntutumu, J. Sim´
o-Reigadas, and R. Patra,
“Comparison of MAC protocols for 802.11-based long dis-
tance networks,” in Proceedings of Workshop Wireless For
Development (WIRELESS4D), Karlstad, Sweden, December
[19] AirMAX TDM SYSTEM, “Product datasheet,” December 2018,
[20] NV2 Manual,” December 2018,
[21] G. Bianchi and I. Tinnirello, “Remarks on IEEE 802.11 DCF
performance analysis,” IEEE Communications Letters, vol. 9,
no. 8, pp. 765–767, 2005.
[22] I. Tinnirello, G. Bianchi, and Y. Xiao, “Refinements on IEEE
802.11 distributed coordination function modeling ap-
proaches,” IEEE Transactions on Vehicular Technology,
vol. 59, no. 3, pp. 1055–1067, 2010.
[23] F. Daneshgaran, M. Laddomada, F. Mesiti, M. Mondin, and
M. Zanolo, “Saturation throughput analysis of IEEE 802.11 in
the presence of non ideal transmission channel and capture
effects,” IEEE Transactions on Communications, vol. 56, no. 7,
pp. 1178–1188, 2008.
[24] F. Daneshgaran, M. Laddomada, F. Mesiti, and M. Mondin,
“Unsaturated throughput analysis of IEEE 802.11 in presence
of non ideal transmission channel and capture effects,” IEEE
Transactions on Wireless Communications, vol. 7, no. 4,
pp. 1276–1286, 2008.
[25] I. Papapanagiotou, G. S. Paschos, S. A. Kotsopoulos, and
M. Devetsikiotis, “Extension and comparison of QOS-
enabled wi-fi models in the presence of errors,” in Pro-
ceedings of Global Telecommunications Conference,
GLOBECOM’07, pp. 25302535, IEEE, Washington, DC,
USA, December 2007.
[26] D. Senthilkumar and A. Krishnan, “Nonsaturation
throughput enhancement of IEEE 802.11 b distributed co-
ordination function for heterogeneous traffic under noisy
environment,” International Journal of Automation and
Computing, vol. 7, no. 1, pp. 95–104, 2010.
[27] J. Martin, B. Li, W. Pressly, and J. Westall, “WiMAX per-
formance at 4.9 GHz,” in Proceedings of Aerospace Conference,
IEEE, Big Sky, MT, USA, March 2010.
[28] Ubiquity networks, “Datasheet Nanostation M5 5GHz,” De-
cember 2018,
[29] Alvarion, “Datasheet 4Motion 802.16e CPE BreezeMAX Si
4000,” December 2018,
[30] NS3 Network Simulator, December 2018, https://www.
[31] A. Botta, A. Dainotti, and A. Pescap`
e, “A tool for the gen-
eration of realistic network workload for emerging net-
working scenarios,” Computer Networks, vol. 56, no. 15,
pp. 3531–3547, 2012.
[32] R. P. Karrer, I. Matyasovszki, A. Botta, and A. Pescap´
“Magnets-experiences from deploying a joint research-
operational next-generation wireless access network test-
bed,” in Proceedings of 3rd International Conference on
Testbeds and Research Infrastructure for the Development of
Networks and Communities, TridentCom 2007, pp. 1–10,
IEEE, Lake Buena Vista, FL, USA, 2007.
[33] R. P. Karrer, I. Matyasovszki, A. Botta, and A. Pescap´e,
“Experimental evaluation and characterization of the
magnets wireless backbone,” in Proceedings of the 1st In-
ternational Workshop on Wireless Network Testbeds, Ex-
perimental Evaluation & Characterization, pp. 2633,
ACM, 2006.
[34] G. Pei and T. R. Henderson, “Validation of OFDM error rate
model in ns-3,” Boeing Research and Technology, Technical
Report, 2010.
[35] N. Baldo, M. Requena-Esteso, J. N ´uñez-Mart´
ınez et al.,
“Validation of the IEEE 802.11 MAC model in the ns3
14 Mobile Information Systems
simulator using the extreme testbed,” in Proceedings of the 3rd
International ICST Conference on Simulation Tools and
Techniques, p. 64, ICST, Malaga, Spain, March 2010.
[36] S. Rinaldi, P. Ferrari, A. Flammini, F. Gringoli, M. Loda, and
N. Ali, “An application of IEEE 802.11 ac to smart grid au-
tomation based on IEC 61850,” in Proceedings of 42nd Annual
Conference of the IEEE Industrial Electronics Society, IECON
2016, p. 4645, IEEE, Florence, Italy, October 2016.
[37] V. Popovskiy, V. Loshakov, O. Philipenko, A. Martinchuk,
and A. Drif, “Results of development of tropospheric com-
munications system,” in Proceedings of Second International
Scientific-Practical Conference Problems of Infocommunica-
tions Science and Technology (PIC S&T), pp. 193–195, IEEE,
Kharkiv, Ukraine, 2015.
[38] C. Rey-Moreno, I. Bebea-Gonzalez, I. Foche-Perez, R. Quispe-
Tacas, L. Liñ´
an Benitez, and J. Simo-Reigadas, “A telemedi-
cine WiFi network optimized for long distances in the am-
azonian jungle of Peru,” in Proceedings of the 3rd Extreme
Conference on Communication: e Amazon Expedition,
ExtremeCom ’11, pp. 91–96, ACM, New York, NY, USA, 2011.
Mobile Information Systems 15
... For example, if a WiFi link is reaching its saturation point, an increase of 1% in the load can increase the latency up to 20 times without even increasing the offered throughput. 10 Usually, this issue is solved in the edges of the network with traffic engineering and QoS rules. 11 However, additional mechanisms are needed to notify the HNB's and modify their admission control policy (ie, blocking current and future exceeding connections), improving eventually the overall users' experience. ...
... This implies that they do not accurately provide the actual nonsaturate throughput in a link, ie, the throughput offered with bounded. 10 There are also a number of passive methods that provide an accurate measurement of the actual traffic in an interface at almost real time. 31,32 Although these approaches are simple and lightweight, they require extra information about the actual maximum capacity of a given link in order to estimate its actual congestion and how far from the saturation point its current load is. ...
... This is because the link always reaches the saturation point before reaching the maximum actual capacity. 10 Precisely, wireless BH links should always work under the saturation point, since although all available throughput is not used, the delay is kept bounded. * Although our figures are lower than the ones obtained in Rattaro and Belzarena, 33 they allow us to be confident enough to use this measurement congestion mechanism in order to estimate the congestion in our ECN-enabled BH tests. ...
Full-text available
Management in wireless backhaul networks is a challenging task, especially in rural and isolated environments. In these scenarios, the backhaul network usually consists of a set of heterogeneous wireless links that provide limited and variable bandwidth to the access networks, often 3G/4G small cells. Because of the highly constrained nature of this type of backhaul network, intelligent and joint management in both the backhaul network and the access network is crucial in order to avoid performance degradation caused by traffic congestion. In order to avoid the saturation in the backhaul network, access networks should consider the backhaul state when taking decisions in the admission control and scheduling procedures. However, no standardized mechanisms currently exist for sharing management information between both networks. In this work, we propose to use the Explicit Congestion Notification (ECN) bit in the outer IP headers present in the Iuh 3GPP IPSec‐enabled interfaces in order to notify the backhaul congestion state to the access network. We analyze for the referred scenario, compatibility and security details, validating our approach by running numerically simulations and implementing the notification mechanism. Our low complexity approach offers 2% accuracy and backhaul update latency lower than 10 ms during 80% of the time, which makes the solution appropriate for admission control and scheduling intervals in small cells. This work proposes to use the Explicit Congestion Notification (ECN) bit in the outer IP headers present in the Iuh 3GPP IPSec‐enabled interfaces in order to notify backhaul congestion state to 3G/ 4G access networks. The solution is especially designed for wireless backhaul networks in rural areas. Our low complexity approach offers 2% accuracy and backhaul update latency lower than 10 ms during 80% of the time.
Full-text available
In the past years, when wireless network improvement occurs from 1G/2G to third generation (3G), the rate in the use of real-time traffic oriented applications for voice, video and data increases. Consequently, the bandwidth to be backhauled from the cell site to the mobile switching center increases rapidly. 3G network is most prevalent in Nigeria with wide area of coverage. However, in recent times, poor subscribers’ mobile broadband experience is still the major challenge faced by many GSM operators. One of the major causes of this challenge is the use of wrong backhaul for radio access network (RAN). This lead to poor traffic throughput, high packet loss or frame loss at the cell edge. To overcome this challenge, the use of appropriate backhaul technology is crucial. Third Generation Partnership Program (3GPP) recommends the use of either asynchronous transfer mode (ATM) or internet protocol (IP) as the backhaul technologies for its RAN. This paper presents the performance analysis and the comparison of the ATM RAN and IP RAN backhaul technologies using six different 3G sites (with 3G base stations) located at Ado local government area of Ekiti State, Nigeria. The performance of each base station with different backhaul technology was evaluated in term of average maximum throughput per day. The effect of frame loss (for ATM RAN network) and packet loss (for IP RAN network) on traffic throughput were also analyzed. The comparison of the overall result analysis shows that the 3G base stations with IP-based RAN backhaul has better performance than the base station with ATM-based RAN backhaul.
Full-text available
Small rural communities around the world have worse or null access to telecommunication services than populated urban areas. This is due to the disadvantageous relationship between the potential revenues for the operators and the cost of required infrastructures. Nowadays there are interesting low-cost solutions based on small cells for providing 3G access to small communities. However, the backhaul infrastructure to rural areas that are far away from cities can be very expensive. This barrier represents an opportunity for rural communities that already own and manage a multi-hop wireless infrastructure that interconnects them to neighboring communities, the Internet, and/or the PSTN. In this paper we propose the opportunity for 3G operators to share community-deployed wireless infrastructures for their 3G backhauling needs. Infrastructure sharing is a win-win solution because it helps rural communities to achieve financial sustainability for their community networks, while allowing operators to extend their coverage to regions where it would not be economically sustainable otherwise. This paper studies the techno-economic feasibility of the proposed solution through a CAPEX/OPEX analysis. It also identifies the negotiation margin that exists for the community networks that provide backhauling services to mobile network operators. In addition, it provides a general architecture for rural broadband community networks, derived from several particular experiences and from the Quality of Service requirements of 3G traffic. Three techniques are considered to ensure the quality of the backhaul while 3G traffic is transported together with non-3G traffic: DiffServ, plain MPLS and MPLS-TE. Experiments on a laboratory test-bed have allowed us to compare the performance of these three techniques in regards to the advantages and drawbacks of each.
Conference Paper
Full-text available
3G small cells are becoming a promising solution for inexpensive 3G coverage in rural small villages, but the high cost of backhaul infrastructures remains an obstacle for the worldwide adoption of this kind of solutions. On the other hand, broadband wireless infrastructures may be found in many rural areas for communications in specific sectors such as health or education services, environmental monitoring, etc. These sectorial broadband communications infrastructures are often multi-hop wireless networks using long-distance WiFi links or WiMAX. We propose the possibility of 3G operators sharing those third parties' infrastructures. It would permit them to have a low-cost backhaul solution for their rural 3G cells in those areas where the expected demand of 3G services does not ensure enough revenues to justify the deployment of dedicated infrastructures. This is a win-win solution because bandwidth-renting may help to achieve financial stability for such sectorial networks, while operators can extend their coverage to regions where this would not be economically sustainable otherwise. This paper studies the technical feasibility of the proposed solution. Different multi-hop broadband rural networks are studied, and a general architecture is derived from them. Then, additional requirements are introduced in order to make such networks support the QoS requirements of 3G voice and data traffic. Three techniques are considered in order to manage the network in such a way that 3G traffic and other traffics may be transported together in the network without compromising the quality of the backhaul: DiffServ, plain MPLS and MPLS-TE. Measurements taken on a real testbed in a laboratory allow us to compare the performance achieved with these three techniques, and the advantages and drawbacks of each of them are presented.
Conference Paper
Full-text available
Isolated rural areas of developing countries often lack of any kind of communications infrastructure. There are some WiFi for Long Distances (WiLD) networks that have been deployed successfully in forests and mountainous regions of countries such as India, Colombia and Peru, for providing Voice over IP and Internet support to health centers. In this work a real telemedicine WiFi network deployed in the Amazonian jungle of Peru is introduced. Technical and social considerations taken into account and devices used for its deployment in such a difficult context are described together with the services provided. Furthermore, performance of this multihop network is analyzed in order to provide a better insight of its behavior. As a conclusion, we describe future and present works related to the improvement of the network.
Full-text available
WiFi based Long Distance (WiLD) networks have been successfully deployed in rural areas of developing countries, thus providing Internet access and VoIP telephony services to previously isolated regions. However, it is well known that the standard CSMA/CA MAC protocol used by WiFi is not well suited for long distances. To achieve better performance, some researchers have proposed replacing the standard CSMA/CA MAC with a TDMA-based MAC, while others have tried to optimize timing parameters in the IEEE 802.11e EDCA standards. This paper is an experimental comparison of the two approaches for point-to-point links. We measure the throughput and delay in various test scenarios by varying the distance of point-to-point links in a controlled environment created using a wireless channel emulator. As far as we know, this is the first experimental comparison between a TDMA-based MAC solution over 802.11 with the IEEE 802.11e EDCA standard optimized for long distances. Finally, we illustrate the tradeoffs between using these two approaches under different conditions.
Full-text available
We validate the IEEE 802.11 MAC layer model in ns3 by means of measurements on the EXTREME testbed. We consider different scenarios: communications within a single pair of nodes, multi-user communications using either VoIP or saturated traffic, and communications in the presence of hidden nodes. For each scenario we describe in detail our testbed and simulation setup, and compare the results pro-vided by the ns3 simulator with the performance measured on the testbed.
Full-text available
In this paper, we propose a mechanism named modified backoff (MB) mechanism to decrease the channel idle time in IEEE 802.11 distributed coordination function (DCF). In the noisy channel, when signal-to-noise ratio (SNR) is low, applying this mechanism in DCF greatly improves the throughput and lowers the channel idle time. This paper presents an analytical model for the performance study of IEEE 802.11 MB-DCF for nonsaturated heterogeneous traffic in the presence of transmission errors. First, we introduce the MB-DCF and compare its performance to IEEE 802.11 DCF with binary exponential backoff (BEB). The IEEE 802.11 DCF with BEB mechanism suffers from more channel idle time under low SNR. The MB-DCF ensures high throughput and low packet delay by reducing the channel idle time under the low traffic in the network. However, to the best of the authors’ knowledge, there are no previous works that enhance the performance of the DCF under imperfect wireless channel. We show through analysis that the proposed mechanism greatly outperforms the original IEEE 802.11 DCF in the imperfect channel condition. The effectiveness of physical and link layer parameters on throughput performance is explored. We also present a throughput investigation of the heterogeneous traffic for different radio conditions. KeywordsIEEE 802.11-medium access control (MAC)-distributed coordination function (DCF)-nonsaturated traffic-heterogeneous traffic-signal-to-noise ratio (SNR)-binary exponential backoff (BEB)
Conference Paper
Smart Grids deployment is growing around the world. The expansion of the existing Smart Grid network infrastructure in order to connect new electrical substations (or end-users) may be impaired by difficult cable deployment. For this reason, recently, wireless communication links have also been used. This paper deals with the deployment, and the first characterization, of an IEEE 802.11ac connection between IEC 61850 electrical automation devices. Some configurations of IEEE 802.11ac have been tested and compared with usual cable connection. The use of GPS based instrumentation combined with IEEE 1588 clock synchronization allows reducing measurement uncertainty and producing meaningful results. For instance, the IEC 61850 Transfer Time between GOOSE (Generic Object Oriented Substation Event) publisher and subscriber over an IEEE 802.11ac wireless link, loaded with 80Mbit/s of traffic, has an average value of 5.4 ms and a maximum value of 10 ms.
Recent years have witnessed a massive penetration of cellular systems in developing countries. However, isolated rural areas (sparsely inhabited by low-income population) have been disregarded because classical access and backhaul technologies do not ensure the return on investment. This article presents innovative techno- economical solutions to provide these areas with cellular voice and data services. We first analyze the general characteristics of isolated rural communities, and based on this information, low-cost solutions are designed for both access (using 3G access points) and backhaul networks (using non-carrier grade equipment as WiFi for long distances or WiMAX in non-licensed bands). Subsequently, a study of population- dependent income vs. costs is presented, and a new business model is proposed involving mobile network operators, rural operators, and infrastructure providers. In order to test these solutions, we have built two demonstration platforms in the Peruvian jungle that have allowed validation of the technical feasibility of the solution, verifying the business model assumptions and the scalability of the initiative.
Internet workload is a mix of many and complex sources. Therefore, its accurate and realistic replication is a difficult and challenging task. Such difficulties are exacerbated by the multidimensional heterogeneity and scale of the current Internet combined with its constant evolution. The study and generation of network workload is a moving target, both in terms of actors (devices, access networks, protocols, applications, services) and in terms of case studies (the interest expands from performance analysis to topics like network neutrality and security). In order to keep up with the new questions that arise and with the consequent new technical challenges, networking research needs to continuously update its tools. In this paper, we describe the main properties that a network workload generator should have today, and we present a tool for the generation of realistic network workload that can be used for the study of emerging networking scenarios. In particular, we discuss (i) how it tackles the main issues challenging the representative replication of network workload, and (ii) our design choices and its advanced features that make it suitable to analyze complex and emerging network scenarios. To highlight how our tool advances the state-of-the-art, we finally report some experimental results related to the study of hot topics like (a) broadband Internet performance and network neutrality violations; (b) RFC-based security and performance assessment of home network devices; (c) performance analysis of multimedia communications.