Sei sulla pagina 1di 8

Radio Resource Allocation in Urban Femto-WiFi

Convergence Scenarios
Paul Fuxjger, Hans Ronald Fischer, Ivan Gojmerac, Peter Reichl
FTW Telecommunications Research Center Vienna
Donau-City-Str. 1, 1220 Vienna, Austria
{fuxjaeger, fischer, gojmerac, reichl}@ftw.at
AbstractThe recent tremendous growth of mobile broadband
traffic demand forces mobile operators to find ways for offloading
their macro-infrastructure. This can be achieved by a combination of two concepts, i.e., introducing/deploying a finer grid of
base-stations (femto-cells) together with increasing the share of
traffic volume that is handled by communication-links operating
in unlicensed bands (e.g., IEEE802.11-based wireless-LAN). Both
concepts are supported by the recent proliferation of wireline
broadband (DSL/cable) which enables the cost effective backhauling of small sites using standard internet access. However,
new challenges in terms of network-planning arise as these small
cells are deployed randomly, their resulting spatial density in
urban environments may get very high, and thus the aspect of
interference-mitigation is of vital importance.
The contribution of this paper is twofold: first we provide a
comparison of licensed-band femto-cell versus unlicensed WiFi
technologies. We try to outline characteristic differences and
respective limitations in terms of efficiency of radio-resource
sharing mechanisms, with a specific focus on macro-cell offloading scenarios. Second, we present the results of a measurement
campaign that allows to quantify the current density of WiFi
access points in the 2.4GHz WiFi band in urban environments.
We subsequently use these data sets to extrapolate the coverage
ratios that are achievable by aggregating low-power WiFi/femto
cells. This can be used to provide novel insights into specific radioresource management problems which a mobile operator needs
to solve when moving towards smaller cells and less-controllable
deployment procedures.

I. I NTRODUCTION
Since the early 1990s, there has been a tremendous growth
of the number of cellular phone users, which can mainly
be attributed to the successful roll-out of GSM networks
in Europe and Asia [3]. At the same time, the usage of
the Internet experienced broad adoption among the general
population, resulting in an explosion of the global amount of
data traffic carried via telco infrastructures [4]. Until the end
of that decade, the two emerging worlds of cellular telephony
networks and fixed-line internet had been coexisting in parallel, without major points of contact, with few exceptions such
as the Wireless Application Protocol (WAP) [5].
This situation started to change with the advent of the
Universal Mobile Terrestrial System (UMTS) technology, i.e.,
the 3rd generation of cellular networks, whose broad rollout started around 2002 in Europe [6]. Back then, the persector data rates offered by the first deployed UMTS release
1 AWARE is a COMET project which has been running since January 2010
as part of a CELTIC activity in close collaboration with Mobilkom Austria,
Alcatel-Lucent Austria, Alcatel-Lucent Bell Labs France and the University
of Versailles, Laboratoire PRiSM. See http://www.ngnlab.at/aware for further
details.

(Rel. 99) were quite limited and far from being sufficient for
shared broadband connectivity [7]. Nevertheless, this can be
considered the first major convergence step towards an AllIP cellular network [2]. Due to the fact that data rates (in
UMTS Release 99) soon proved to be insufficient for mass
data service deployment in mobile networks, emphasis has
been put on the further development of the evolution of UMTS
technology, which has eventually led towards the specification
of High Speed Uplink and Downlink Packet Access (HSPA)
in 3GPP Release 5 and 6 [9], [10]. HSPA deployment allowed
mobile operators to offer mobile internet connectivity to their
users with significantly higher bit-rates, which is generally
known as mobile broadband [11] and already starting to
experience strong growth in some countries [1].
However, the convenience and reliability of mobile broadband is expected to soon lead to an explosion of the observed
overall data rates (i.e., network utilization), resulting in severe
congestion especially in metropolitan areas. This phenomenon
has particularly been exacerbated by the often fierce market
competition which forces operators to offer mobile broadband
connectivity at very low prices, bringing them into direct
competition with fixed line DSL or cable Internet Service
Providers (ISP) [1].
Thus, the operators are faced with two urgent points of
action: on one hand, they need to ensure the profitability of
their connectivity offerings, on the other hand they also have
to make sure that their networks offer reasonable quality of
experience levels [2]. As neither future releases of UMTS
(e.g., HSPA+ [12]) nor Long Term Evolution (LTE) as the
next generation of mobile network technologies are likely
to fundamentally change this general setting, we expect that
the out-lined problems will lead to substantial changes of
traditional network planning paradigms.
Therefore, the recently started project AWARE 1 (Aggregation of Wireless Access Resources) focuses on the investigation of novel strategies for network planning and operations,
aiming at the offloading of the macro-cell network via wireless
access points to be deployed at the customer premises, relying
either on IEEE802.11 (WiFi) or the upcoming femto-cell
technologies. The first goal within the project was to have
an exhaustive state-of-the-art overview and to systematically
identify strengths and weaknesses of both technologies.
This paper documents the current state of knowledge within
AWARE and it provides an initial outline of several fixedmobile convergence scenarios, leading to a bundle of open
issues concerning the feasibility of nomadic and/or mobile
usage of small-cell networks. As a second contribution we

report on a recent coverage-study on the current extent of


unmanaged personal WiFi deployment in Vienna, Austria,
which may serve as a typical case of a major city in central
Europe, and draw some initial conclusions.
The remainder of the paper is structured as follows: In
section II we briefly review the state of the art in terms
of technologies and current trials. Section III provides a
general discussion of problem dimensions and outlines two key
scenarios for macro-cell offloading. In section IV, we address
basic offloading feasibility aspects for nomadic and mobile
users by presenting detailed statistical results about current
WiFi coverage in urban Vienna. Section V concludes the paper
with a summary and an outlook on future work.
II. M ACRO -C ELL O FFLOAD : S TATE OF THE A RT,
T ECHNOLOGIES AND C URRENT T RIALS
There are basically two triggers for the deployment of
femto-cells. On one hand, in countries which suffer from
poor in-door coverage increasing signal quality is considered
the main factor for the deployment of femto-cells [13]. On
the other hand, the significant growth of mobile broadband
traffic [14] and the resulting impact on the macro-cell network
infrastructure causes network operators to offload traffic via
femto-cells deployed at customer premises [26].
In general, there is a twofold motivation behind using
customer premises-based access points for macro-cell offload:
Firstly, a dense deployment of low power wireless access
points has the potential to greatly increase the overall network
capacity. Secondly, this extra capacity achieved by a multitude
of low power access points comes at a minimum cost for
the mobile network operator, as the access points backhaul
in most operational models is provided by the customers
themselves, and as the cost of the access points is expected
to be easily affordable, i.e., in the order of current consumergrade WiFi access points [15].
There already exists a multitude of commercial femto-cell
offerings by a number of international equipment vendors.
Whereas some traditional equipment vendors only act as
resellers of specific technologies in this context [16], other,
mostly smaller vendors develop the corresponding solutions
in-house. The most well known manufactures include Airvana, Thomson, RadioFrame Networks, Ubiquisys, Netgear
and Pirelli [17]. The products offered by these companies
in most cases already take into account the technological
challenges which become relevant in large-scale deployments,
e.g., radio interface mitigation, macro-femto handovers, security, scalability (especially with respect to the core network),
access control (closed group vs. open group), and femtocell activation and management (while issues like plug and
play and operator-customer access point co-management still
represent open issues) [18].
Complementary to the growing number of existing femtocell solutions for 3G technologies in recent years, substantial
research efforts have been made addressing the remaining
unresolved problems. Currently, the research community is
mainly focusing on the following issues: co-channel interference (due to the use of identical frequencies by femto-cells and

macro-cells) [19], auto-configuration mechanisms [20], as well


as general radio resource management [21] and performance
optimization [22] problems.
These ongoing research and development efforts are complemented by a number of technology field trials that have
been recently initiated jointly by network equipment vendors
and operators at different levels of scale and with different
techno-economical objectives in mind [23]. The most important deployment at the moment is performed by Sprint which
has been started under the name "Sprint Airave" in 2007
as a limited rollout (Denver and Indianapolis) of a homebased femto-cell built by Samsung. Mid of 2008, Airave has
been rolled out on a nationwide basis. In November 2008,
Starhub rolled out its first nation-wide commercial 3G femtocell services. In January of 2009 Verizon rolled out its femtocell network extender, it is based on the same design as
the Sprint/Samsung system. In July of 2009 Vodafone also
released a femto-cell network extender.
Additionally, a number of operators have run field trials
in 2008 and 2009, including O2, Softbank, TeliaSonera, and
Vodafone. AT&T has started femto-cell testing as a limited
roll-out in Raleigh and Charlotte under the name "3G MicroCell" as of 28 October 2009, based on equipment from Cisco
Systems. In Central Europe, mobilkom austria has been the
first to launch a pilot project for femto-routers with 35 testhouseholds in 2009 [24].
III. S CENARIO A NALYSIS
Due to the significant differences in the current deployment
state of the different mobile operators, it is unlikely to be able
to devise a "one-fits-all" solution. This section will discuss
several operational scenarios which seem to be relevant both
from a technological and business point of view. The following
fundamental problem fields need to be considered:
User mobility level: we distinguish between three major
user groups, i.e., stationary home users who use mobile
broadband mostly at their private premises, nomadic users
using mobile broadband frequently at different locations
in a quasi-stationary manner, and truly mobile users who
use mobile internet on the go.
Spectrum availability: depending on the instantaneous
level of radio resource utilization, technology might
enable users to automatically make use of different
parts/types of spectrum, e.g., either the ISM (industrial,
scientific and medical) band or the 3G band, while
participating in macro-cell offloading.
Backhaul characteristics: the relevant criteria concerning the backhaul include sufficient up- and downstream
bandwidth, low delay and jitter, Quality-of-Service (QoS)
mechanisms at the provider edge, high connection availability, and level of control by the mobile operator.
Customer/operator co-management: as currently users
operate their own LANs also as home networking infrastructure (a trend which will even increase e.g., due to
network attached storage (NAS) for data backup and
media streaming), it is virtually unthinkable for them to
give up functionalities and control of their local network,

which therefore mandates strategies for customer/operator


co-management of the deployed access points.
Type of service: while the voice-only case has already
been addressed extensively by the Unlicensed Mobile
Access (UMA) / Generic Access Network (GAN) 3GPP
[26] initiative, the more complex case is represented by
the offloading of voice and data which imposes different
constraints on the target system. The data-only case again
is considered less complex.
Security, authentication and integrity: depending on
the type of scenario addressed and technology employed,
security, integrity and authentication will represent important issues: in the WiFi case, ensuring stringent authentication and security of transmission are open problems. On
the other hand, for femto solutions protecting the integrity
of the device is a key aspect.
Business models: there are three main constraint dimensions: (1) current deployment status with respect to
coverage and capacity, (2) relationship between mobile
operator and associated fixed-line ISP (which is sometimes even the owner of the mobile operator) or viceversa, and (3) competitive situation on the market.
Operator vs. user incentives: as already mentioned
earlier, macro-cell offloading is expected to allow for
a significant increase in network capacity at low cost,
whereas the benefits for the user include better Quality
of Experience together with potential tariff reductions
and, last but not least, the much better performance of
WiFi/femto access points in terms of electromagnetic
compatibility (contrary to public opinion). Hence, in
general there is potential to achieve a win-win situation
for users and providers alike.
The following subsections present a brief overview of two
elementary ways to introduce macro-cell offloading. Also, the
relevant aspects for two distinct deployment variants (WiFi
and femto) are highlighted and compared:
Scenario 1: Macro-Cell Offload for Stationary Users
Description: The goal of the mobile operator is to offload
users from the macro-cell network who mainly use mobile
broadband within their home premises (as a replacement of
fixed-line Internet). Due to the expensive nature of macrocell Internet provisioning, there is a natural incentive to use
potentially available fixed-line data backhauling to the user
homes. In this scenario, the traffic load generated by nomadic
and truly mobile users (which is not considered as critical)
will still be handled by the macro-cell infrastructure.
Prerequisites: The user needs an appropriate backhaul into
the home as well as a wireless access-point. Provisioning a
(probably cross-financed) backhaul might provide an interesting business case especially for mobile operators with a close
business relationship to a fixed-line ISP.
Deployment variants:
WiFi: Assuming that there is sufficient capacity on the
ISM band available, all traffic can be offloaded via
WiFi access points which are assigned to a specified
closed group of users (typically a household etc.). An

additional functional distinction can be made between


the access-point acting as an infrastructure device within
the user premises (locally administrated by the user)
versus a designated offloading device (not configureable
by the user). If the 3G data modem does not support
WiFi, an additional WiFi-enabled data card has to be
provided. However, deploying WiFi devices in large
numbers (e.g., one per household in densely populated
buildings) will likely generate interference problems since
the flat management hierarchy that is being adopted by
the IEEE802.11 standard family lacks a robust method to
efficiently share the common radio resource (e.g. no coordinated channel assignment, no stringent transmission
power control, no frequency division duplex mode).
Femto: The association between the data modem and the
deployed femto-cell (realized via the SIM card and the
cell ID of the femto-cell) triggers the usage of the femtocell instead of the macro-cell, and all data traffic will be
routed in this way. As in the WiFi case, dense deployment
easily leads to severe interference problems between the
femto-cell and the macro-cell network as well as between
individual femto-cells. But the hierarchical structure of
the underlying 3GPP protocol suite offers systematic
ways to apply efficient radio-resource management algorithms, centrally controlled by the operator via the so
called Femto Management System (FMS).

Scenario 2: Macro-Cell Offload for Nomadic/Mobile Users


Description: This case corresponds to the scenario when
due to the large number of active users and their growing
demand for bandwidth, the resulting overall volume of mobile
Internet usage cannot be satisfied by the existing macrocell infrastructure (not even by upcoming LTE-and-beyond
technologies). Therefore, the mobile operator aims at relieving
its macro-cell network at specific locations of high user density
and thus allows for nomadic and/or mobile usage of wireless
access points. In this scenario, the user is not confined to
her private premises but moving around in a quasi-stationary
(nomadic) or even truly mobile manner.
Prerequisites: The main prerequisite is the dense deployment of wireless access points (e.g., femto/WiFi) at crowded
sites of still manageable size, which are attracting huge
user densities (for instance: cafes, restaurants, lecture halls,
cinemas, school rooms, public or office spaces). The need for
backhauling infrastructure is similar to Scenario 1, however,
based on the traffic volume created by a potentially larger
group of users within a single femto/WiFi-cell, the provided
backhaul-bandwidth has to be usually much higher.
Deployment variants:
WiFi: As in Scenario 1, sufficient availability of unused ISM spectrum has to be assumed. Moreover, the
employment of stringent authentication mechanisms is
of central importance. Free and unencrypted WiFi is
attractive for owners of public premises (like shops,
restaurants etc.), however it is not an interesting solution
from an operators point of view due to the lack of access
control which inhibits profitable business cases. As far

as mobility is concerned, special care has to be taken to


achieve efficient seamless or even soft handover.
Femto: In this case, the femto-cell has to operate in
an open-group mode, which allows access for all users
covered by the cell. Especially for mobile users, the
femto solution provides the advantage of integrated handover mechanisms for all cases (femto-to-femto, femtoto-macro, macro-to-femto). Specific attention may have to
be put on differentiating traffic the owner of the premises
where the femto-cell is located from traffic originating
from third party users. Another relevant issue concerns
access point co-management, which allows transferring
part of the responsibility for the device to the user.

Potential Extensions to Scenario 1 and 2


WiFi over Licensed Band: High density of WiFi access
points in addition to heavy user-generated WiFi traffic volume
leads to the complete saturation of the ISM spectrum which
for the users is therefore no longer accessible with acceptable
quality. This prevents the mobile operator from offloading
the macro-cells via the unlicensed spectrum. The mobile
operator is interested in a solution without major impact to
core network dimensioning. As a prerequisite Wifi chipsets
are not only supporting the ISM band, but also can operate
in other parts of the spectrum (e.g., within the licensed 2.1
GHz bands). We propose to use the WiFi stack in the licensed
band owned by the mobile operator who therefore can take
care of related radio network planning issues like, e.g., the
distribution of access points in certain areas. Using WiFi
in the licensed spectrum is attractive as it leads to benefits
because of the simplicity of this protocol stack. But it also
leaves some complex problems unsolved: the coordination
between different access points due to the lack of interference
mitigation, and the general friction originating from the use
of a foreign protocol in the licensed band.
Local Breakout: With mobile broadband based on 3GPP
specifications, Internet traffic is routed via the Gateway GPRS
Support Node (GGSN). However, due to huge traffic demand
in the core network, central components like the GGSN
become bottlenecks of the system, and the mobile operator
is interested in offloading the traffic as early as possible in the
network (also known as local breakout). Local breakout means
that a functional block in the access point itself is able to
decide whether part of the traffic has to pass the core network
under all circumstances (e.g., for charging purposes). If this
is the case, the access point forwards this part of the traffic
through an encrypted tunnel directly to the operators core
network. All the remaining other traffic is offloaded directly
to the Internet via the backhaul of the access point. Note that
this provides a relatively simple solution for off-loading mass
amounts of traffic.
Device-Level Convergence: WiFi access point and femtocell base-station are physically and logically integrated. The
mobile operator aims at jointly managing the scarce radio
resources in an efficient manner. This is especially interesting
if the ISM band availability changes dynamically. In this
case, the decision for licensed versus unlicensed band as

well as for the employed technology (WiFi vs 3GPP stack)


is made ad-hoc and transparent to the user, based on the
instantaneous availability of sufficient radio resources in the
respective bands. The modem has to be capable to run both
the 3GPP stack and the WiFi stack in parallel. In addition, it
must be capable to operate both in the ISM and licensed band.
The device-level convergence approach is perfectly capable
of handling legacy functionality (e.g., support for closed
local area networks) and enables user-transparent sharing of
traffic between licensed and unlicensed bands. To maximize
performance and robustness both protocol stacks need to be
merged, or at least closely coordinated. Ideally, functional
blocks in the legacy part of the stack (e.g., frequency
and power control in IEEE802.11) could be controlled
by exploiting the more hierarchical network structure of
UMTS/LTE.
Summarizing, the above scenarios and their extensions
highlight a broad spectrum of relevant cases for macro-cell
offloading. From them it becomes apparent that the spatial
distribution of wireless access points is of key importance for
increasing offloading efficiency. However, for the femto case
it turns out to be almost impossible to gather related data due
to the current lack of large-scale femto-cell deployments. On
the other hand, it is interesting to observe that both WiFi and
femto technologies operate on frequencies (2.4 GHz versus
2.1 GHz) with comparable path-loss and radio-propagation
properties (which, by the way, partly applies to the 2.6 GHz
band of LTE as well). At the same time, also the transmission
powers employed by WiFi and femto access points are of the
same order of magnitude. Thus, the current spatial distribution
and radio-propagation properties of WiFi access points may
offer precious information to be used for gaining additional
insight into the femto case.
Therefore, in the next section of the paper we will report on
a recent war-walking campaign in Vienna, tracking in detail
the WiFi-coverage and signal-strength originating from inhouse access points as measured on the street. These results
will be helpful to find an answer to the question of whether
mobile or at least nomadic usage of mobile broadband based
on a small-cell infrastructure is feasible under currently given
deployment constraints. Secondly, our results also indicate
that there are upcoming potential femto-macro interference
problems if both systems are operated in the same frequency
band.
IV. E XPLORING T HE F EASIBILITY OF M ACRO -C ELL
O FFLOADING IN U RBAN E NVIRONMENTS
Our war-walking campaign has been performed on March
18, 2010, in the 2nd district of Vienna which can be characterized as a typical densely populated urban area with buildings
erected mainly in the early 20th century (exterior wall thickness typically 50-60 cm, average building height of 3-4 levels,
i.e., approx. 15 m) and an estimated population density of
10,000 inhabitants per square kilometer. During three hours
each existing street in this area has been walked along with
an average speed of around 1.2 m/s, resulting in a total routelength of approximately 10 km. For a satellite view of the

0.07

0.06

PMF of Peak Signal Strength

0.05

0.04

0.03

0.02

0.01

0
100

Figure 1. Resulting peak-signal-strength positions with GPS-based longitude


and latitude information. In the following analysis only entries in the yellow
square (area of 0.25 km2 ) will be considered.

Figure 2.
values.

95

90

85

80
75
70
Peak Signal Strength [dBm]

65

60

55

50

The probability mass function of recorded peak-signal-strength

0.07

0.06

0.05

PMF of Range

region (taken from the current Google Earth database), we


refer to Figure 1.
Apparently, the outcome of such a measurement campaign
depends significantly on the receiver-sensitivity. In order to
achieve results which can be considered representative for
mobile broadband hardware actually in use, we have chosen
an iPhone 3GS as measurement device. Due to its popularity,
we consider it a good candidate for equipment which will
be used by a large community in the future. We are aware
of the fact that using measurement devices with improved
reception capabilities (e.g., larger and/or external antennas)
will lead to different results, however, in our context we
consider the usage of a "real-life device" to be crucial for
drawing practically relevant conclusions. During the entire
measurement period, an application called WiFi-Where was
active which is able to detect the signal-strength of WiFi
beacon frames. All detection events have been tagged with
geographic coordinates provided by the internal GPS receiver
of the iPhone. In contrast to similar measurement campaigns
which only record the geographic coordinates of the first
detection event related to an individual access point, our
approach was to continuously evaluate the varying signal
strengths and use the coordinates of peak signal strength
which allows a much more precise estimation of the actual
location of the WiFi device. On top of that, the tracked data
allows to derive an approximation for coverage ranges of the
individual SSIDs based on the geographical distance between
first and last detection coordinates. For further technical details
of the measurements, please refer to Table I. In Figure 1, the
geographic locations of peak signal strengths are marked as red
pins. In order to avoid inaccuracies at the edges of our route,
the following statistical evaluations refer to the area marked
by the yellow square which covers a total of 0.25 km2 . In
this area, we have detected a total number of 1166 unique
access points, which corresponds to an average of one device
per 214 m2 . From a practical point of view, this is equivalent
to detecting roughly one new access point every 5 m along
the entire walking route.
Figure 2 depicts the statistical distribution (probability mass

0.04

0.03

0.02

0.01

10

20

30

40

50

60
70
Range [m]

80

90

100

110

Figure 3. The probability mass function of coverage range values which are
based on the geographical distance between first and last detection coordinates.

function) of the peak-signal-strength values (in dBm) as they


have been provided by the iPhoneOS-specific API. Observe
that the sensitivity-threshold is around -100 dBm, whereas the
majority of the signal strengths is to be found in the range
between -90 dBm and -70 dBm (median -79 dBm).
It is interesting to observe that there is no direct correlation between the individual peak-signal-strengths and the
corresponding coverage ranges estimated from the continuous
SSID tracking as described above. Figure 4 demonstrates this
unexpected result by depicting (Range, Peak-signal-strength)
pairs which turn out to be quite evenly distributed instead of
being concentrated along a straight line. This indicates that
there is only a weak relation between the strength of the
signal of individual access points and their coverage range.
It is therefore not justified to assume a simple (e.g., exponentbased) path-loss model. For the corresponding histogram of
coverage range estimates, we refer to Figure 3.
Figure 5 provides a detailed analysis of the coverage properties within the area under investigation. Figure 5(a) depicts
the measured peak location points, which in a second step
are related to the corresponding range estimates and thus
lead to the aggregated coverage map of Figure 5(b). Here,
color saturation is used as an indicator for the number of

100

100

Latitudinal Position [m]

Latitudinal Position [m]

200

300

100

200
300
Longitudinal Position [m]

400

500

500

(a) Measured peak location points in the area of interest. Note that each
marked position corresponds to at least one detected peak point - for
most measurement points multiple peaks coincide at one point due to
the long sampling interval of 4 seconds.
1

100

100

200

300

400

500

100

200
300
Longitudinal Position [m]

400

500

(b) The pixels corresponding to public street-space are colored in blue


if at least one WiFi beacon is received at that point.

Latitudinal Position [m]

Latitudinal Position [m]

300

400

400

500

200

200

300

400

100

200
300
Longitudinal Position [m]

400

500

(c) This figure is similarly generated as 5(b), but here only WiFi devices
that correspond to customers of fixed-line ISP Telekom Austria are
considered.

500

100

200
300
Longitudinal Position [m]

400

500

(d) Here only WiFi devices that correspond to customers of the cable-ISP
provider UPC Austria are considered.

Figure 5. These are the resulting coverage maps, based on estimated range and position information of individual access points. In this analysis a spot is
said to be in coverage-range if at least one WiFi beacon can be received there. One can see that in the considered area of Vienna, the WiFi coverage ratio is
100 percent if all access points were collectively aggregated (see subfigure 5(b))

55

0.9

60

0.8

65

0.7

70

0.6

Coverage ratio

Peak Signal Strength [dBm]

50

75

80

0.5

0.4

85

0.3

90

0.2

95

0.1

100

20

40

60
Range [m]

80

100

120

Figure 4.
This figure demonstrates that peak-signal-strength values and
corresponding range values are not significantly correlated.
Table I
M EASUREMENT PARAMETERS
Measurement Device
WiFi chipset
WiFi antenna
Measurement time
Application used for GPS-tagging
Update interval of SSID-list
Average walking speed
Type of frames
Typical beacon format
MPDU length
Typical rate
Frequency band
Channel numbers

iPhone 3GS (FW V3.1.2)


Broadcom BC4325
Internal
March 18, 2010, 5pm-8pm
WiFi-Where
4 seconds
1.2 m/s
IEEE802.11 beacons
DSSS-BPSK 1Mbit/s
92-104 Bytes
10 Frames per sec
2.412 2.472 GHz
1-13

individual WiFi devices seen simultaneously at the respective


coordinates. From this figure, we may conclude that the
coverage of this area is exhaustive, i.e., 100%.
In a next step, we have filtered our measurement data
with respect to similarities in the SSID strings in the beacon
frames. As a result, we have been able to identify two major
consistent subgroups which can be explicitly assigned to two
different ISPs (Telekom Austria with a total share of 18%
and UPC Austria with a total share of 24%). Figures 5(d)
and 5(c) depict the resulting ISP-specific coverage maps. Note
that exhaustive coverage is not reached any longer; in total,
the Telekom Austria coverage equals 98% whereas the UPC
Austria coverage equals 96%.
Finally, we would like to address the relationship between
the conversion ratio and the coverage ratio as an initial
estimate of the effort for the deployment of offloading infrastructures. To this end, the following random experiment
has been performed: take all access points operated by one
(or both) of these operators and replace randomly a certain
ratio by offload-enabled equipment. This leads to resulting
coverages ratios as depicted in Figure 6. It is interesting to
observe that the initial slope of these graphs is rather steep,
for instance in the combined case it would be sufficient to
update 20% of the equipment to achieve a coverage ratio of
almost 90%.
Summarizing our measurement results, we observe already

Telekom Austria
UPC Austria
Combined
0

0.1

0.2

0.3

0.4

0.5
0.6
Conversion ratio

0.7

0.8

0.9

Figure 6.
Relation between conversion ratio and coverage ratio. Both
operators would individually be able to cover 90 percent of the street area
just by converting half of their CPE equipment to support roaming. If both
operators collaborate in this effort, a coverage of 100 percent would be feasible
by converting only half of the device-base.

today an impressive density in terms of access points per km2


which indicates a considerable overcrowding of the 2.4GHz
ISM band. In general, the signal strengths in urban canyons
are higher than we originally expected, and also the aggregated
percentage of access points operated by the two major local
ISPs (Telekom Austria and UPC Austria) turns out to be larger
than assumed.
V. S UMMARY AND F UTURE W ORK
The enormous growth of mobile broadband usage in Europe
over the last years has driven mobile operators to continuously
scale up their macro-cell network infrastructure. While the
introduction of the latest high capacity 3G technology releases
like HSPA has up until now allowed operators to retain
satisfactory levels of QoS, the still-growing interest in mobile
broadband (further fueled by falling end-user pricing levels) is
very likely to soon cause capacity bottlenecks in urban areas,
in spite of the introduction of novel technologies like, e.g.,
LTE. Based on these observations, operators have recently
begun looking into small-cell network technologies which may
complement their networks in high user-density areas and in
this way offload the traffic from the macro-cell infrastructure.
In Section II of this paper we especially focus on the stateof-the-art in femto technologies as the latest candidate for
macro-cell offloading, followed by an overview of the current
femto technology-trials conducted jointly by the operators
and the network equipment vendors. Addressing the fact that
mobile network deployments of the individual operators are
likely to necessitate different strategies, Section III outlines a
broad spectrum of scenarios for macro-cell offloading based
on WiFi and femto technologies, taking into account a number
of identified problem fields and constraints. Motivated by the
apparent insight that the spatial distribution of wireless access
points is of key importance for increasing offloading efficiency,
Section IV explores the feasibility of nomadic and/or mobile
macro-cell offloading in urban environments based on the
measurements of current user-controlled WiFi coverage in

densely populated areas of Vienna, as we believe that it


represents a good extrapolation of future operator-deployed,
home-based femto-sites due to the similar path-loss properties
of the operational frequency bands as well as the comparable
transmission powers used.
Our measurement results show that the high number of
the deployed WiFi access points attributable to two major ISPs could offer huge potential for full outdoor smallcell signal coverage, assuming a reasonable rate of WiFi
device-conversion from purely end-user controlled to end-useroperator co-managed access points.
Further interesting results include the discovery of an unexpected lack of direct correlation between range and peaksignal-strength and a detailed aggregated coverage map for
typical urban areas in Vienna.
On the other hand, due to the increase of WiFi-installments
and the unexpected high signal-strengths observed in the
street-canyons we also conjecture that this could easily lead
to performance-problems due to the interference-load.
Overall, our study of state-of-the-art femto technologies,
together with the outlined deployment scenarios and the conducted measurements, demonstrates the good feasibility and
high attractiveness of nomadic and mobile offloading strategies
based on femto and/or WiFi technologies for mobile network
operators, especially considering the tough competitive situation of mobile broadband in many European markets.
Our future work includes a more detailed technical specification of the most relevant offloading scenarios as discussed
initially in this paper as well as the investigation of interference
problems while using licensed bands (e.g., 2.1GHz).
ACKNOWLEDGEMENTS
This research has been funded by the Austrian Government
and the City of Vienna within the competence center program
COMET. The authors would like to thank their colleagues
Georg Lffelmann, Thomas Baumgartner, Olivier Marc, Barbara Orlandi for their input and all members of the AWARE
team for many fruitful discussions.
R EFERENCES
[1] GSM Association, White paper "Case study series MKA", http://www.
gsmworld.com/documents/26052009105307.pdf, February 2007
[2] P. Reichl: From Quality-of-Service and Quality-of-Design to Qualityof-Experience: A Holistic View on Future Interactive Telecommunication
Services. Invited Paper, 15th IEEE ComSoc International Conference on
Software, Telecommunications and Computer Networks (SoftCOM07),
Split, Croatia, September 2007.
[3] T. Dunnewijk, S.Hulen: A Brief Histrory of Mobile Telekommuncation
in Europe, http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.122.
1419&rep=rep1&type=pdf
[4] L.G. Roberts, Beyond Moors Law: Internet Growth Trends, Computer,
IEEE Journal Volume 33 Issue 1, January 2000
[5] Wireless Application Forum Ltd.: Wireless Application Protocol Architecture Specification, http://www.openmobilealliance.org/tech/affiliates/wap/
wap-210-waparch-20010712-a.pdf, July 2001
[6] Mobilkom Austria press: Mobiltel Carries Out First UMTS Video Call
Between Sofia and Vienna - Technical Launch of UMTS, http://www.
mobilkomgroup.org/en/press/20050914a, September 2005
[7] ETSI MMC: Overview of 3GPP Release 1999 V0.1.1,
http://www.3gpp.org/ftp/Information/WORK_PLAN/Description_
Releases/Rel-99_description_20100214.zip, February 2010

[8] R. Love, A. Ghosh, X. Weimin, R. Ratasuk: Performance of 3GPP


high speed downlink packet access (HSDPA), Vehicular Technology
Conference, 2004. VTC2004-Fall. 2004 IEEE 60th, April 2005
[9] 3rd Generation Partnership Project (3GPP): 3GPP TS 25.308 - High
Speed Downlink Packet Access (HSDPA), (Release 5), www.3gpp.org
[10] 3rd Generation Partnership Project (3GPP): 3GPP TS 25.321 - Medium
Access Control (MAC) protocol specification (Release 6), www.3gpp.org
[11] GSM Association: Mobile Broadband Evolution, http://gsmworld.com/
documents/05032009134226.pdf, 2009
[12] 3rd Generation Partnership Project (3GPP): 3GPP TR 25.999 - High
Speed Packet Access (HSPA) evolution (Release 7), www.3gpp.org
[13] The Register: Femtocells boost cell coverage in the home, http://www.
theregister.co.uk/2010/01/19/vodafone_femtocell/, February 2008
[14] GSM Association: Mobile Broadband Investment Set to Soar as HSPA
Connections Pass 200 Million, http://www.gsmworld.com/newsroom/
press-releases/2010/4621.htm, February 2010
[15] ThinkFemtoCell: Femtocell Leadership Interview with Shlomo
Gadot Percello, http://www.thinkfemtocell.com/Femtocell-Interview/
femtocell-leadership-interview-with-shlomo-gadot-percello.html, March
2009
[16] Nokia Siemens Networks (NSN): Striding towards an open market with
3G Femto Home Access Solution, http://w3.nokiasiemensnetworks.com/
de/Portfolio/Products/Radio-access/3G_Femto_HomeAccessSolution.
htm?languagecode=de, June 2008
[17] Nokia Siemens Networks (NSN): Reaching the mass market with
3G Femto Home Access, http://w3.nokiasiemensnetworks.com/de/
Portfolio/Products/Radio-access/3G_Femto_HomeAccessSolution.htm?
languagecode=de, March 2008
[18] Airvana: Femtocells Transforming The Indoor Experience, http://www.
airvana.com/files/Airvana_Femtocell_White_Paper_Oct_2007.pdf
[19] K. Tae-Hwan, L. Tea-Jin: Throughput Enhancement of Macro and
Femto Networks By Frequency Reuse and Pilot Sensing, Performance,
Computing and Communications Conference, 2008. IPCCC 2008. IEEE
International, Austin Texas, December 2008
[20] R. Lan-lan, M. Luo-ming, Q. Xue-song: Information Modeling and
Autonomic Configuration of Femto Network, Wireless Communications,
Networking and Mobile Computing, 2009. WiCom 09. 5th International
Conference, September 2009
[21] P. Kulkarni, W. Hau Chi, T. Farnham: Radio Resource Management
Considerations for LTE Femto Cells, ACM SIGCOMM Computer Communication Review, January 2010
[22] K. Sundaresan, S. Rangarajan: Efficient Resource Management in
OFDMA Femto Cells, MobiHoc09, New Orleans USA, May 2009
[23] Fierce Wireless: European Femtocell Deployment Status, http://www.
fiercewireless.com/europe/story/european-femtocell-deployment-status/
2009-03-12, March 2009
[24] Mobilkom Austria press: Mobilkom Austria starts first Femto router pilot
project in Central Europe, http://www.mobilkom.at/en/press/20081114,
September 2008
[25] 3rd Generation Partnership Project: Generic Access Net-work (GAN)
Release 6, http://www.3gpp.org/ftp/Specs/html-info/43318.htm, June
2008
[26] D. Calin, H. Claussen, H. Uzunalioglu: On Femto Deployment Architectures and Macrocell Offloading Benefits in Joint Macro-Femto
Deployment, IEEE Communication Magazine Volume 48, January 2010

Potrebbero piacerti anche