Content uploaded by Se-Young Yun
Author content
All content in this area was uploaded by Se-Young Yun on Aug 29, 2014
Content may be subject to copyright.
IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 30, NO. 3, APRIL 2012 595
The Economic Effects of Sharing Femtocells
Se-Young Yun, Yung Yi, Dong-Ho Cho, and Jeonghoon Mo
Abstract—Femtocells are a promising technology for handling
exponentially increasing wireless data traffic. Although extensive
attention has been paid to resource control mechanisms, for
example, power control and load balancing in femtocell networks,
their success largely depends on whether operators and users
accept this technology or not. In this paper, we study the
economic aspects of femtocell services for the case of monopoly
market, and aim to answer questions on operator’s revenue,
user surplus, and social welfare by considering practical service
types and pricing strategies. We consider three user subscription
services, that is, users can access only macro BSs (mobile-only),
or deploy femto BSs in their house and open / exclusively use
their femto BSs (open- / closed-femto). For pricing strategies,
flat pricing and partial volume pricing are exploited. The main
messages include the following: 1) open-femto service is beneficial
to both users and providers; 2) in flat pricing, the impact on
operator revenue of allowing or blocking the access of mobile-
only users to open femto BSs is minor; and 3) compared
with partial volume pricing, flat pricing is advantageous to the
operator when users are sensitive to price.
Index Terms—Femtocell, pricing, market model.
I. INTRODUCTION
A. Motivation
THE DEMAND for wireless data traffic is dramatically
growing and the monthly demand has been forecasted to
reach 6.3 EB 1on 2015, a 26-fold increase over 2010 [1]. This
unprecedented growth, which is driven by the introduction
of smart mobile devices and the diversity of multimedia
applications, throws up both challenges and opportunities for
technical and business communities. In order to cope with this
growing demand, many capacity enhancement solutions have
been proposed. These include the system-wide upgrade to the
4G infrastructure, for example, LTE and WiMax, mainly by
adopting enhanced physical layer technologies [2] or ad-hoc
solutions such as offloading to WiFi [3], [4]. However, more
dominant factor for capacity increase to handle traffic explo-
sion in cellular systems is efficient frequency spatial reuse by
reducing the cell sizes, e.g., micro, pico, and femto cells [5].
The key difference between micro/pico and femto cells largely
Manuscript received 10 March 2011; revised 1 September 2011. This
research was supported by the KCC (Korea Communications Commission),
Korea, under the R&D program supervised by the KCA (KCA-2011-11913-
05004), the Korea Science and Engineering Foundation (KOSEF) grant funded
by the Korea government (MEST) (No. 2009-0075757), and MKE/KEIT (No.
KI001865). Part of this work has been published at the proceedings of IEEE
Infocom Mini-conference, 2011.
S. Yun, Y. Yi, and D. Cho are with the Department of Electrical En-
gineering, KAIST, South Korea (e-mail: syyun@comis.kaist.ac.kr, {yiyung,
dhcho}@ee.kaist.ac.kr).
J. Mo is with the Department of Information and Industrial Engineering,
Yonsei University, South Korea (e-mail: j.mo@yonsei.ac.kr). Corresponding
author.
Digital Object Identifier 10.1109/JSAC.2012.120409.
11EB or 1 exabyte is 1018 bytes.
Fig. 1. Femtocell services: mobile-only,mobile+open femto,andmo-
bile+closed femto
lies in who deploys and controls the cell: micro/pico cell
deployment is driven by operators, whereas users of femtocells
tend to individually decide on deployment.
The femtocell technology, configuring a very small cell and
the residential broadband backhaul, is economically attractive
because it can achieve high spectral efficiency at a viable cost.
Many researchers that have worked on femtocells have focused
on technical issues [6]–[14], for example, spectrum sharing be-
tween macro and femtocells, interference management through
power control, or handoffs and association for load balancing.
However, limited attention has been paid to economic and
business aspects, which is yet another important factor in
the success of femtocell technology. Recently, Shetty and
Walrand [15] proposed an economic framework and analyzed
the economic impacts of adopting femtocells on the providers’
revenue2, which inspired our work. After that, the authors in
[16] investigates the economic value of femtocells in view of
spectrum allocation algorithms.
However, an important issue has remain under-explored:
openness of femtocells. Since femtocells are typically installed
in personal indoor environments, their capacity of femtocells
tends to be higher than that of macrocells. Thus, it is expected
that the utilization of the femtocells seems to be relatively
low only with the femto owners’ traffic. Therefore, it may
be economically beneficial to users and providers to allow
“guest” users to utilize the femto BSs that are open. However,
it is far from clear how beneficial the open femtocell service
is, depending on the factors, which have significant impact on
the effect of femtocells on coping with mobile data explosion.
2We use ‘operator’ and ‘provider’ interchangeably throughout this paper.
0733-8716/12/$25.00 c
2012 IEEE
596 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 30, NO. 3, APRIL 2012
B. Summary
We propose an analytical market model between the oper-
ator and users to understand the economic impact of sharing
femtocells, whose major features are summarized as follows:
Service. As illustrated in Fig. 1, we consider three sub-
scription services for accessing the network, namely mobile-
only,mobile+open femto,andmobile+closed femto3.Asthe
names imply, the users of mobile-only services can access
only macro BSs, whereas the users of mobile+open femto
and mobile+closed femto services have access to macro BSs
as well as femtocells. The users of mobile+closed femto
services exclusively use their femto BSs, whereas those of
mobile+open femto services open their femto BSs which can
be shared by other users.
Policies for sharing open femto BSs. Operators can choose
one of two policies for users of mobile+open femto to share
their femto BSs: open-to-all and open-to-femto.Usersof
mobile-only can access open femto cells under the open-to-
all policy, but not the open-to-femto policy, which are open
only to the users subscribing to the femto service.
Pricing. We consider two pricing schemes: flat and partial
volume pricing. In both pricing schemes, a fixed service fee
is charged for accessing femto BSs. This is because there
is a practical concern that a complex metering may not be
possible in femto BSs. The difference between two schemes
lies in that in partial volume pricing users pay the service
fee in proportion to the amount of the service by macro
BSs, whereas in flat pricing the fixed fee is also charged for
using macro BSs. The current femto service market trends
support our choice of flat femto pricing. According to [17],
for example, Vodafone in Spain offers a monthly pay plan and
other providers such as Sprint and Docomo monthly charge the
cost of BS equipment. KDDI and Softbank in Japan provide
the femto services without additional charge. Verizon and
AT&T in the US have a similar pricing scheme to the Japanese
companies with minor difference in that they charge only once
when the femtocell BS is installed.
Metric and market types. The major metrics are users’
surplus, operator’s revenue,andsocial welfare. We consider
a monopoly market, where a single operator dominates and
fully controls the market price to maximize its own revenue.
Users simply follow the operator’s price control and select the
service that optimizes personal use.
The main messages of this paper are summarized as follows:
1) It is beneficial to both providers and users to have open-
femto BSs rather than just closed ones. The difference
between two services tends to grow as the number of users
and the coverage of femto BSs increases.
2) The impact of open policies is minor in flat pricing. The
differences to users and the provider are not significant
whether or not the provider limits the access of femto
BSs to mobile-only users. The subsidy for incentivising
the femto users to open their femto BSs does not have to
be large, and only needs to be approximately 10%-20% of
the price of the closed-femto service.
3We si mpl y u se open-femto and closed-femto to refer to mobile+open femto
and mobile+closed femto, respectively.
TAB L E I
SUMMARY OF MAJOR NOTATI ON (EXTERNAL PARAMETERS AND SYS TEM
VARIABLES ARE SEPARATED)
External Parameter Description
NNumber of users per one macro BS
γ, ¯γ,θ user type, max. of user type, price sensitivity
βfraction of a femto BS’s coverage
δoprobability that a user is outside
CM,C
Fcapacities of macro and femto BSs
CC,C
Ocapacities of closed and open femto BSs
ηfemto capacity reduction factor
cfmaintenance cost of a femto BS
qofraction of coverage of all open femto BSs
Uj,Φjexpected utility and service fee of service type j
α=(αm,α
o,α
c)user subscription ratios
R, S, W revenue, user surplus, social welfare
3) The provider can achieve higher revenue with the flat
pricing than it can with partial volume pricing when
users are price-sensitive. However, these differences are
significant with less sensitive users.
C. Related Work and Organization
The economic aspects of access networks have been ad-
dressed by many researchers [18]–[22]. The paper [18] per-
formed financial analysis of femtocell networks and claimed
that there could be significant cost savings using a femtocell
network. Shetty et al also showed the benefit of femtocell
using a mathematical model [15]. However, these work did not
address openness of femtocell. Pricing in Wi-Fi networks has
been addressed in [19], [20]. In [19], the authors studied the
economic incentives of WiFi network operators connecting to
ISPs. The paper [20] studied the economic interaction between
the WiFi and WiMax network providers.
The openness of access network is addressed in Wi-Fi based
networks. In Korea, one of major provider, LG U+, announced
a plan to open WiFi APs at homes for other guest users to
extend WiFi connectivity [21]. FON 4is another good example
of open WiFi APs, where a user can use any WiFi APs of FON
with no charge, if they allow other users’ access subscribing
to FON. The paper [22] dealt with business dynamics of open
Wi-Fi networks. They claim that the evolution of such network
depends on many factors such as initial coverage, subscription
fee and user preference. Our work focuses on the economic
aspect of users’ willingness to choose and/or open the femto
service. Note that we do not intend to compare the femto
service with other access solutions such as operator-deployed
pico/micro cells or FONs.
The rest of this paper is organized as follows: In Section II,
we describe the system model. Sections III and IV provide
the economic analysis for flat and partial volume pricing
schemes, followed by the the numerical results in Section V.
We conclude the paper in Section VI.
II. MODEL
A. System Model
Consider a wireless network consisting of macro and femto
BSs, where Nusers/macro-cells are served by a monopoly
4http://www.fon.com.
YUN et al.: THE ECONOMIC EFFECTS OF SHARING FEMTOCELLS 597
operator. We assume a simple model of BSs, that is, macro
and femto BSs provide the fixed capacities CMand ηCF,
where CFis the “pure” capacity of a femto BS and η∈(0,1]
is an interference factor. The value of ηdepends on spectrum
sharing and femto open policy, whose details will be discussed
in Section II-C.
Users are always guaranteed to be under the coverage of
a macro BS, but not of a femto BS. We also assume that
femtocell equipment is identical and that the coverage size
of a femto BS is the fraction βof that of a macro BS.
We do not consider the handover effects on the users, and
assume that each user already has a backhaul connection based
on residential Internet service, e.g., DSL (Digital Subscriber
Line). We adopt this simple model to purely focus on the
economic aspects of the system and to enable meaningful
analysis.
Note that the instantaneous BS capacities and the volume
of users’ delivered data may largely depend on factors such
as resource allocation mechanisms, and other factors (e.g.,
power control policy, channel conditions, user’s distance to
BS). However, simplifying the model does not overly change
the key messages and insights because we focus on economic
benefits and pricing, which typically take effect over a longer
time-scale.
B. Monopoly Operator and Services
As mentioned in Section I, the operator provides three
services, namely, mobile-only,mobile+open femto,andmo-
bile+closed femto, and two femto open-policies: open-to-all
and open-to-femto.Weuse{m, o, c}to indicate these service
types.
The operator charges φL
l(x)for generating trafficratexin
the BS type Lto the user subscribing to service l∈{m, o, c}.
The index Lin the charging function aims to show the
dependence of charging by the serving BS type. We consider
two types of tariffs: flat pricing and partial volume pricing.
We use {M, O, C}to index to macro, open-femto, and closed-
femto BSs.
In flat pricing, users’ payments are constant regardless of
data usage, that is, for any BS type L∈{M, O,C},
φL
l(x)=pl,l∈{m, o, c},(1)
where plis the constant charge for service l∈{m, o, c}.
In partial volume pricing, users pay pM
vper unit data rate
when they are served by macro BSs, whereas they pay a
fixed service fee pm,p
o,p
cfor using femto BSs. This hybrid
setup is motivated by the practical reasons that low-cost femto
BSs may not be appropriately equipped for complex per-data
operations. Recall that operators do not use volume pricing
for femto services [17]. Thus, for all l∈{m, o, c}, the pricing
structure is represented as follows:
φM
l(x)=pM
vx+pl,(2)
φL
l(x)=pl,L∈{O, C },(3)
pm=0,(4)
where pm=0is due to the fact that (pure) volume pricing is
applied when mobile-only users access macro BSs but cannot
access femto BSs.
C. Capacity and Interference Model
The interference between macro and femto BSs depends
on a spectrum sharing policy and service types. Under the
separate carriers in which femto and macro BS do not share the
spectrum, we can assume that the macro and femto capacities
are fixedtobeCMand CF. However, when they share
the carriers, e.g., partially shared carriers [14], the “actual”
capacities differ depending on the type of the femto BSs. In
the case of closed femto, macro users around the closed femto
BS can interfere the femto BS, possibly degrading the femto
capacity. However, if a femto cell is open, mobile-only users
can handoff to the open femto BS. To reflect this, we introduce
the interference factor ηand model a femto BS’ capacity as:
CC=ηCFclosed femto BS,
CO=CFopen femto BS,
where CFis again the “pure” femto BS capacity, CCand
COare “actual” capacities of the femto BS under closed and
under open-femto policies, respectively, η∈(0,1].We assume
that the macro capacity CMis not affected, which can be
justified from the fact that interference mitigation scheme can
be employed, e.g., partially shared carriers [14].
D. Users
To model the behavior of a user, we adopt an iso-elastic
utility function 5u(x;γ),givenby
u(x;γ)=γxθ,(5)
where xis traffic, γis a user type value, and θis an elas-
ticity parameter. The utility function is an increasing concave
function of traffic volume x, but with a decreasing marginal
payoff. The user type parameter γis introduced to model
different willingness to pay. As the higher γusers have the
more payoff for the same data-rate, they can afford more to
subscribe to a service. We assume γis uniformly distributed
between [0,γ
max],γ
max >0.The parameter θis closely
related to elasticity of demand, that is, the percent change of
demand to the percent change of price. A higher value of
elasticity means more changes in demand to the price change.
It is known that the elasticity for the given utility function
u(x;γ)is 1
1−θ. The iso-elastic function for data trafficisused
in [23]–[25].
As the service rates are dissimilar for macro, open and
closed femto BSs, we introduce expected utility and expected
service fee functions for service type l,Ul,and Φl, as follows:
Ul(x;γ)=Eu(xL;γ)=γ
L∈{M,O,C}
(xL)θπL
l,(6)
Φl(x)=EφL(xL)=
L∈{M,O,C}
φL
l(xL)πL
l.(7)
where πL
lis the fraction of time or probability that users of
service type luse a type LBS to get a service, and x=
(xM,x
O,x
C)is a vector that represents the traffic volume
generated in each type of BS.
5An utility function U(x)is said to be iso-elastic if for all k>0,U(kx)=
f(k)U(x)+g(k)for some functions f(k),g(k)>0.
598 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 30, NO. 3, APRIL 2012
TAB L E I I
PROBABILITY πL
lFOR OPEN-TO-FEMTO AND OPEN-TO-ALL POLI CIES
Open-to-femto Open-to-all
M O C M O C
m 1 0 0 δo(1 −qo)+δiδoqo0
oδo(1 −qo)δoqo+δi0δo(1 −qo)δoqo+δi0
cδo(1 −qo)δoqoδiδo(1 −qo)δoqoδi
Then, the net-utility ˜
Ulof service type lis given by
˜
Ul(x;γ)=Ul(x;γ)−Φl(x),l∈{m, o, c}.(8)
Users move and connect to different types of BSs over time.
Users achieve different data rates, which also depends on the
service type. Under our system model, when there are nopen-
femto users, the fraction of area qocovered by the open femto
BSs is given by:
qo1−(1 −β)n.(9)
Users’ average mobility statistics are assumed to be equal.
This is denoted by δi, which is the probability of being
“inside,” where δo=1−δi.To users of femto services,
δicorresponds to the fraction of time that they are under
the coverage of their own femto BSs. The mobile-only users
rely on macro BSs even when they are inside because of the
absence of their own femto BSs. We ignore the possibility
that the mobile-only users utilize neighboring femto BSs when
they are inside for simplicity. When users are outside, they can
access either a macro or an open-femto BS. They access an
open-femto BS with a probability qoor a macro BS with a
probability 1−qo.
Table II shows (πL
l:l∈{m, o, c},L ∈{M,O,C})
under different open policies. Under the open-to-femto policy,
mobile-only users cannot access open femto BSs and can only
access macro BSs, as shown on the first line. Open-femto users
access open femto BSs with a probability δoqo+δiand macro
BSs with probability δo(1 −qo).Theclosed-femto user case
is shown in a similar manner. Under the open-to-all policy,
even mobile-only users can access the open femto BSs.
E. Operators and Regulators
According to the user type γand charging schemes, a user
selects a service type and decides on the data demand. Let
α=(αl:l∈{m, o, c})be the vector of user fractions
subscribing to each service type l. The service type and traffic
rate vector of the user type γare denoted by l∗(γ)and x(γ),
respectively. Thus, operator revenue (R),socialwelfare(W),
and user surplus (S)are computed as
R=Φl∗(γ)(x(γ))Ndγ −(αo+αc)Ncf,(10)
W=Ul∗(γ)(x(γ); γ)Ndγ −(αo+αc)Ncf,(11)
S=˜
Ul∗(γ)(x(γ); γ)Ndγ =W−R, (12)
respectively, where cfis the cost of a femto BS for the service
provider.
III. FLAT PRICING MARKET
A. Market Model
We first consider a market model under the flat pricing
scheme. In this market model, the operator decides on the
price vector p=(pj:j∈{m, o, c})in order to maximize
the revenue Rby solving the following problem:
Provider :max
pm,po,pc≥0R. (13)
In the flat pricing scheme, the revenue in (10) is simplified to
R=N
j
pjαj−cf·(αo+αc).(14)
The subscription ratios αvary by price. We assume that
users are selfish and try to maximize individual (expected)
utility. Thus, a user of type γselects the service j∗(γ)that
maximizes his or her net-utility:
User :j∗(γ) = arg max
j∈{m,o,c}
˜
Uj(x;γ),(15)
when his or her maximum net-utility is positive, and he or
she does not select any service, otherwise. This market model
can be modeled as a two-stage sequential game, where the
operator determines the price vector to maximize the revenue
in the first stage, and then, users select one of the services (or
exit from the market) according to the price vector provided
by the operator.
B. Traffic
We assume that users are saturated and have sufficient data
to transmit whenever possible. The (average) amount of data
generated by each user depends on service type, capacities
CM,C
O,and CC,and the scheduling discipline of BSs for
competitive users. In particular, we simply assume that a
BS serves its served users equally 6. Under this fairness
assumption, the average service rate of a user served by macro
BSs is inversely proportional to the number of users in a macro
BSs, given by
xM=CM/(1 +
j∈{m,o,c}
πM
jαjN),(16)
where the denominator in (16) corresponds to the total number
of users in a macro BS, whereas a user is in the macro BS’s
service. Similarly, the service rates for users with open and
closed femto BSs are given by
xO=CO/(1 +
j∈{m,o,c}
πO
jαjN/αoN),(17)
xC=CC,(18)
respectively, where note that πO
jαjN/αoNis the average
number of users visiting an open femto BS.
6in view of long term average, users have the same data rate when their
mobility patterns are homogeneous.
YUN et al.: THE ECONOMIC EFFECTS OF SHARING FEMTOCELLS 599
C. Equilibrium
User subscription ratios α=(αj:j∈{m, o, c})are a
function of price level pand data rate x.Letγibe a point
˜
Ui(x;γi)=0for all i∈{m, o, c}and γij be a point such
that ˜
Ui(x;γij )= ˜
Uj(x;γij )for all i, j ∈{m, o, c}.
Finding the equilibrium of the flat pricing market is difficult
because of the complex inter-play between αand p. We typ-
ically use a backward induction to solve the sequential game,
that is, for a given p, we solve the user’s problem to find the
corresponding α(p). Then, we optimize the first stage game
to decide on the equilibrium price pby solving maxpR(p)
from (14). However, our problem requires a Newton-type
backward induction to numerically solve a complex fixed point
problem because of the lack of a closed form.
Note that even for a given p,computing αby solving the
problem User is difficult. This is because the net-utility of
each user affects αbecause of the problem User,andalso
because αaffects the net-utility owing to the achieved data
rates’ dependence on α.In order to explicitly represent this
dependency, we denote x(α)=(xk(α):k∈{M,O,C})for
agivenα.
Theorem 3.1 enables us to compute the equilibrium effi-
ciently (the proof is presented in Appendix). The basic idea is
that we express revenue R=R(α)as a function α, not pby
finding p’s closed form w.r.t. α.Then, revenue R(α)can be
simply maximized. Note that for a given α,there may exist
multiple values of pthat lead to the same subscription ratio α.
The set of all such pis denoted by P(α).Theorem 3.1 also
states that it is sufficient to consider one p∈P(α)because
all price vectors in P(α)lead the same.
For ease of presentation of Theorem 3.1, we will use the
index variables i, j,andkto distinctly refer to one of the
service types m, o, and c.
Theorem 3.1: Let Tl:= K∈{M,O,C}(xK)θπK
l,l ∈
{m, o, c}.Consider the following set A:
A{α|Ti≤Tj≤Tk}.
For any α∈A,there exists a p=(p
m,p
o,p
c)∈P(α),such
that
p
i=Ui(x(α); γi),
p
j=Uj(x(α); γij )−Ui(x(α); γij )+p
i,
p
k=Uk(x(α); γjk)−Uj(x(α); γjk )+p
j,(19)
where γi=1−αi−αj−αk,γij =1−αj−αk,and
γjk =1−αk.Moreover, for any p∈P(α),the provider’s
revenue is identical.
IV. PARTIAL VOLUME PRICING MARKET
A. Market Model
We also consider a partial volume pricing market model.
The market is slightly different from the flat pricing market
in that the provider should decide on the volume-based price,
pM
v,when a user is served by a macro BS, and a user should
also decide on the elastic data demand xM.In this section,
we only consider the case of the open-to-femto policy for the
following reason. When pm=0,in the open-to-all policy,
mobile-only users can use a free open-femto service, in which
case the provider’s revenue is significantly reduced because of
free-riding.
The provider selects the optimal prices that maximize the
following problem:
Provider :max
pM
v,po,pcR
s.t pM
v,p
o,p
c≥0,(20)
Then, a user with type γfirst determines the data demand
for macro BSs xM(γ)by maximizing the corresponding
surplus subject to the macro BS capacity constraint. He or
She then selects a service type to maximize the net-utility.
User :xM(γ) = arg max
xγxθ−pM
vx,
j∗(γ) = arg max
j∈{m,o,c}
˜
Uj(x;γ).(21)
The total amount of traffic must be less than the capacity of
the macro BS. Thus, every user can be served according to his
or her entire demand xM(γ), when the following condition is
satisfied.
TMNπM
j(γ)xM(γ)dγ ≤CM,(22)
where TMdenotes the total macro BS traffic generated by
users. When TMexceeds CM,asinflat pricing, we assume
that the fair scheduler controls the serving rate. Thus, the
service rate is suppressed by an upper bound xM
max,where
TM=NπM
j(γ)min{xM
max,x
M(γ)}dγ =CM.(23)
Note that unlike flat pricing where some users exit from the
market and subscribe to no services, every user selects one
of the services in partial volume pricing. The revenue of the
operator simply reads:
R=TMpM
v+N(po−cf)αo+(pc−cf)αc.(24)
B. Equilibrium
Computing the equilibrium in partial volume pricing is even
harder than it is in flat pricing. This difficulty is caused by
the hybrid structures of the two pricing schemes, where the
volume pricing for macro BSs often causes the net-utility
to be non-linear. When the net-utility graph has only one
intersection point between any two lines, it is relatively easy
to find the equilibrium point (as in the flat pricing case).
However, this non-linear net-utility graph sometimes generates
multiple intersections between any two net-utility curves,
leading to difficulties in finding the relationship between α
and γ, which is the first step in computing the equilibrium.
For simplicity, we consider the case when ¯γ=1. Similarly
to flat pricing, we develop a theorem to compute the equi-
librium. We again use the notations xk(α),k∈{O, C }and
xM(α;γ)to explicitly show the dependence of the data rates
on α.We use x(α)to denote the vector of xk(α),and omit
γfor notational simplicity, unless required.
In the description of Theorem 4.1, similarly to Theorem 3.1,
we use the index variables i, j. When xO(α)>x
C(α),i=
‘o’,j=‘c’, otherwise, i=‘c’,j=‘o’.
600 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 30, NO. 3, APRIL 2012
(a) Revenue
(b) User surplus
(c) Social welfare
Fig. 2. Flat pricing: value-added of the femto services (N= 200,β =
0.0048,θ=0.5,η=1)
Theorem 4.1: We d efine A
A{α|xM(α;γ)
≤min{xO(α),x
C(α)},for all γ∈(0,1]}.(25)
Then, for all γ∈(0,1] and any given α∈A,
(i) The γmi and γij are unique and given by:
γmi =1−αi−αj,γ
ij =1−αj.(26)
(ii) The piand pjare then expressed as a closed form of α
in the following manner:
pi=Ui(x;γmi)−Um(x;γmi)+pM
vxM(α;γmi),
pj=pi+Uj(x;γij )−Ui(x;γij ).(27)
(iii) pM
v=∞maximizes the provider’s revenue if
θ
2−θ
(1 −πM
i)(γmi)2−θ
1−θ+πM
i
(1 −πM
i)(γmi)1
1−θ(1 −γmi)<1.(28)
(a) Revenue
(b) User surplus
(c) Social welfare
Fig. 3. Partial volume pricing: value-added of the femto services (N= 200,
β=0.0048,θ=0.5,η=1).
Otherwise, the following pM
vmaximizes the revenue:
pM
v=θ1−θ
2−θ(1 −πM
i)γ
2−θ
1−θ
mi +πM
i N
CM1−θ
.
(29)
The proof is presented in the Appendix. Note that A
contains all αwhere femtocells give better throughput to users
than that of macrocells. Therefore, Aincludes all cases where
femtocells generate additional revenue to the operator.
Theorem 4.1(i) states that a relationship between the sub-
scription ratio and the type γcan be simply characterized.
Theorem 4.1(ii) represents poand pcas simple functions of
α.In Theorem 4.1(iii), the equations (28) and (29) are the
functions of α,for example, γmi and πM
iare determined for
agivenα.Thus, a simple optimization can be used to compute
the equilibrium as in Theorem 3.1.
YUN et al.: THE ECONOMIC EFFECTS OF SHARING FEMTOCELLS 601
Fig. 4. Impact of Femto capacity reduction factor on the Revenue (N= 200,
β=0.0048,θ=0.5,c
f=0)
In Theorem 4.1(iii), the left-hand side of the condition (28)
is more likely to be satisfied when there are more femto users,
because γmi declines for such a case. Thus, it means that with
many femto users, pM
vshould be large to increase revenue,
because with small pM
v, more users tend to subscribe to the
mobile-only service. In such a case, the provider will decrease
poand pcto attract more femto users, resulting in an overall
decrease in revenue. Note that for a small number of femto
users (i.e., (28) is violated), the provider gets more revenue
with smaller pM
v,wherein the data demand for macro BSs will
grow. Thus, in order to maximize revenue, pM
vis decreased to
(29) until the data demand at macro BSs reaches its capacity.
Note that even when pM
vis very large, where users decrease
the traffic demand for macro BSs, γmi can be some positive
value.
V. N UMERICAL RESULTS
A. Setup
We now provide numerical results, where in most cases,
we plot the provider’s revenue, user surplus, social welfare,
and user subscription ratio for different values of femto
costs, pricing schemes, and users’ price-sensitivities. We tested
different values and observed similar trends to those presented
in this section.
We consider a cellular network with N users/cells, where
Nis tested ranging from 100 to 350. Both CMand CF
are set equal to 1. Note that the actual numbers of CMand
CFare not critical, because revenue, user surplus, and social
welfare just scale with those numbers; our main interest lies
in investigating the relative ratios and changes of the metrics.
The CFcan vary according to the transmission rate of the
backbone network or the transmission power level of the femto
BS. However, the ratio of CMto CFseems realistic, because
the power level of femto BSs is set to give the same SINR to
users on the boundary between femtocells and macrocells [18].
The probability of users being inside is set to be 0.4 7.The
value β, the coverage of femto BSs (normalized by that of a
macro BS) is tested over [0.0048,0.03], where we use 0.0048
unless explicitly mentioned. This value is obtained for macro
7According to [1], the probability of being at home is 40% and being at
work is 30%, roughly. As our focus is on the individual user’s acceptance of
femto BSs, we use the probability of being home as that of being inside.
Fig. 5. Normalized revenue of open-femto BSs by that of just closed-femto
BSs (θ=0.5,η=1,cf=0). The percentages on the top of bars indicates
the revenue increasing by open femto BSs.
and femto cells with radiuses of 500 m and 20 m, respectively,
and macrocells exploit a three-sector topology. The value of
maximum user type, ¯γis set to 1. For all simulations, unless
explicitly mentioned, the price sensitivity is chosen as 0.5,
which is the median of the interval [0,1].We vary the femto
capacity reduction factor ηin the interval [0.2,1],where we
use η=1unless explicitly mentioned.
B. Value-added of the Open-Femto Service
Figs. 2 and 3 show the impact of open-femto services on
the revenue and user surplus, when η=1.We compare
three different cases: 1) no femto, 2) only with closed-femto
BSs, and 3) with both closed and open femto BSs. We first
observe that both revenue and user surplus increase with the
introduction of closed-femto services for all pricing schemes,
as also reported in [15]. From these results, we can compute
social welfare, which also increases thanks to the closed-femto
service, because social welfare is the sum of revenue and
user surplus. The introduction of open-femto services, further
increases revenue and user surplus because opening femtocells
increases the total capacity of the system 8.
Fig. 4 shows the impact of interference between femto BSs
and the macro BS. The x-axis and y-axis correspond to the
interference factor and the revenue, respectively. The revenue
with the open femto services is constant while that with the
closed femto service decreases as ηdecreases. The differences
between the open femto and the closed femto increase even
more as ηdecreases, which shows the superiority of the open
femto policy. As more macro users interfere femto BSs, the
value of closed-femto BSs decreases linearly.
The value-added of open femtocells changes depending on
the coverage βof open femto BSs and the number of users
N. In Fig. 5, the normalized revenue of open-femto BSs by
that of just closed-femto BSs is monotonically increasing, as
βand Nincrease. This is because more offloading can be
achieved by open-femto BSs than closed-femto BSs.
The value-added of open femto services decreases with
increasing femto costs. This reduction is because more users
select the macro-only service with high femto costs, as also
8It is called positive externality in economics.
602 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 30, NO. 3, APRIL 2012
Fig. 6. Users’ subscription ratios in flat pricing (N= 200,β=0.0048,
θ=0.5,η=1)
shown in Fig. 6. In our environments, for femto costs higher
than 0.4, no value-added is observed. Note that the cost 0.4 is
very high in that 0.13 is the price that maximizes the revenue
without femto BSs, that is, the cost 0.4 is approximately
three times larger than the price of the macro-only service.
Theorem 5.1 supports the observations above, stating that
when the femto-cost is not significantly high, the open-femto
service generates higher revenue than the macro-only service
under flat pricing does. For a given femto cost c,RM(c),
RA(c)and RF(c),denote the maximum revenues with no
open-femto service, open-femto with open-to-all,andopen-
femto with open-to-femto, respectively.
Theorem 5.1: Under flat pricing, there exist numbers ¯cA
and ¯cFsuch that RA(c)≥RM(c)for any 0≤c≤¯cAand
RF(c)≥RM(c),for any 0≤c≤¯cF.Moreover, ¯cA>¯cF.
The proof and detailed expressions of ¯cAand ¯cF,are
presented in the Appendix. In order to offer practical insight,
we numerically computed ¯cAand ¯cFfor θ=0.5:¯cA=0.38
and ¯cF=0.36.Thus, in flat pricing, the open-to-all policy
is more economically robust to the femto cost, and, in our
environments, it is verified that the bounds in Theorem 5.1
are a good match for those from the numerical computation.
Fig. 7 shows the Return on Investment (RoI) of femto BSs
over femto costs for both pricing schemes. RoI intuitively
refers to the increase in revenue compared to the invested
capital. We assume that the investment is proportional to
the number of deployed femto BSs. Thus, we define RoI as
(R−Rmacro)/(αo+αc)N, where Rand Rmacro denote operator
revenue and the revenue only with macro BSs, respectively.
In practice, RoI is a useful metric for making decisions on
investment, which gives an insight into how long it will take to
recover the investment to the femtocell services [26]. For both
pricing schemes, the RoIs of the three deployment scenarios
are similar (see Fig. 7), whereas, in the revenue graph, the
revenue of “with closed femto BS only” is significantly smaller
than other “with ‘open to femto’ policy”. This is because open
femto BSs are promoted by a subsidy so that the number of
femto BSs can increase, although revenue increases more with
open femto BSs.
C. Impact of Femto Cost on User Behavior, Price, and Subsidy
The operator may want to provide a subsidy to motivate
femto users to open their femto BSs since the utility of closed-
Fig. 7. Return on Investment (ROI) of femto BSs (N= 200,β=0.0048,
θ=0.5,η=1)
Fig. 8. Prices in open-to-all policy in flat pricing. The percentages on top
of the bar graphs represent the subsidy, calculated by pc−po
pc.(N= 200,
β=0.0048,θ=0.5,η=1)
femto service always exceeds that of open-femto.Wedefine
subsidy as the fraction pc−po
pc.This definition is adopted to
reflect the fact that since open-femto users induce positive
externalities, operators discount subscription fees for the femto
service. Then, interesting questions include (i) how much
subsidy is necessary, and (ii) how may users will subscribe
to each service for the given subsidy.
Fig. 6 shows how the user behavior changes when femto
costs or femto open policies vary under flat pricing. We
observe that when the femto cost is low (cf≤0.2), the
majority of users join the open-femto or closed-femto services,
whereas the subscription ratio decreases significantly as femto
cost increases (cf>0.2). As shown in Fig 8, the provider
whose objective is revenue maximization selects low prices
for the mobile-only service for high femto cost, in which case,
for the provider it is hard to attract more users subscribing to
femto services.
Our numerical study suggests that the subsidy ranges be-
tween 10% and 20%, as shown in Fig. 8. We also observe
that the provider may still start an open-femto business despite
(po−pm)<c
f,which implies that it should pay more money
to install and maintain a femtocell than the increased price
from introducing femtocells. This is illustrated in Fig. 8 for the
femto cost >0.2and the open-to-all policy. Once again, the
reason for this is the positive externalities of open femto BSs.
Under the regime of non-negligible portion of mobile-only
YUN et al.: THE ECONOMIC EFFECTS OF SHARING FEMTOCELLS 603
(a) Revenue
(b) User surplus
(c) Social welfare
Fig. 9. Impact of pricing schemes (N= 200,β=0.0048,η=1)
users, the provider can increase the price pmand thus increase
the revenue earned from the mobile-only users. Despite a
sufficient subsidy, the operator sustains high revenues, because
more open-femto users lead both femto users and mobile-
only users to increase their utilities and thus a high price is
acceptable to users.
When interference factor ηis considered, e.g., η<1,open-
femto users start to appear without any subsidy, since the larger
capacity of open femto BSs provides enough incentive to open
BSs. For example, in our simulation, when η<0.7,no closed-
femto users exist without subsidy. Thus, we can conclude
that restrictive use of resource for closed femto BSs due to
interference can be an enough incentive to open.
D. Open-to-all vs. Open-to-femto Policies
Fig. 2 shows that the plots for the two polices are close
for all values of femto costs, where a small economic gain
is observed in the open-to-all policy over the cost range
[0.2,0.4]. When the femto cost is less than 0.2, it is trivial that
there is no difference between open-to-all and open-to-femto
because users do not subscribe to the mobile-only service
as shown in Fig. 6. On the other hand, over the cost range
[0.2,0.4],mobile-only and femto service users can coexist
and thereby, mobile-only users influence the economic aspects
of the system. Over this cost range, the open-to-all policy
produces only positive effects on user surplus because open
femto BSs offload more data. However, the gain on revenue is
minor because under open-to-all the price for femto services
should be discounted due to the loss on their utility by the
mobile-only users’ access, whereas the revenue earned from
mobile-only users increases.
However, the impacts of these open policies on user behav-
ior is noteworthy. When cfis low, user behavior is almost
identical regardless of the policy. Differences exist only when
the femto cost is high. Under the open-to-femto policy, users
that subscribe to the femto service decide not to share their
femto services, whereas under the open-to-all policy, users
choose the open-femto service instead of the closed-femto
service. Users seem to have higher incentives to share their
femto BSs under the open-to-all policy than they do under the
open-to-femto policy. Because sharing helps macro users as
well as femto users, the operator can provide enough subsidy
to persuade users to share their femto BSs.
Decrease in subsidy when cf=0.4does not have a
strong impact on the overall analysis, because for such a high
cost, the users of femto services are extremely small or even
disappear. We model the total femto operational costs as being
linearly proportional to the number of femto users, as seen in
(10). However, this model may not reflect certain practical
cases. Thus, the results for high femto costs, for example,
cf>0.3may not deserve much attention.
E. Flat vs. Partial Volume
We now study the impact of pricing schemes. As shown
in Fig. 9, we observe that in flat pricing the revenue is no
less than that with partial volume pricing over most values
of elasticity parameter, θ. In particular, smaller θresults in a
significant gap in revenue between two pricing schemes. We
interpret this result as follows:
In flat pricing, it is widely known that the users with
higher willingness to pay, that is, γ, tend to dominate the
network resources [1], [27]. This “negative externality” (i.e.,
congestion) due to users’ heterogeneity in terms of willingness
to pay in flat pricing can be alleviated in various ways which
includes QoS-provisioning mechanism, i.e., imposing the max-
imum rate on the users with high demands or guaranteeing
the minimum rate to the users with small demands. Volume
pricing can clearly be another solution that lets the users with
higher demand pay more. Adding QoS control to flat pricing
often leads to larger revenue than volume pricing [28]. In our
model, scheduling across users in a cell is assumed to be fair,
and thus each user is served with a similar rate differently
from the actual demand, which behaves like a QoS-control
mechanism mentioned above. Note that this assumption is not
significantly impractical. For example, in Korea, operators still
adopt flat pricing (i.e., unlimited plan), but, with a QoS control
which constrains users’ maximum usage per day.
604 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 30, NO. 3, APRIL 2012
The revenue difference between two pricing schemes be-
comes larger with smaller θ. Note that θis an elasticity
parameter, i.e., as θgoes to 1, the traffic demand significantly
varies as the user type γ,whereas,asθgoes to 0, the traffic
demand is insensitive to the user type γ, and thus whether
users are served or not itself affects the provider’s revenue,
not the traffic demand. From a simple calculation, the traffic
demand over a macro BS under volume pricing, xM(γ),fora
given user type γ, is given by xM(γ)=(
γθ
p)1/(1−θ).Thus, for
low θ, users tend to transmit a (relatively) small volume of data
with volume pricing, even if the price pM
vis low. However,
under flat pricing, the maximum price which guarantees a
positive net-utility is γxθ(due to the condition of γxθ−p≥0)
and γxθincreases as θdecreases (due to x<1). Thus, users
are willing to subscribe to the service even with high price,
which leads to higher revenue in flat pricing.
In regard to user surplus, we observe that a sharp increase
of surplus for some specific value of θ:θ=0.4for cf=0and
θ=0.1for cf=0.3in Fig. 9. We resort to Theorem 4.1 to
interpret this. For small values of θ, the LHS of the condition
(28) is likely to be met, where the provider chooses pM
v=∞
to maximize the revenue. In that case, no users generate data
traffic at macro BSs, which connects large decrease in surplus.
As θincreases, the condition (28) starts to be unsatisfied, then
users accordingly start to use macro BSs with the price of (29)
and experience the increase in surplus. For very high θ(very
low price-sensitivity), the provider can attract the users with
the increased pM
v,again reduces the user surplus.
VI. CONCLUDING REMARKS
In this paper, we developed an analytical framework in order
to study the economic aspects of the femtocell services on
monopoly market. In particular, we focus on investigating the
economic benefits of openness of femtocell networks based
on the decision of users when the investment on femtocell
networks are already taken. Under the developed model we
drew the following conclusions:
1) With an enhanced network capacity driven by open-femto
BSs, the open-femto service also benefitsusersaswellas
providers because users can enjoy cheaper services (i.e.,
subsidy) with better quality. Social welfare also increases
in this scenario.
2) When open-femto services are offered, blocking the access
of mobile-only users to femto BSs does not significantly
influence revenue, user surplus or social welfare. Providers
can choose either option depending on their preferences
without the loss of economic benefit.
3) When users are price-sensitive, providers can achieve
higher revenues with flat pricing than they can with partial
volume pricing. However, the difference becomes negligi-
ble for users that have low price-sensitivity.
We comment that our conclusion 2) is based on the assump-
tion that both femto/macro BS’s capacity is not differentiated
between two open-femto policies. In practice, in open-to-all
policy the actual system capacity may be larger than that in
open-to-femto policy due to interference. Thus, there may exist
a gap in terms of revenue between two policies, which is left
for future work. Future work also includes the study of the
relation between the user subscription ratio αand the femto
capacity reduction factor η, where the value of ηmay differ
for the closed femto BSs and the open femto BSs, under
open-to-femto policy. Extending this work, it is interesting
to see the followings: (i) the study of the cases when there
exist multiple providers and the impact of their cooperation
and competition 9and (ii) comparison with other competitive
technologies (e.g., picocell and WiFi).
APPENDIX
Proof of Theorem 3.1. For each of γi,γ
ij ,and γjk,we find
that ˜
Ui(x;γi)=0,˜
Ui(x;γij )= ˜
Uj(x;γij ),and ˜
Uj(x;γjk)=
˜
Uk(x;γjk).Thus, at p, the equilibrium exists with αsince
Ti≤Tj≤Tk.For the given α,pl,wherel∈{m, o, c},can
take other values only when αl=0.
Proof of Theorem 4.1. At the piand pj,˜
Ui(x;γij )=
˜
Uj(x;γij ). Moreover, the relationship ˜
Ui(x;γ)>˜
Uj(x;γ)
holds when γ<γ
ij ,and the relationship ˜
Ui(x;γ)<˜
Uj(x;γ)
holds when γ>γ
ij .Thus, ˜
Ui(x;γ)and ˜
Uj(x;γ)intersect
uniquely at γij .
The utility difference between service type iand mobile-
only service is given by:
˜
Ui(x;γ)−˜
Um(x;γ)=γ
K∈{O,C}
(xK)θπK
i−
(1 −πM
i)γ(xM(γ))θ−pM
vxM(γ)−pi.(30)
For γb<γ
a,wehave:
˜
Ui(x;γa)−˜
Um(x;γa)+pi(31)
=γa
K∈{O,C}
(xK)θπK
i−
(1 −πM
i)γa(xM(γa))θ−pM
vxM(γa)(32)
≥γb
K∈{O,C}
(xK)θπK
i−
(1 −πM
i)γb(xM(γa))θ−pM
vxM(γa)(33)
≥γb
K∈{O,C}
(xK)θπK
i−
(1 −πM
i)γb(xM(γb))θ−pM
vxM(γb)(34)
=˜
Uo(x;γb)−˜
Um(x;γb)+po,(35)
where Eq. (33) is obtained from the condition min{xO,x
C}≥
xM(γa)and Eq. (34) is obtained from the fact that xM(γb)
maximizes γbxθ−pV
mx. Thus, ˜
Ui(x;γ)−˜
Um(x;γ)is in-
creasing in γ, and γmi and γij are uniquely determined with
γij =1−αjand γmi =γij −αi.
Assume that the capacity for macro BSs is not constrained.
Remarking that:
dUl(x;γ)
dxM=πM
lθγ(xM)θ−1−pM
v,l∈{m, o, c}
(36)
9The duopoly case is an ongoing work [29].
YUN et al.: THE ECONOMIC EFFECTS OF SHARING FEMTOCELLS 605
the service rate of macro BSs for users with type γis xM(γ)=
(θγ
pM
v)1
1−θ.We also find that:
dR
dpM
v
=Nθ 1
1−θ(pM
v)1−2θ
1−θ(1 −πM
i)(γmi)1
1−θ(1 −γmi)
−θ
2−θ((1 −πM
i)(γmi)2−θ
1−θ+πM
i).(37)
Thus, when the condition (28) is met, the revenue increases
with pM
v.Otherwise, owing to the capacity constraint, the
revenue is maximized when pM
vsatisfies following:
CM=πM
j(γ)(θγ
pM
v
)1
1−θdγ
=N1−θ
2−θ(1 −πM
i)(γmi)2−θ
1−θ+πM
i θ
pM
v
1
1−θ.
This concludes the proof.
Proof of Theorem 5.1. We prove the theorem by finding the
conditions on the femto costs for open-to-all and open-to-
femto, such that revenue when there are only mobile-only users
increases along with the change from a mobile-only user to a
femto user.
Initially, let αbe the subscription ratio for the system with
only mobile-only users. Subsequently, by jointly solving (13)
and (15), Rin (14) is given by: R=(1−α)(CM/(αN ))θαN,
where we get:
dR
dα =NCM
Nθα−θ(1 −θ−(2 −θ)α),(38)
d2R
dα2=−NCM
Nθ(1 −θ)(2 + θ(1 −α)α−1).(39)
Since Eq. (39) is non-positive for α>0,θ∈[0,1],Ris
concave in α. Thus, Ris maximized when α∗:= 1−θ
2−θ.Let
γb
mbe 1−α∗.Then, when no femto BS exists, according
to Theorem 3.1, the maximum revenue is Um(x(0) ;γb
m)α∗N,
where x(0) is a traffic rate vector.
Let us assume that, under the open-to-all policy, a user,
whose type value is ¯γ, changes his or her service to open-
femto. Then, there are α∗N−1mobile-only users and an
open-femto user. From Theorem 3.1, the prices for mobile-
only and open-femto are defined as
p(1)
m:= Um(x(1);γb
m),(40)
p(1)
o:= Uo(x(1);1)−Um(x(1) ;1)+Um(x(1);γb
m),(41)
where x(1) is a traffic rate vector when there are α∗Nsub-
scribers and they are mobile-only users except one subscribing
to the open-femto service. Thus, the revenue increment by
introducing an open-femto user is computed as
R∗=(Um(x(1) ;γb
m)−Um(x(0);γb
m))α∗N
+(Uo(x(1);1)−Um(x(1) ;1))−c. (42)
If R∗is positive, the provider can gain more revenue with an
open femto BS. Let the bound ¯cOdenote the maximum femto
cost that guarantees more profit with an open femto BS. Then
the bound ¯cOis represented as
¯cO= ((1 −α∗)α∗Nδoβ+δi)( CO
1+δi+δoβα∗N)θ−
(1 −α∗)α∗N(CM
1+α∗N)θ+ ((1 −α∗)α∗N(1 −δoβ)−δi)
×(CM
1+δi(α∗N−1) + δo(1 −β)α∗N)θ.(43)
Similarly, we can also compute the bound ¯cCwhich is the
maximum femto cost where a provider gains revenue by
introducing one closed-femto user.
¯cC=δiCθ
C+ ((1 −α∗)α∗N−δi)( CM
1+α∗N−δi
)θ
−(1 −α∗)α∗N(CM
1+α∗N)θ.(44)
Thus, under the open-to-all policy, ¯cA=max(¯cO,¯cC).
Moreover, under the open-to-femto policy, ¯cF=¯cC,because
only one open-femto user has the same characteristic as only
one closed-femto user. Finally, we can conclude that ¯cA≥¯cF.
This concludes the proof.
REFERENCES
[1] Cisco, “Cisco visual networking index: Global mobile
data traffic forecast update, 2010-2015,” February 2011,
http://www.cisco.com/en/US/solutions/collateral/ns341/ns525/ns537/
ns705/ns827/white_paper_c11-520862.html.
[2] R. Irmer, H.-p. Mayer, A. Weber, V. Braun, M. Schmidt, M. Ohm,
N. Ahr, A. Zoch, C. Jandura, P. Marsch, and G. Fettweis, “Multisite field
trial for LTE and advanced concepts,” IEEE Commun. Mag.,vol.47,
no. 2, pp. 92 –98, february 2009.
[3] A. Balasubramanian, R. Mahajan, and A. Venkataramani, “Augmenting
mobile 3g using wifi,” in Proc. ACM Mobisys, 2010.
[4] K. Lee, Y. Yi, J. Lee, I. Rhee, and S. Chong, “Mobile data offloading:
How much can wifideliver?” in Proc. ACM CoNext, 2010.
[5] M. Dohler, R. Heath, A. Lozano, C. Papadias, and R. Valenzuela, “Is
the phy layer dead?” IEEE Commun. Mag., vol. 49, no. 4, pp. 159 –165,
april 2011.
[6] J.-H. Yun and K. G. Shin, “CLTR: A self-organizing femtocell manage-
ment architecture for co-channel deployment,” in Proc. ACM MobiCom,
2010.
[7] C.-W. Chen, C.-Y. Wang, S.-L. Chao, and H.-Y. Wei, “DANCE: A game-
theoretical femtocell channel exchange mechanism,” in Proceedings of
ACM MobiCom, 2009.
[8] M. Andrews, V. Capdevielle, A. Feki, and P. Gupta, “Autonomous
spectrum sharing for mixed LTE femto and macro cells deployments,”
in Proc. IEEE INFOCOM, 2010.
[9] K. Sundaresan and S. Rangarajan, “Efficient resource management in
OFDMA femto cells,” in Proc. ACM MobiHoc, 2009.
[10] Z. Bharucha, H. Haas, G. Auer, and I. Cosovic, “Femto-cell resource
partitioning,” in Proc. IEEE GLOBECOM, November 2009.
[11] H.-S. Jo, C. Mun, J. Moon, and J.-G. Yook, “Interference mitigation
using uplink power control for two-tier femtocell networks,” IEEE
Trans. Wireless Commun., vol. 8, no. 10, pp. 4906–4910, October 2009.
[12] V. Chandrasekhar, J. G. Andrews, T. Muharemovic, Z. Shen, and
A. Gatherer, “Power control in two-tier femtocell networks,” IEEE
Trans. Wireless Commun., vol. 8, no. 8, pp. 4316–4328, August 2009.
[13] P. Kulkarni, W. H. Chin, and T. Farnham, “Radio resource management
considerations for LTE femto cells,” SIGCOMM Computing Communi-
cation Review, vol. 40, no. 1, pp. 26–30, 2010.
[14] J. D. Hobby and H. Claussen, “Deployment options for femtocells and
their impact on existing macrocellular networks,” Bell Labs Technical
Journal, vol. 13, no. 4, pp. 145 – 160, february 2009.
[15] N. Shetty, S. Parekh, and J. Walrand, “Economics of femtocells,” in
Proc. IEEE Globecom, 2009.
[16] L. Duan, J. Huang, and B. Shou, “Economic viability of femtocell
service provision,” in Proc. Gamenets, 2011.
[17] informa telecoms &media, “Femtocell market status,” Febru-
ary 2011, http://www.femtoforum.org/femto/pdfsend01.php?file=023%
20Informa%20Femtocell%20Market%20Status%20Q1%202011.pdf.
606 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 30, NO. 3, APRIL 2012
[18] H. Claussen, L. T. W. Ho, and L. G. Samuel, “An overview of the
femtocell concept,” Bell Labs Technical Journal, vol. 13, no. 1, pp. 221
–245, may 2008.
[19] J. Musacchio and J. Walrand, “Wifiaccess point pricing as a dynamic
game,” IEEE/ACM Trans. Netw., vol. 14, pp. 289 – 301, April 2006.
[20] D. Niyato and E. Hossain, “Wireless broadband access: Wimax and
beyond - integration of wimax and wifi: Optimal pricing for bandwidth
sharing,” IEEE Commun. Mag., vol. 45, no. 5, pp. 140 – 146, May 2007.
[21] The Korea Herald, “LG Uplus to expand wireless networks,”
Nov. 2010, http://www.koreaherald.com/national/Detail.jsp?newsMLId=
20101129001028.
[22] M. H. Manshaei, J. Freudiger, M. Félegyházi, P. Marbach, and J.-
P. Hubaux, “On wireless social community networks,” in INFOCOM,
2008, pp. 1552–1560.
[23] L. A. DaSilva, “Pricing for QoS-enabled networks: A survey,” IEEE
Commun. Surveys Tutorials, vol. 3, no. 2, pp. 2 – 8, 2000.
[24] D.Mitra, K.G.Ramakrishnan, and Q.Wang, “Combined economic mod-
eling and traffic engineering: Joint optimization of pricing and routing
in multi-service networks,” in Proc. ITC, 2001.
[25] J. Huang and X. Huang, “Revenue management for cognitive spectrum
underlay networks: An interference elasticity perspective,” in Proc.
APCC, 2009.
[26] V. Chandrasekhar, J. Andrews, and A. Gatherer, “Femtocell networks:
asurvey,”IEEE Commun. Mag., vol. 46, no. 9, pp. 59 –67, september
2008.
[27] J. MacKie-Mason and H. Varian, “Pricing congestible network re-
sources,” IEEE J. Sel Areas Commun., vol. 13, no. 7, pp. 1141 –1149,
sep 1995.
[28] J. Altmann and K. Chu, “How to charge for network services - flat-rate
or usage-based?” Computer Networks, vol. 36, no. 5-6, pp. 519 – 531,
2001.
[29] S. Yun, Y. Yi, D. Cho, and J. Mo, “On the efficiency of open femtocell
services,” technical Report, KAIST, http://lanada.kaist.ac.kr/femto.pdf.
Se-Young Yun (S’10) received the B.S. in electrical
engineering from the KAIST, Daejeon, Republic of
Korea, in 2006. He is currently working toward
the Ph.D degree in electrical engineering at KAIST.
His research interests include network economics,
future wireless communication systems, femtocell
networks, and green communications.
Yung Yi (S’04, M’06) received his B.S. and the
M.S. in the School of Computer Science and En-
gineering from Seoul National University, South
Korea in 1997 and 1999, respectively, and his Ph.D.
in the Department of Electrical and Computer En-
gineering at the University of Texas at Austin in
2006. From 2006 to 2008, he was a post-doctoral
research associate in the Department of Electrical
Engineering at Princeton University. Now, he is an
associate professor at the Department of Electrical
Engineering at KAIST, South Korea. He has been
serving as a TPC member at various conferences including ACM Mobihoc,
Wicon, WiOpt, IEEE Infocom, ICC, Globecom, and ITC. His academic
service also includes the local arrangement chair of WiOpt 2009 and CFI
2010, the networking area track chair of TENCON 2010, and the publication
chair of CFI 2010, and a guest editor of the special issue on Green Networking
and Communication Systems of IEEE Surveys and Tutorials. He also serves
as the co-chair of the Green Multimedia Communication Interest Group
of the IEEE Multimedia Communication Technical Committee. His current
research interests include the design and analysis of computer networking and
wireless systems, especially congestion control, scheduling, and interference
management, with applications in wireless ad hoc networks, broadband access
networks, economic aspects of communication networks economics, and
greening of network systems.
Dong-Ho Cho (M’85, SM’00) received the B.S.
degree in electrical engineering from Seoul National
University, Seoul, Korea, in 1979 and the M.S.
and Ph.D. degrees in electrical engineering from
the KAIST, Daejeon, Republic of Korea, in 1981
and 1985, respectively. From 1987 to 1997, he
was a Professor in the Department of Computer
Engineering at Kyunghee University. Since 1998, he
has been with KAIST, where he is a Professor in
the Department of Electrical Engineering. Also, he
was a Director of KAIST Institute for Information
Convergence from 2007 to 2011 and has been a Director of Online Electric
Vehicle Project since 2009. His research interests include wireless communi-
cation, magnetic communication, bio informatics and wireless power transfer.
Jeonghoon Mo received the BS degree from Seoul
National University, Korea, and the MS and Ph.D.
degrees from the University of California, Berkeley.
He is a professor in the department of Information
and Industrial Engineering at Yonsei University.
Before joining Yonsei University, he was with AT&T
Bell Laboratories, Middletown, New Jersey, and
with start-ups in San Jose, California. He worked
for voice-over-IP (VoIP) quality assessment and per-
formance analysis of network processor. He is the
author of “Performance Modeling of Communica-
tion Networks with Markov Chains”. His research interests include network
economics, optimization, performance analysis, transport/media access control
(MAC) design of communication networks, WiMax, and Wi-Fi.
