Spectrum and Infrastructure Sharing in Millimeter Wave Cellular Networks: An Economic Perspective
Spectrum and Infrastructure Sharing in Millimeter Wave Cellular Networks: An Economic Perspective
Spectrum and Infrastructure Sharing in Millimeter Wave Cellular Networks: An Economic Perspective
Fraida Fund, Shahram Shahsavari, Shivendra S. Panwar, Elza Erkip, Sundeep Rangan
I. I NTRODUCTION
The millimeter wave (mmWave) bands represent one of
the largest unlicensed bandwidths ever allocated, presenting
a tremendous opportunity for both technical and policy innovation. The appropriate licensing model for this band remains
the subject of considerable debate. Replies to an FCC notice
of inquiry [1] requesting comments on usage of bands greater
than 24 GHz in the United States reveal disagreement on how
to best utilize this spectrum, with economic considerations
playing a significant role. Major industry players argued in
favor of exclusive use licensing on a geographic service area
basis, primarily on the grounds that this offers sufficient
certainty to motivate major capital investment. Several of these
explicitly asked the FCC to reject licensing mechanisms that
require spectrum sharing on some bands [2][8]. Others argued
that unlicensed use maximizes efficient spectrum use, and
encourages innovation and competition by lowering barriers to
entry [9], [10]. A recent notice of proposed rulemaking [11]
for these bands involves 3,850 MHz of spectrum, but does
not move on an additional 12,500 MHz of potentially useful
spectrum in bands above 24 GHz.
Beyond these business concerns, technical properties of
mmWave bands favor spectrum and infrastructure (base station) sharing. While cellular frequencies have traditionally
been allocated with geographic area exclusive use licenses,
the physical characteristics of mmWave signals suggest that
All of the authors are with the Department of Electrical and Computer
Engineering, NYU Tandon School of Engineering.
C. Paper Organization
The rest of this paper is organized as follows. We begin
with a brief introduction to the economic framework used in
this paper, in Section II. In Section III, we describe the system
model and simulation results showing the benefit of resource
sharing in mmWave networks with respect to fifth percentile
rate. Section IV describes mmWave service as a network good
(in the economic sense), and uses simulation results to quantify
the network externalities associated with increasing network
size. We build on results from Section III and Section IV in
Section V to show how demand for mmWave network services
evolves as a service provider increases its network size, and
we compare the likelihood of market entry with and without
open resources such as unlicensed spectrum or an open
deployment of neutral small cells. In Section VI, we describe a
duopoly game involving two vertically differentiated mmWave
network service providers, and compare their profits with and
without resource sharing, for simultaneous market entry and
sequential market entry. Finally, in Section VII, we conclude
with a discussion of the implications of this work and areas
of further research.
II. E CONOMIC FOUNDATIONS
We briefly summarize here the economic framework used
in the rest of this paper. We define a network good and show
how its demand is fundamentally different from demand for
non-network goods, give equilbria and conditions for reaching
critical mass in a market for a network good, explain the
concept of compatibility, and describe a model of vertically
differentiated network good. For a more detailed overview of
this area of economics, see [13].
A. Network goods
In economics, a network good or service [13] is a product
for which the utility that a consumer gains from the product
varies with the number of other consumers of the product
(the size of the network). This effect on utility - which is
called the network externality or the network effect - may be
direct or indirect. The classic example of a direct network
effect is the telephone network, which is more valuable when
the service has more subscribers. The classic example of an
indirect effect is the hardware-software model, e.g. a consumer
who purchases an Android smartphone will benefit if other
consumers also purchase Android smartphones, because this
will incentivize the development of new and varied applications for the Android platform. The network externality may
also be negative, for example, if an Internet service provider
becomes oversubscribed, its subscribers will suffer from the
congestion externality.
p (price)
1.0
1.0
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
n'
n''
n=0
0.0
0.0
0.2
0.4
0.6
0.8
1.0
B
i , and a set of user equipment (UEs) whose locations are
modeled by an independent hPPP with intensity U
i .
Both BSs and UEs use antenna arrays for directional beamforming. For the sake of tractability, we approximate the actual
array patterns using a simplified pattern as in [24], [35]. Let
G() denote the simplified antenna directivity pattern depicted
in Fig. 2, where M is the main lobe power gain, m is the back
lobe gain and is the beamwidth of the main lobe. In general,
m and M are proportional to the number of antennas in the
array and M/m depends on the type of the array. Furthermore,
is inversely proportional to the number of antennas, i.e., the
greater the number of antennas, the more beam directionality.
We let GB () (which is parameterized by M B , mB , and
B ) be the antenna pattern of the BS, and GU () (which is
parameterized by M U , mU , and U ) be the antenna pattern
of the UE.
Fig. 2. Simplified antenna pattern with main lobe M , back lobe m and
beamwidth .
TABLE I
N ETWORK PARAMETERS
Parameter
Value
Frequency
Total bandwidth (W1 + W2 )
B
Total BS density (B
1 +2 )
U)
Total BS density (U
+
1
2
BS transmit power P
B
BS antenna model (M ,mB ,B )
UE antenna model (M U ,mU ,U )
Rate model (, )
UE noise figure Nf
Noise PSD N0
Simulation duration T
73 GHz
1 GHz
100 BSs/km2
500 UEs/km2
30 dBm
(20 dB, -10 dB, 5 )
(10 dB, -10 dB, 30 )
(0.2, 0.5)
7 dB
-174 dBm/Hz
105 slots
B. Scheduling
C. Performance Evaluation
In order to establish benefits of resouce sharing from
a technical perspective, we present simulation results of a
mmWave network with two NSPs (i {1, 2}) operating in
the 73 GHz band. We fix BS transmit power (P ) and loss
factor () as 30 dBm and 0.5, respectively, as in [36]. Table I
shows the specific parameters we use in the simulations. We
assume two different cases: symmetric NSPs and asymmetric
NSPs. In the symmetric case, NSPs are identical in terms of
network resource (BS density, bandwidth) and UE density.
In the asymmetric case, one of the NSPs has more network
resources and UE density than the other. Parameters such as
B
UE and BS densities (U
i , i ), and bandwidth (Wi ) will be
specified for each NSP separately.
1) Symmetric NSPs: Fig. 4 shows the cumulative distribution function (CDF) of the UE rate, for symmetric NSPs, each
with 500 MHz of 73 GHz spectrum licensed for exclusive use,
50 BSs, and 250 UEs in a single square kilometer. All UEs
1.00
BS Sharing Only
Empirical CDF
Full Sharing
0.75
No Sharing
Spectrum Sharing Only
0.50
0.25
0.00
102
101
100
101
benefit the most from when the NSPs pool both spectrum and
BSs. However, both spectrum sharing alone and BS sharing
alone improve UE rate relative to the case where no resources
are shared. BS sharing has a greater effect on rate than
spectrum sharing for UEs outside the coverage range or with
a poor signal quality. For UEs in outage, BS sharing improves
coverage probability. For UEs with a low SNR, there is little
benefit to adding bandwidth because they are in a powerlimited regime. UEs with a good signal quality (high SNR)
are bandwidth-limited and benefit more from spectrum sharing
than from BS sharing. At this BS density (100 BSs total per
square kilometer), the effects of interference are neglible due
to the directional nature of the transmissions, so there is no
negative effect due to spectrum sharing without coordination.
When two NSPs pool their BS and spectrum resources,
they offer UEs a higher data rate. This is consistent with
the results described in [23], [25][27]. However, the early
work on mmWave resource sharing in [23], [25][27] does
not consider NSPs, which we address next.
2) Asymmetric NSPs: Fig. 5 shows the CDF of UE rate
for consumers of two asymmetric NSPs operating in the same
geographic area. The larger NSP has 70% of resources and
subscribers: 700 MHz of 73 GHz spectrum licensed for exclusive use, 70 BSs, and 350 UEs in a single square kilometer.
The smaller NSP has 30% of resources and subscribers: 300
MHz of spectrum, 30 BSs, and 150 UEs.
When there is no resource sharing, the larger NSP gains
market power by offering its subscribers a higher rate than
BS Sharing Only
Empirical CDF
Full Sharing
0.75
No Sharing
Spectrum Sharing Only
0.50
0.25
0.00
102
101
100
101
102
101
100
101
No sharing
B
i
Wi
Sharing
B
i
Wi
No open resources
Open BS deployment
Open
spectrum
ni B
max
ni Wmax
P
B
P jI nj max
n
W
max
j
jI
B
max
ni Wmax
ni B
max
Wmax
P
B
max
P
jI
nj B
max
jI nj Wmax Wmax
9
1.25
0.4
Outage
NLOS to LOS
LOS, increasing signal power
0.2
Round Robin
1.00
0.75
0.50
100
200
300
250
100
150
200
250
0.2
0.0
250
500
750
1000
follows:
20
50
0.4
0.6
0.25
Rate
Rate without interference
0.0
Opportunistic
0.6
500
750
1000
Bandwidth (MHz)
10
0.5
No Open Resources
Open BS Deployment
0.4
Open Spectrum
0.3
0.2
0.1
0.0
0.00
0.25
0.50
0.75
1.00
11
No Open Resources
Open BS Deployment
Open Spectrum
0.5
Fulfilled Expectations
Demand: p(n;n)
Revenue: np(n;n)
0.8
0.4
0.3
0.8
0.8
0.4
0.2
0.4
0.2
0.0
0.2
0.4
0.6
0.8
0.0
Demand at given n e
0.4
0.6
0.2
0.1
0.6
0.6
1.0
0.0
0.2
0.4
0.6
0.8
1.0
0.2
0.0
0.2
0.4
0.6
0.8
1.0
Fig. 10. The fulfilled expectations demand curve p(n; n) and NSP revenue np(n; n) for a mmWave network deployment, where the value of the network
externalities function for a network of size n, h(n), is computed by simulation as the fifth percentile rate of UEs in a network of that size (in Fig. 9). We
consider three scenarios: one where the size of the network, n, is proportional to the number of BSs it has deployed and its spectrum resources (left), one
where 100 BSs are open to all networks and n is proportional to the networks spectrum resources (middle), and one where 1 GHz of open spectrum is used
by all networks and n is proportional to the number of BSs the network has deployed (right).
(2)
(4)
(5)
(6)
n2 =
(7)
n1 =
We can solve (4), (5), (6), and (7) for n1 , n2 , , and , and
thus determine the decisions of the consumers and the market
share of each NSP given pi , qi , i {1, 2}.
12
2
q1
(8)
>
q2
(
)(
2)
n1 =
0<<<
(9)
then both NSPs have market share greater than zero. When
both (8) and (9) hold, then there is a unique Nash equilibrium
in which both NSPs earn non-zero profit. We restrict our
attention to these circumstances, since these are of primary
interest to us.
If (8) and (9) hold and the NSPs do not share resources,
then according to [34] their equilibrium prices p1,N S , p2,N S
are as follows:
#
"
(
1)[2q1 (
)2 q2
(2
)]
> q1
p1,N S = q1 1 +
4q1 (
)2 q2
2
# (10)
"
2
(
1)[q
(
)(
2)
]
1
2
> q2
p2,N S = q2 1 +
4q1 (
)2 q2
2
(11)
q1,N
S =q
q2,N
S =
q(
)2 11
10
(12)
i
3(3
2 +28
202 )
2
2 (7
5)
< q
(13)
p1,S = q1 1 +
> q1
(14)
(4
3)q1
q2
"
#
1)(q
q
)
1
2
p2,S = q2 1 +
> q2
(15)
(4
3)q1
q2
q2,S
=
q1,S
= q
(16)
q(4
3)
< q
7
6
(17)
q1 + q1 n
1 p1 = 0
p1,M =
q1 (
1)
2
(18)
(19)
(20)
q1,M
= q
(21)
).
Fig. 12 shows the market share of each NSP and total market
coverage (i.e., share of consumers who subscribe to either
NSP) under the same set of circumstances. When
(and the
dispersion of consumers willingness to pay) is small, sharing
offers the best overall market coverage. When
is large, the
best market coverage is achieved by not sharing resources. The
value of
at which the benefit of market segmentation begins
to dominate the benefit of network effects is greater when the
intensity of the network effect is high (large ).
VII. C ONCLUSIONS
In this paper, we have connected economic models of the
strategic decision making of cellular network service providers
and subscribers, to detailed simulations of mmWave networks,
with and without resource sharing. While we have confirmed
the benefits of resource sharing from a purely technical view
(without considering the effect on demand), with the economic
analysis we have illustrated that resource sharing is not always
the preferred strategy of service providers, and some kinds
of resource sharing may be preferred over others. We have
shown that open deployments of neutral small cells make it
easier for networks to reach critical mass, encouraging market
entry more than open spectrum would. Furthermore, we have
shown that the leading service provider in a duopoly market
prefers to share resources only when sharing gains are small
or the market is highly segmented. Our technical simulations
of asymmetric service providers have hinted at this, with
13
Monopoly
Profit
No Sharing
2
Sharing
Monopoly
Highend provider
Lowend provider
2
10
10
10
Fig. 11. Profit of each NSP for different cases of the intensity of the network effect, . The dotted line shows the profits of each NSP if they choose not to
share their respective resources (though they may still use any available open resources). The dashed line shows their profits if they share resources. The
solid line shows the profits in a monopoly market, for comparison. The horizontal axis indicates the dispersion of the consumers valuation of the services.
(
q = 1.5)
No Open Resources (0.7)
Market Share
1.00
Monopoly
No Sharing
0.75
Sharing
0.50
Monopoly
0.25
Highend provider
Lowend provider
0.00
2
10
10
10
Total
Fig. 12. Market share of each NSP and total market coverage for different cases of the intensity of the network effect, . The dotted line shows the market
share of each service provider and their total market coverage if they choose not to share their respective resources (though they may still use any available
open resources). The dashed line shows their market shares and total market coverage if they share resources. The solid line shows the market share of a
single firm in a monopoly market, for comparison. The horizontal axis indicates the dispersion of the consumers valuation of the services. (
q = 1.5)
greater gains for the smaller service provider than the market
leader. However with a purely technical approach, one would
conclude that resource sharing is always beneficial (albeit less
beneficial for the market leader), while the economic analysis
with consideration of price and demand in addition to technical
gains has suggested a different conclusion.
We briefly discuss here some assumptions of our approach.
Our results are predicated on an assumed indirect network
effect benefitting consumers subscribing to a large service
provider. That is, we assume that the resources held by a
service provider in a given market scale together with the number of subscribers it serves. Practically, building out physical
infrastructure and licensing spectrum requires a tremendous
capital investment. A service provider is unlikely to build
out a very large network, at great cost, when it has few
subscribers and so a limited revenue stream. For this reason,
we consider it justified to tie the level of investment in the
network - and thus, the size of the network resources - to the
number of subscribers. Another assumption is that consumers
are homogeneous in their preference for one firm or the
other, given their overall valuation of network service, i.e.,
that consumers with the same will make the same choice
between service providers, given their price, network size, and
inherent quality. Actually, consumers and are not identical
in their valuations of competing services. However, despite
this common simplifying assumption, the general economic
framework we have applied in this paper has been empirically
14
[10] Google Inc., Comments of Google Inc. GN Docket No. 14-177, Jan.
2015.
[11] Federal Communications Commission, In the matter of use of spectrum
bands above 24 GHz for mobile radio services, Notice of Proposed
Rulemaking in FCC 15-138, Oct. 2015.
[12] T. S. Rappaport, R. W. Heath Jr., R. C. Daniels, and J. N. Murdock,
Millimeter Wave Wireless Communications. Pearson Education, 2014.
[13] N. Economides, The economics of networks, International journal of
industrial organization, vol. 14, no. 6, pp. 673699, 1996.
[14] N. Economides and C. P. Himmelberg, Critical mass and network size
with application to the US fax market, NYU Stern School of Business
EC-95-11, 1995.
[15] M. L. Katz and C. Shapiro, Network externalities, competition, and
compatibility, The American economic review, vol. 75, no. 3, pp. 424
440, 1985.
[16] N. Economides and F. Flyer, Equilibrium coalition structures in markets
for network goods, Annales dEconomie et de Statistique, pp. 361380,
1998.
[17] J. S. Panchal, R. Yates, and M. M. Buddhikot, Mobile network resource
sharing options: Performance comparisons, IEEE Trans. on Wireless
Communication, vol. 12, no. 9, pp. 44704482, 2013.
[18] S. Hua, P. Liu, and S. S. Panwar, The urge to merge: When cellular
service providers pool capacity, in IEEE ICC 2012, 2012, pp. 5020
5025.
[19] J. Kibilda, P. Di Francesco, F. Malandrino, and L. A. DaSilva, Infrastructure and spectrum sharing trade-offs in mobile networks, in IEEE
DySPAN 2015, 2015, pp. 348357.
[20] E. A. Jorswieck, L. Badia, T. Fahldieck, E. Karipidis, and J. Luo, Spectrum sharing improves the network efficiency for cellular operators,
IEEE Communications Magazine, vol. 52, no. 3, pp. 129136, 2014.
[21] J. Luo, J. Eichinger, Z. Zhao, and E. Schulz, Multi-carrier waveform
based flexible inter-operator spectrum sharing for 5G systems, in IEEE
DYSPAN 2014, 2014, pp. 449457.
[22] J.-C. Guey, P.-K. Liao, Y.-S. Chen, A. Hsu, C.-H. Hwang, and G. Lin,
On 5G radio access architecture and technology [industry perspectives], IEEE Wireless Communications, vol. 22, no. 5, pp. 25, 2015.
[23] A. K. Gupta, J. G. Andrews, and R. W. Heath Jr., On the feasibility of
sharing spectrum licenses in mmwave cellular systems, arXiv preprint
arXiv:1512.01290, 2015.
[24] T. Bai and R. W. Heath, Coverage and rate analysis for millimeter-wave
cellular networks, IEEE Trans. on Wireless Communications, vol. 14,
no. 2, pp. 11001114, 2015.
[25] F. Boccardi, H. Shokri-Ghadikolaei, G. Fodor, E. Erkip, C. Fischione,
M. Kountouris, P. Popovski et al., Spectrum pooling in mmwave
networks: Opportunities, challenges, and enablers, arXiv preprint
arXiv:1603.01080, 2016.
[26] M. Rebato, M. Mezzavilla, S. Rangan, and M. Zorzi, Resource sharing
in 5G mmwave cellular networks, arXiv preprint arXiv:1603.02651,
2016.
[27] , The potential of resource sharing in 5G millimeter-wave bands,
arXiv preprint arXiv:1602.07732, 2016.
[28] G. Li, T. Irnich, and C. Shi, Coordination context-based spectrum
sharing for 5G millimeter-wave networks, in CROWNCOM 2014, 2014,
pp. 3238.
[29] P. D. Francesco, F. Malandrino, T. K. Forde, and L. A. DaSilva,
A sharing- and competition-aware framework for cellular network
evolution planning, IEEE Trans. on Cognitive Communications and
Networking, vol. 1, no. 2, pp. 230243, June 2015.
[30] C. Singh, S. Sarkar, A. Aram, and A. Kumar, Cooperative profit sharing
in coalition-based resource allocation in wireless networks, IEEE/ACM
Trans. Netw., vol. 20, no. 1, pp. 6983, Feb. 2012.
[31] J. Markendahl and B. G. Mlleryd, On co-opetition between mobile
network operators: Why and how competitors cooperate, in 19th ITS
Biennial Conference 2012. ITS, 2012.
[32] D.-E. Meddour, T. Rasheed, and Y. Gourhant, On the role of infrastructure sharing for mobile network operators in emerging markets,
Comput. Netw., vol. 55, no. 7, pp. 15761591, May 2011.
[33] V. Nikolikj and T. Janevski, State-of-the-art business performance
evaluation of the advanced wireless heterogeneous networks to be
deployed for the "TERA age", Wirel. Pers. Commun., vol. 84, no. 3,
pp. 22412270, Oct. 2015.
[34] P. Baake and A. Boom, Vertical product differentiation, network externalities, and compatibility decisions, International Journal of Industrial
Organization, vol. 19, no. 1, pp. 267284, 2001.
[35] A. M. Hunter, J. G. Andrews, and S. Weber, Transmission capacity
of ad hoc networks with spatial diversity, IEEE Trans. on Wireless
Communications, vol. 7, no. 12, pp. 50585071, 2008.
[36] M. R. Akdeniz, Y. Liu, M. K. Samimi, S. Sun, S. Rangan, T. S. Rappaport, and E. Erkip, Millimeter wave channel modeling and cellular
capacity evaluation, IEEE Journal on Sel. Areas in Communications,
vol. 32, no. 6, pp. 11641179, 2014.
[37] S. Shahsavari and N. Akar, A two-level temporal fair scheduler for
multi-cell wireless networks, IEEE Wireless Communications Letters,
vol. 4, no. 3, pp. 269272, June 2015.
[38] Y.-F. Kuo, C.-M. Wu, and W.-J. Deng, The relationships among service
quality, perceived value, customer satisfaction, and post-purchase intention in mobile value-added services, Computers in Human Behavior,
vol. 25, no. 4, pp. 887 896, 2009.