HAN_LAYOUT_Layout 1 11/14/14 2:37 PM Page 6

5G Wireless Backhaul Networks:

Challenges and Research Advances
Xiaohu Ge, Hui Cheng, Mohsen Guizani, and Tao Han

Abstract
5G networks are expected to achieve gigabit-level throughput in future cellular net-
works. However, it is a great challenge to treat 5G wireless backhaul traffic in an
effective way. In this article, we analyze the wireless backhaul traffic in two typical
network architectures adopting small cell and millimeter wave communication tech-
nologies. Furthermore, the energy efficiency of wireless backhaul networks is com-
pared for different network architectures and frequency bands. Numerical
comparison results provide some guidelines for deploying future 5G wireless back-
haul networks in economical and highly energy-efficient ways.

T o meet the challenges of the expected traffic volume

increase in wireless communications (say that of 2020
as compared to the 2010 level [1]), research on the
next (fifth) generation cellular networks (or 5G net-
works) is highly anticipated in the next decade. Moreover,
some potential transmission technologies are emerging to sup-
port 1000× wireless traffic volume increment in future wire-
cell networks with guaranteed quality of service (QoS) and
affordable energy consumption by sustainable systems.
With the exponentially increasing demand for wireless data
traffic in recent years, it is infeasible for current cellular system
architectures to satisfy gigabit-level data traffic in an economi-
cal and ecological way [3]. One of the solutions is the small
cell network, which is densely deployed by self-organizing, low-
less communications. Massive multiple-input multi-output cost, and low-power small cell base stations (SBSs). In early
(MIMO) antenna technology is validated to improve the spec- studies, a low number of small cells is adopted to improve the
trum efficiency in the same frequency bandwidth10–20×. The signal-to-interference-and-noise ratio (SINR) of wireless links
millimeter wave communication technology is explored to be in limited hot areas, embedded in conventional cellular net-
applied to cellular networks, which can provide more than 100 works. In this case, a little burst backhaul traffic originating
MHz frequency bandwidths. Considering wireless signal prop- from small cells can be forwarded into the core network by the
agation characteristics, the massive MIMO antenna and mil- traditional backhual link of cellular networks. When small cells
limeter wave communication technologies will obviously are ultra densely deployed in cellular networks, it is a key
reduce cell coverage [2]. Therefore, small cell networks are problem to forward massive backhaul traffic into the core net-
emerging in 5G networks. In this case, the 5G network is not work. Moreover, there is concern that the large number of
a simple upgrade of its predecessor, by adding additional small cells will cause the signaling load on the network nodes
spectrum and thus boosting the capacity, or replacing it with to increase due to frequent handovers and degraded mobility
advanced radio technologies. It needs rethinking from the sys- robustness due to increased handover and radio link failures
tem and architecture levels down to the physical layer. In [4]. The impact of small cell deployments on mobility perfor-
addition, we need to be able to answer the question of how to mance in Long Term Evolution (LTE)-Advanced systems was
forward hundreds of gigabits backhaul traffic in ultra dense investigated by system-level simulations [5]. Simulation results
implied that the handover optimization technique can effec-
tively decrease the handover failure rate. With rapidly develop-
The corresponding author is Tao Han. ing point-to-point microwave technologies, the wireless
backhaul solution is becoming an attractive alternative for
Xiaohu Ge, Hui Cheng, and Tao Han are with Huazhong University of small cell networks. Based on simulation and measurement
Science and Technology. results, the microwave backhaul technology at high frequencies
was a viable high-performance solution for wireless small cell
Mohsen Guizani is with Qatar University. backhaul links in non-line-of-sight (NLOS) [6]. Moreover, the
high-performance NLOS backhaul link using higher frequen-
The authors would like to acknowledge the support from the Interna- cies compared to sub-5 GHz frequencies can provide higher
tional Science and Technology Cooperation Program of China (Grant antenna gain for similar antenna sizes. This makes it possible
No. 2014DFA11640 and 0903), the National Natural Science Founda- to design small, compact, point-to-point fixed backhaul links
tion of China (NSFC) (Grant No. 61271224, 61301128 and 61471180), with hundreds of gigabits per second throughput. 60 GHz and
NFSC Major International Joint Research Project (Grant No. 70–80 GHz millimeter wave communication technologies for
61210002), the Hubei Provincial Science and Technology Department high-capacity last mile and pre-aggregation backhaul were
(Grant No. 2013BHE005), the Fundamental Research Funds for the explored in [7]. In addition, orthogonal frequency-division mul-
Central Universities (Grant No. 2013ZZGH009 and 2014QN155), and tiplex (OFDM) access passive optical networks were discussed
EU FP7-PEOPLE-IRSES (Contract/Grant No. 247083, 318992 and as the optical technology complement for enabling flexible
610524). cost-efficient hybrid coverage. According to network simula-

6 0890-8044/14/$25.00 © 2014 IEEE IEEE Network • November/December 2014

HAN_LAYOUT_Layout 1 11/14/14 2:37 PM Page 7

Macrocell

Macrocell

W ckh
ba
ire au
les l
Macrocell

s
s
les l
i re hau
SBS W ck
ba
Uplink
Downlink
FTTC
Small cell UE
FTTC
UE: User equipment
(a) (b)

Figure 1. The central solution of 5G wireless backhaul networks: a) the central scenario; b) the logical architecture of the central
scenario.

tion results for demanding urban small cell backhaul applica- central solution. S1 serves as a feeder for user data from the
tion, flexible high-capacity hybrid millimeter wave/optical advance gateway to the MBS, and the advance gateway is the
mobile backhaul networks presented a highly promising entrance to the core network. X2 enables mutual information
approach for future mobile backhaul networks. Coordinated exchange among small cells. The detailed scenario and logical
multipoint (CoMP) technology is adopted in small cells to architecture are illustrated in Figs. 1a and 1b.
decrease the inter-site interference and improve the spectrum The second backhaul solution is marked as the distribution
efficiency. However, additional backhaul traffic is generated solution in Fig. 2. Compared with the central solution in Fig.
due to the possibility of sharing data among cooperative small 1, there is no MBS to collect all backhaul traffic from small
cells. The backhaul bandwidth required by different coordina- cells, and all backhaul traffic is relayed to a specified SBS. In
tion technologies was discussed in [8]. On the other hand, the Fig. 2, all SBSs are assumed to be uniformly distributed in a
energy efficiency of small cell networks is of great concern as given area. The backhaul traffic of an SBS is relayed by adja-
the base station (BS) density will be increased significantly. cent SBSs using millimeter wave communications. All back-
Based on the random spatial network model, the energy effi- haul traffic from adjacent SBSs will be cooperatively
ciency of small cell networks was analyzed in [9]. Numerical forwarded to a specified SBS that is connected to the core
results showed that the energy efficiency of small cell networks network by FTTC links. There are two logical interfaces, S1
critically depends on the BS power consumption model. and X2, which are used for backhaul traffic in the distribution
Different from other studies in 5G networks, we focus on the solution. Functions of S1 and X2 are the same in the central
throughput and energy efficiency of 5G wireless backhaul net- and distribution solutions. The detailed scenario and logical
works considering ultra dense small cells and millimeter wave architecture are illustrated in Figs. 2a and 2b.
communications. In detail, we first configure two typical small
cell scenarios for comparison analysis. Then, we evaluate the
wireless backhaul traffic models based on two typical small cell
Backhaul Traffic Models
scenarios with different spectrum efficiencies. Furthermore, the Although the backhaul traffic comprises different types of
energy efficiency of 5G wireless backhaul networks in two typical traffic in small cell networks, most traffic originates from user
scenarios is analyzed by numerical results. Moreover, the impact data traffic. However, the overhead backhaul traffic caused by
of different frequency bands of wireless backhaul links on the transmission protocols at S1 interfaces and the handover
energy efficiency of 5G backhaul networks is investigated for two backhaul traffic between adjacent small cells are important
typical scenarios. Finally, future challenges of 5G wireless back- parts of the backhaul traffic. In addition, the wireless traffic
haul networks are discussed, and conclusions are drawn. used for management and synchronization is ignored in this
article since that traffic is obviously less than other traffic in
small cell networks. Considering ideal wireless backhaul links
The System Model between small cells and the MBS or the specified SBS, user
With massive MIMO and millimeter wave communication data traffic is assumed to be only related with the bandwidth
technologies, the small cell scenario is an unavoidable solution and average spectrum efficiency in every cell. Without loss of
for the future 5G network. In this article, two small cell back- generality, all small cells are assumed to have the same band-
haul solutions are presented to evaluate the throughput and width and average spectrum efficiency. In this case, the back-
energy efficiency of 5G wireless backhaul networks. The first haul throughput of a small cell is simplified as the production
backhaul solution is marked as the central solution in Fig. 1. of the bandwidth and average spectrum efficiency in the small
A macrocell BS (MBS) is assumed to be located in the macro- cell [10, 11]. According to results in [12], the 10 percent over-
cell center, and SBSs are assumed to be uniformly distributed head backhaul traffic is assumed to be generated at S1 inter-
in the macrocell. All SBSs are configured with the same trans- faces and the 4 percent handover backhaul traffic is assumed
mission power and coverage. In Fig. 1, the wireless backhaul to be generated at X2 interfaces in small cell networks.
traffic of small cells is transmitted to the MBS by millimeter
wave communication links, and then the aggregated backhaul Backhaul Traffic Model in Central Solutions
traffic at the MBS is forwarded to the core network by fiber The backhaul traffic of the central solution includes the
to the cell (FTTC) links. There are two logical interfaces, S1 uplink and downlink traffic in the macrocell and small cells.
and X2, which are used for forwarding backhaul traffic in the The uplink throughput of a small cell is denoted by THcentra
small-up

IEEE Network • November/December 2014 7

HAN_LAYOUT_Layout 1 11/14/14 2:37 PM Page 8

Small cell

Wi
FTTC bac reless
less kha
Wire haul ul
b a c k
SBS

UE
SBS Uplink
Downlink
FTTC

(a) (b)

Figure 2. The distribution solution of 5G wireless backhaul networks: a) the distribution scenario; b) the logical architecture of the
distribution scenario.

= 0.04 ◊ B sccentra ◊ S centra , where B centra is the bandwidth of a

sc sc where P OP is the BS operating power and T lifetime is the BS
small cell and Ssc centra is the average spectrum efficiency of a lifetime. Without loss of generality, the BS operating power is
small cell [11]. The downlink throughput of a small cell is cal- assumed as the linear function of the BS transmission power
small-down = (1 + 0.1 + 0.04) ◊ B sc
culated by TH centra centra ◊ S centra , PTX, which is expressed as POP = a ◊ PTX + b, a > 0 and b >
sc
which is transmitted through the S1 interface of backhaul net- 0. In general, the transmission power depends on the radius of
works. The uplink throughput of a macrocell is denoted by coverage and the signal propagation fading.
macro-up = 0.04 ◊ Bmc
THcentra centra ◊ S centra, where B centra is the macro- To simplify the model derivation, the MBS transmission
mc mc
cell bandwidth and Smc centra is the average spectrum efficiency of power is normalized as P0 = 40 W with coverage radius r0 = 1
a macrocell [10]. The downlink throughput of a macrocell is km [7]. Similarly, the BS transmission power with coverage
macro-down = (1 + 0.1 + 0.04) ◊ Bmc
calculated by THcentra centra ◊ S centra,
mc radius r is denoted by PTX = P0 ◊ (r/r0)a, where a is the path
which is transmitted through the S1 interface of backhaul net- loss coefficient. Furthermore, the BS operating power with
works. Assume that the backhaul traffic is balanced in every coverage radius r is expressed as P OP = a ◊ P 0 ◊ (r/r 0) a + b.
small cell. The total number of small cells in a macrocell is The BS embodied energy includes the initial energy and the
configured as N. The total uplink backhaul throughput of the maintenance energy, which is calculated by EEM = EEMinit +
central solution is calculated as THcentra sum-up = N ◊ TH small-up +
centra EEMmaint.
THcentra
macro-up and the total downlink backhaul throughput of the In the central solution, the system energy consumption is
central solution is calculated as THcentra sum-down = N ◊ TH small-down
centra expressed by
centra
+ THmacro-down. As a consequence, the total backhaul through-
put of the central solution is summed as THsum
centra
+ THsum-down.
centra = THcentra
sum-up
centra
Esystem = EEM
macro
+ EOP
macro
+ N ⋅ EEM
small
+ EOP(
small
)
= macro
EEMinit + EEMma
macro
int + POP
macro
⋅ Tlifetime
macro
(1)
Backhaul Traffic Model in Distribution Solutions
In the distribution solution, adjacent small cells cooperatively +N⋅ ( small
EEMinit + EEMma
small
int + POP
small
⋅ Tlifetime
small
).
forward the backhaul traffic to a specified SBS. Therefore,
not only the channel information but also the user data are Considering the wireless backhaul throughput in the central
shared in adjacent cooperative SBSs. Without loss of generali- solution, the energy efficiency of the central solution is
ty, adjacent cooperative small cells are structured in a cooper- defined as hcentra = THsum
centra/Ecentra .
system
ative cluster, and the number of adjacent small cells in a
cluster is assumed to be K. Without including the specified In the distribution solution, the system energy consumption
SBS that collects all backhaul traffic from adjacent small cells, is expressed by
the spectrum efficiency of a cooperative cluster is denoted by
SscComp = (K – 1)Ssc dist, where S dist is the spectrum efficiency of
sc
dist
Esystem (
= K ⋅ EEM
small
+ EOP
small
)
= K ⋅(E )
the small cell in the cooperative cluster. Considering the (2)
cooperative overhead in the cooperative cluster, the uplink
small
EMinit + EEMma
small
int + POP
small
⋅ Tlifetime
small
.
backhaul throughput of a cooperative small cell is denoted by
small-up = 1.14 ◊ Bsc ◊ Ssc , where Bsc is the bandwidth of
THdist dist dist dist Considering the wireless backhaul throughput in the distri-
the small cell [12]. The downlink backhaul throughput of a bution solution, the energy efficiency of the distribution solu-
dist
cooperative small cell is denoted as THsmall-down = 1.14 ◊ Bsc
dist ◊ tion is defined as hdist = THsum
centra/Ecentra .
system
dist + S comp). Therefore, the total backhaul throughput of a
(Ssc To analyze the energy efficiency of 5G wireless backhaul
sc
distribution solution is summed as THsum dist = K ◊ (THdist
small-up + networks in two backhaul solutions, default parameters are
dist
THsmall-down ). configured as follows: the radius of the small cell is 50 m, the
macrocell radius is 500 m, the bandwidth of the macrocell and
small cell is 100 Mb/s, the average spectrum efficiency of the
Energy Efficiency of 5G Wireless Backhaul macrocell is 5 b/s/Hz [12], and the path loss coefficient is 3.2
for the urban environment [14]. In macrocells, parameters of
Networks BS operating power are configured as a = 21.45 and b =
The energy consumption of cellular networks should include 354.44 W, respectively. In small cells, parameters of BS oper-
the operating energy and the embodied energy [13]. In this ating power are configured as a = 7.84 and b = 71.50 W,
article, the operating energy is defined as EOP = POP ◊ Tlifetime, respectively. The lifetime of MBSs and SBSs are assumed to

8 IEEE Network • November/December 2014

HAN_LAYOUT_Layout 1 11/14/14 2:37 PM Page 9

Wireless backhaul frequencies 5.8 GHz 28 GHz 60 GHz

be 10 and 5 years, respectively [15]. Other parameters amacro 21.45 21.45 21.45
are listed in Table 1.
First, the throughput of wireless backhaul networks bmacro 354 W 354 W 354 W
with respect to the number of small cells considering dif-
macro
ferent average spectrum efficiency is compared in Fig. 3. PTX (coverage radius is 500m) 10 W 233 W 1070 W
In Fig. 3a, the backhaul throughput linearly increases
with the increase of small cell numbers in the central macro
POP (coverage radius is 500m) 568 W 5352 W 23305 W
solution. In Fig. 3b, the backhaul throughput exponential-
ly increases with the increase of small cell numbers in the macro
EEMinit 75 GJ 75 GJ 75 GJ
distribution solution. The exponentially increasing feature
is caused by sharing cooperative traffic among small cells macro
in the distribution solution. When the number of small EEMmaint 10 GJ 10 GJ 10 GJ
cells is fixed, the backhaul throughput increases with the
macro
increase of average spectrum efficiency in small cells. Tlifetime 10 years 10 years 10 years
Second, the energy efficiency of wireless backhaul
networks with respect to the number of small cells con- asmall 7.84 7.84 7.84
sidering different frequency bands is illustrated in Fig. 4.
In Fig. 4a, the energy efficiency of wireless backhaul bsmall 71 W 71 W 71 W
networks logarithmically increases with the increase in
the number of small cells in the central solution. In Fig. small
PTX (coverage radius is 50m) 6.3 mW 147 mW 675 mW
4b, the energy efficiency of wireless backhaul networks
linearly increases with the increase of the number of small
small cells in the distribution solution. When the num- POP (coverage radius is 50m) 71 W 72 W 76 W
ber of small cells is fixed, the energy efficiency of wire-
small small
less backhaul networks decreases with the increase of EEMinit + EEMmaint (percentage in
20% 20% 20%
frequency bands. However, obviously there are energy total energy consumption)
efficiency gaps for 5.8 GHz, 28 GHz, and 60 GHz fre-
quency bands in the central solution. small
Tlifetime 5 years 5 years 5 years
Finally, the energy efficiency of wireless backhaul net-
works with respect to the path loss coefficient consider- Table 1. Parameters of wireless backhaul networks.
ing different small cell radii is shown in Fig. 5. When the
radius of small cells is less than or equal to 50 m, the
energy efficiency of wireless backhaul networks increases with Shannon capacity theory, the increase of path loss coefficients
the increase of the path loss coefficient. When the radius of have a slight attenuation effect on the wireless capacity when
small cells is larger than 50 m, the energy efficiency of wireless the radius of small cells is less than or equal to 50 m. In con-
backhaul networks decreases with the increase of the path loss trast, the increasing path loss coefficients obviously have an
coefficient. The reason for this result is that based on the attenuation effect on the wireless capacity when the radius of

120 6000
Spectral efficiency is 5 b/Hz Spectral efficiency is 5 b/Hz
Spectral efficiency is 10 b/Hz Spectral efficiency is 10 b/Hz
Spectral efficiency is 20 b/Hz Spectral efficiency is 20 b/Hz

100 5000
Backhaul throughput (Gbps)
Energy efficiency (Mb/s/GJ)

80 4000

60 3000

40 2000

20 1000

0 0
0 10 20 30 40 50 0 10 20 30 40 50
Number of small cells Number of small cells
(a) (b)

Figure 3. Throughput of wireless backhaul networks with respect to the number of small cells considering different average spec-
trum efficiencies: a) the centralized scenario; b) the distributed scenario.

IEEE Network • November/December 2014 9

HAN_LAYOUT_Layout 1 11/14/14 2:37 PM Page 10

40 4500
Frequency band is 5.8 GHz Frequency band is 5.8 GHz
Frequency band is 28 GHz Frequency band is 28 GHz
Frequency band is 60 GHz Frequency band is 60 GHz
35 4000

3500
30
Energy efficiency (Mb/s/GJ)

Energy efficiency (Mb/s/GJ)

3000
25

2500
20
2000

15
1500

10
1000

5 500

0 0
0 10 20 30 40 50 0 10 20 30 40 50
Number of small cells Number of small cells
(a) (b)

Figure 4. Energy efficiency of wireless backhaul networks with respect to the number of small cells considering different frequency
bands: a) the centralized scenario; b) the distributed scenario.

small cells is larger than 50 m. When the system energy con- handover in small cells is the second challenge. To solve this
sumption is fixed, the energy efficiency is proportional to the challenge, the concept of a cooperative small cell group is
wireless capacity in wireless backhaul networks. Compared presented to support high -speed user handover among small
with central and distribution solutions in Figs. 5a and 5b, the cells. In this case, multiple small cells must cooperatively
energy efficiency of the central solution is obviously less than transmit traffic to a high-speed user. When the high-speed
the energy efficiency of the distribution solution under the user departs a small cell, other cooperative small cells still
same radius of small cells and the path loss coefficient. cover its track and consecutively transmit the high traffic to
the high-speed user. Moreover, the new small cell is added
into the cooperative small cell group based on the high-speed
Future Challenges user track. However, there are many issues that need to be
In existing studies, the small cell network is presented to satisfy solved to realize this idea, such as how to organize a dynamic
the high capacity requirement for hot areas in cellular net- cooperative cell group and how to decrease the overhead of
works. Therefore, the large wireless traffic is just transmitted in sharing data in the cooperative small cell group.
limited hot areas. In this case, a little burst backhaul traffic can Even if the massive wireless backhaul traffic can be transmit-
be returned directly back to the core network through the con- ted back to the core network with a specified QoS, it is a key
ventional cellular network architecture. With the massive challenge to deploy it in a highly energy-efficient way. Some
MIMO and millimeter wave communication technologies studies indicate that densely deployed low-power BS will
emerging into 5G networks, the cell size of 5G networks has to decrease the energy consumption in 5G networks. However,
become smaller. Furthermore, the large wireless traffic is trans- based on our analytical results, different architectures of back-
mitted in 5G networks. Moreover, the 5G network will become haul networks have different energy efficiency models. For
an ultra dense cell network with small cells. As a consequence, example, in the central solution, the energy efficiency of wire-
it is a great challenge for future 5G wireless backhaul networks less backhaul networks reach a saturation limit when the densi-
to forward massive wireless traffic to core networks in a low- ty of small cells is larger than a specified threshold. Optimizing
cost and highly energy-efficient manner. Some potential chal- the energy efficiency of wireless backhaul networks is very
lenges are presented in the following context. important for future 5G networks. Some potential solutions are
The first challenge is how to design a new backhaul network presented to face this energy efficiency challenge; for example,
architecture and protocols for ultra dense cell deployment sce- hybrid backhaul networks including wireless and fiber links is a
narios. As discussed in the last paragraph, small cells and ultra valuable solution. Moreover, the new sleeping model in small
dense deployment will become the main features in future 5G cells and adaptive power control of SBSs are effective approach-
networks. In this case, the number of small cells will obviously es for saving energy in 5G wireless backhaul networks. Howev-
increase in the unit area. As a consequence, the corresponding er, more studies should be carried out in the near future.
backhaul traffic will increase exponentially at the gateway if the To overcome the above challenges in 5G wireless backhaul
conventional centralized control model is adopted in the 5G networks, some potential research directions are summarized
backhaul network architecture. The massive backhaul traffic to solve these issues:
not only creates congestion but also collapses the backhaul net- • Distribution cell architecture and protocols can be explored
work. It looks as though the distributed control model has to be to forward wireless backhaul traffic in the ultra dense small
adopted in the 5G backhaul network architecture. However, cell networks.
another challenge then is whether or not existing network pro- • Millimeter wave communications are recommended to
tocols can support the massive backhaul traffic of wireless links. transmit the massive backhaul traffic in 5G wireless back-
For high-speed users, overcoming the effect of frequently haul networks.

10 IEEE Network • November/December 2014

HAN_LAYOUT_Layout 1 11/14/14 2:37 PM Page 11

6.5 300
Path loss coefficient α is 3 Path loss coefficient α is 3
Path loss coefficient α is 4 Path loss coefficient α is 4
6 Path loss coefficient α is 5 Path loss coefficient α is 5

5.5 250

5
Energy efficiency (M/ps/GJ)

Energy efficiency (Mb/s/GJ)

4.5 200

3.5 150

2.5 100

1.5 50
30 40 50 60 70 80 90 100 30 40 50 60 70 80 90 100
Radius of small cells (m) Radius of small cells (m)
(a) (b)

Figure 5. Energy efficiency of wireless backhaul networks with respect to the path loss coefficient considering different small cell
radii: a) the centralized scenario; b) the distributed scenario.

• Cooperative small cell groups should be investigated to [9] L. Chang, J. Zhang, and K. B. Letaief, “Energy Efficiency Analysis of Small
solve handover issues in small cell networks. Cell Networks,” Proc. IEEE ICC, June 2013, pp. 4404–08.
[10] J. Robson, “Guidelines for LTE Backhaul Traffic Estimation,” NGMN
• Highly energy-efficient transmission technologies should be White Paper, July 2011, pp. 1–18.
developed to guarantee 5G wireless backhaul networks [11] J. Robson, “Small Cell Backhaul Requirements,” NGMN White Paper,
being deployed in a low-energy-consumption way. June 2012, pp. 1–40.
[12] V. Jungnickel et al., “Backhaul Requirements for Inter-Site Cooperation in
Heterogeneous LTE-Advanced Networks,” Proc. IEEE ICC Wksps.), June
Conclusion 2013, pp. 905–10.
[13] I. Humar et al. , “Rethinking Energy Efficiency Models of Cellular Net-
5G networks are expected to satisfy rapid wireless traffic works with Embodied Energy,” IEEE Network, vol. 25, no. 2, Mar./Apr.
growth. Massive MIMO, millimeter wave communications, and 2011, pp. 40–49.
[14] S. Misra, S. S. Chatterjee, and M. Guizani, “Stochastic Learning
small cell technologies are presented to achieve gigabit trans- Automata-Based Channel Selection in Cognitive Radio Dynamic Spectrum
mission rates in 5G networks. In this article, we study how to Access for WiMAX Networks,” Int’l. J. Commun. Systems, Jan. 2014.
promote 5G wireless backhaul networks in high throughput and [15] C. Khirallah, J. S. Thompson, and H. Rashvand, “Energy and Cost
low energy consumption ways. Two typical small cell scenarios Impacts of Relay and Femtocell Deployments in Long-Term-Evolution
Advanced,” IET Commun., vol. 5, no. 18, July 2011, pp. 2617–28.
are configured to analyze the wireless backhaul traffic in future
5G networks. Furthermore, the energy efficiency of wireless
backhaul networks is compared in two typical small cell scenar- Bigoraphies
XIAOHU GE [M’09, SM’11] (xhge@mail.hust.edu.cn) is currently a professor
ios. Numerical results imply that the distribution solution has with the Department of Electronics and Information Engineering at Huazhong
higher energy efficiency than the central solution in 5G wireless University of Science and Technology (HUST), China. He received his Ph.D.
backhual networks. However, a veritable challenge would degree in communication and information engineering from HUST in 2003. In
indeed emerge if the new distribution network architecture is January 2013, he was honored as a Huazhong scholarship professor. He
serves as an Associate Editor for IEEE ACCESS , Wireless Communications
adopted in future 5G wireless backhaul networks. and Mobile Computing Journal, and others.
References HUI CHENG (hc_cathy@mail.hust.edu.cn) received her Bachelor’s degrees in
[1] T. Taleb, “Towards Carrier Cloud: Potential, Challenges, and Solutions,” information engineering from Wuhan University of Technology, China, in
IEEE Wireless Commun., vol. 21, no. 3, June 2014, pp. 80–91. 2013. She is currently working toward her Master’s degree in communication
[2]N. Bhushan et al., “Network Densification: the Dominant Theme for Wire- and information systems at HUST. Her research interests are in the fields of
less Evolution into 5G,” IEEE Commun. Mag., vol. 52, no. 2, Feb. 2014, mobile backhaul traffic and user mobility models for small cell networks.
pp. 82–89.
[3] M. Chen, S. Mao, and Y. Liu, “Big Data: A Survey,” ACM/Springer MOHSEN GUIZANI [S’85, M’89, SM’99, F’09] (mguizani@ieee.org) is currently
Mobile Networks and Applications , vol. 19, no. 2, Apr. 2014, pp. a professor and Associate Vice President for Graduate Studies at Qatar Uni-
171–209. versity. He received his B.S. (with distinction) and M.S. degrees in electrical
[4] J. Hoydis, M. Kobayashi, and M. Debbah, “Green Small-Cell Networks,” engineering, and M.S. and Ph.D. degrees in computer engineering in 1984,
IEEE Vehic. Tech. Mag., vol. 6, no. 1, Mar. 2011, pp. 37–43. 1986, 1987, and 1990, respectively, from Syracuse University, New York.
[5] T. Yamamoto and S. Konishi, “Impact of Small Cell Deployments on Mobil- His research interests include computer networks, wireless communications and
ity Performance in LTE-Advanced Systems,” Proc. IEEE 24th Int’l. Symp. mobile computing, and optical networking. He currently serves on the editorial
Personal, Indoor and Mobile Radio Commun., Sept. 2013, pp. 189–93. boards of six technical jJournals and is the founder and Editor-in-Chief of the
[6] M. Coldrey et al. , “Non-Line-of-Sight Small Cell Backhauling Using Wireless Communications and Mobile Computing journal published by Wiley
Microwave Technology,” IEEE Commun. Mag. , vol. 51, no. 9, Sept. (http://www.interscience.wiley.com/jpages/1530-8669/). He is a Senior
2013, pp. 78–84. Member of ACM.
[7] D. Bojic et al., “Advanced Wireless and Optical Technologies for Small-
Cell Mobile Backhaul with Dynamic Software-Defined Management,” IEEE TAO HAN [M’13] (hantao@hust.edu.cn) received his Ph.D. degree in communi-
Commun. Mag., vol. 51, no. 9, Sept. 2013, pp. 86–93. cation and information engineering from HUST in December, 2001. He is cur-
[8] D. Samardzija and H. Huang, “Determining Backhaul Bandwidth Require- rently an associate professor with the Department of Electronics and
ments for Network MIMO,” Proc. European Signal Processing Conf., Aug. Information Engineering, HUST. His research interests include wireless commu-
2009, pp. 1494–98. nications, multimedia communications, and computer networks.

IEEE Network • November/December 2014 11