Content uploaded by Jon Mark
Author content
All content in this area was uploaded by Jon Mark on Jun 10, 2014
Content may be subject to copyright.
a
Modulation
and coding a
Green
energy
Energy
efficiencly
Hardware
n-
IEEE Wireless Communications • October 2011
58 1536-1284/11/$25.00 © 2011 IEEE
LIN X. CAI AND H. VINCENT POOR, PRINCETON UNIVERSITY
YONGKANG LIU, TOM H. LUAN, XUEMIN (SHERMAN) SHEN, AND JON W. MARK,
UNIVERSITY OF WATERLOO
DIMENSIONING NETWORK DEPLOYMENT AND
RESOURCE MANAGEMENT IN
GREEN MESH NETWORKS
INTRODUCTION
The unprecedented expansion of ubiquitous
broadband communication networks and the
increasing demand of multimedia services have
led to a significant growth in the energy con-
sumption of communication networks. Facing
the fact that the cost of energy continues to rise,
energy sustainability in the future has become
one of the most important research directions in
the information and communication technology
(ICT) industry. To address this challenging issue,
green radio communication networks using
renewable energy sources have been emerging as
a promising solution to achieve sustainable oper-
ation of communication networks.
In general, the development of a green radio
communication network involves interdisci-
plinary research activities spanning multiple
dimensions, including:
• Energy saving hardware and devices
• Energy-efficient communication techniques
• Energy-aware network architecture and proto-
col design
• Energy-friendly software and applications
• The development of alternative eco-friendly
green energy, and so on
Figure 1 shows a block diagram of the solutions
for green communication networks. First, a
green communication network comprises a vari-
ety of electrical equipment, including network
devices, network peripherals, customer electron-
ics, electrical fans or other cooling systems, and
more. The efficiency of the hardware devices as
an organic system plays an essential role in
reducing the energy consumption of the system.
Furthermore, by exploiting advanced communi-
cation techniques (smart antenna, ultra-wide-
band communications, adaptive modulation and
coding, cooperative communications, etc.),
power transmission efficiency can be significantly
improved. Intelligent energy management soft-
ware and applications can also be developed to
allow users to further optimize energy efficiency
of the system, such as energy audit software and
dynamic voltage control. In addition to the
aforementioned, considerable energy reduction
can also be achieved by upgrading the network
architecture, optimizing resource allocation and
network capacity planning, and improving data
transmission, switching, routing protocols, and so
forth. Recently, the United Kingdom’s Mobile
Virtual Center of Excellence (VCE) has under-
taken research to improve the energy efficiency
of a cellular network by reducing the number of
active base stations (BSs) and reallocating their
wireless users, or switching active BSs to operate
at low-frequency bands [1] when the traffic load
is low. It is shown that operators can achieve
70–80 percent energy savings by shifting to low-
frequency operation. They also find that the
deployment of power-efficient small-size femto-
ABSTRACT
In this article, network deployment and
resource management issues are revisited in the
context of green radio communication networks
with sustainable energy supply. It is argued that
under the green network paradigm powered by
renewable energy, the fundamental design crite-
rion and main performance metric have shifted
from energy efficiency to energy sustainability.
As an effort to this end, in this article, new net-
work solutions are proposed with an objective of
improving network sustainability; the proposed
solutions ensure that dynamically harvested
energy can sustain the traffic demands in the
network. Specifically, the placement issue of
green access points (i.e., APs powered by sus-
tainable energy sources) is investigated to meet
the energy and QoS demands of mobile users;
and an adaptive resource management scheme is
proposed to address the unreliability of renew-
able energy in QoS provisioning. It is shown that
by mitigating the energy depletion probability of
green APs, sustainable network performance can
be significantly improved.
The authors argue
that under the
green network
paradigm powered
by renewable
energy, the
fundamental design
criterion and main
performance metric
have shifted from
energy efficiency to
energy sustainability.
TECHNOLOGIES FOR GREEEN RADIO
COMMUNICATION NETWORKS
CAI LAYOUT 10/11/11 1:25 PM Page 58
IEEE Wireless Communications • October 2011 59
cells in a macrocellular network can greatly
reduce the power consumption per user in the
network [2]. Reference [3] presents a new archi-
tecture of cell zooming in mobile cellular net-
works. By adaptively adjusting the cell size
according to the traffic loads and channel condi-
tions of users, energy consumption of a mobile
cellular network can be greatly reduced. Alterna-
tively, [4] designs a cross-layer approach to mini-
mize the network energy consumption by jointly
considering optimal power control, link layer
scheduling, and multihop routing protocols.
Existing solutions largely target minimizing ener-
gy consumption to attain an energy-efficient
communication network. Another important
solution is using renewable and clean energy
sources, such as solar or wind power, to power
off-grid networks [5].
It is expected that sustainable energy sources
will be applied widely to meet the growing user
demands on multimedia services in future wire-
less networks [2]. However, it is important to
note that the renewable energy harvested from
such sources, although sustainable, is highly vari-
able and often unpredictable in terms of avail-
ability and capacity, which makes the network
resource management and traffic scheduling
tasks very challenging when these sources of
energy are applied to power the communication
network. This situation dictates that the focus of
the network design should shift away from mini-
mizing the total energy consumption toward the
energy sustainability of communications, that is,
whether the harvested energy can sustain the
traffic demands and meet the quality of service
(QoS) requirements of end users in the network.
To this end, it is essential to revisit the existing
energy-efficient solutions under the new green
communication paradigm. In summary, by recog-
nizing the distinct dynamic and long-term inex-
haustible nature of sustainable energy sources, it
is beneficial to focus attention on energy sustain-
ability by addressing relevant issues such as net-
work architecture, deployment, capacity
planning, and resource management to achieve
an overall sustainable system.
This article describes the characteristics of
sustainable energy supplies and elaborates on
the fundamental design criteria of a green mesh
network in which mesh access points (APs) are
powered by green energy. Based on the design
criteria, we study the network deployment and
resource management issues. Specifically, we
first investigate how to cost-effectively deploy
APs in a network and use the harvested energy
to fulfill the QoS requirements of users, and
then we present a resource management scheme
to adaptively distribute traffic demands across
the network to achieve the maximal energy sus-
tainability of the green mesh network.
GREEN ENERGY SUPPLY IN NEXT-
GENERATION WIRELESS NETWORKS
Green energy, also referred to as clean or sus-
tainable energy, is energy generated from
sources such as solar, wind, tidal, and geother-
mal power that cause minimal pollution. Such
sources are usually renewable and can be replen-
ished without exhausting finite fuel supplies.
Notably, as solar and wind energy generation are
already relatively mature technologies, they are
good candidates for widespread deployment in
the next-generation wireless networks. It can be
envisioned that green APs, powered by green
energy, can be deployed to connect domestic
wireless users to the Internet, for example, a
home or office network where solar panels or
wind turbines are installed on the roof to harvest
energy and power the communications to the
Internet. Furthermore, a green mesh backhaul
can be formed among multiple green APs. In
this case, each AP would use the recharged
energy to serve the up- and downlink traffics of
the domestic users and forward the traffic of
other APs or wireless local area networks
(WLANs) to the gateway. Therefore, mesh-con-
nected green APs can serve as relay nodes or
network portals and gateways that connect to
other networks and provide ubiquitous broad-
band access for wireless users. With the prolifer-
ation of customer electronics and multimedia
services, mobile users may carry a variety of mul-
timedia applications with diverse QoS demands
in terms of flow throughput, transmission delay,
packet loss, and so on. In addition, the reception
and transmission of multimedia traffic consume
variable energy, which is determined by the
power reception/transmission efficiency of the
underlying communication techniques and the
transmission environment. To provide satisfacto-
ry and continuous services to end users, green
APs need to not only fulfill users’ QoS require-
ments, but also meet the energy demands for
traffic delivery.
To deploy a sustainable network, it is impor-
tant to note that green energy sources are inher-
ently dynamic and unstable; as green energy is
harvested from the nearby environment such as
sunlight and air currents, the underlying energy
source is by nature variable and intermittent,
which leads to varying power output. For
instance, it is well known that a wind turbine
provides intermittent and unreliable power,
whereas a solar panel can supply relatively con-
Figure 1. Solutions for green communication networks.
Voltage
control
Network
device
Energy
audit
Smart
antenna
Routing
Admission
control
Customer
electronic
Modulation
and coding
Cooperative
comm.
Resource
allocation
Green
energy
Energy
efficiencly
Hardware
Software
Commun-
ication
Network
protocol
CAI LAYOUT 10/11/11 1:25 PM Page 59
IEEE Wireless Communications • October 2011
60
tinuous power with varying output throughout
the day and through the seasons. To combat the
intermittent and variable nature of the green
energy supply, a battery to store the harvested
electric power for future use is desirable. That is
to say, a practical solution to the usage of green
energy in future wireless networks is to combine
green energy technology with large-capacity
rechargeable batteries to provide a reliable and
sustainable power supply. In this context, the
research emphasis is on managing the dynami-
cally charged energy in the battery (or energy
buffer) to support the application requirements
of mobile users.
From the perspective of applications, it is
well known that multimedia traffic flows typi-
cally exhibit bursty characteristics, and the
energy used for multimedia transmissions is
therefore also a dynamic process. Note that the
energy consumption of each node includes the
energy used for receiving a message, processing
it, and forwarding it to the next hop. Usually
the receiving and processing energy can be con-
sidered to be constant while the transmission
energy can be adapted to ensure a desirable bit
error rate at the receiver when adaptive coding
and modulation are used. That is, for a given
signal-to-noise ratio requirement, the minimal
energy used for transmitting one bit, denoted
et, should be proportional to the path loss, et∝
dα, where dis the transceiver distance and αis
the path loss exponent. Without loss of general-
ity, in this article we model the energy buffer-
ing (or battery) as a G/G/1 queue with arbitrary
arrivals (charge) and departures (consumption)
of unit energy. The energy buffering evolution
is thus a random process determined by both
energy charging, C(t), and discharging, D(t),
processes, as shown in Fig. 2a. When the ener-
gy charging rate is larger than the discharging
rate, C(t) > D(t), the length of the energy
buffer increases, and vice versa. The energy
buffer evolution is shown in Fig. 2b. When an
AP depletes its energy (i.e., the energy buffer
reaches 0), the AP becomes temporarily out of
service until it is replenished. A simple but
coarse design criterion to ensure the sustain-
able performance of a green mesh network is
E[C(t)] > E[D(t)]; that is, the average charging
rate over time should be greater than the aver-
age energy consumption rate so that all traffic
can be served eventually with the harvested
energy at the AP. However, because of the ran-
dom dynamics in both the energy charging and
discharging processes, APs may deplete or
drain the energy and go out of service from
time to time, even if the long-term recharged
energy is higher than the energy demands from
users. Unavailable APs due to energy depletion
are not able to serve the traffic demands with
intolerable transmission delays and severe
packet losses, and finally degrade the QoS per-
formance of mobile users. Therefore, in a green
mesh network, it is necessary and important to
ensure a low energy depletion probability of
APs, and provide mobile users consistent and
guaranteed services.
NETWORK DEPLOYMENT IN A
GREEN WLAN MESH NETWORK
In general, the development of a green com-
munication network involves network architec-
ture planning, network deployment, and resource
management issues. The main focus in network
planning and deployment is the implementation
cost of the network infrastructure because radio
network controllers (RNCs, i.e., BSs, APs, etc.)
are usually much more expensive and consume
far more power than nomadic and mobile users.
Thus, the foremost issue is how to economically
deploy RNCs (e.g., APs) to meet the QoS
demands of mobile users with the minimum
physical investment. The conventional AP place-
ment problem can typically be modeled as an
optimization problem: to find an optimal set of
APs with the minimum deployment cost to pro-
vide full coverage radio access for all users and
fulfill their QoS demands. Most previous works
on the issue of AP placement mainly focus on
minimizing the total placement cost and/or pro-
visioning biconnectivity between APs and users
without considering the energy efficiency of the
system [6, 7]. Some recent studies jointly consid-
er energy-efficient RNC placement and power
control to minimize the total network energy
consumption by covering mobile users in a given
area [8]. As discussed above, when sustainable
green energy is used to power APs, we need to
revisit the AP placement problem under the new
energy sustainability constraint [9]. Our focus is
no longer on minimizing the energy consump-
tion of APs, since green energy is renewable and
Figure 2. Energy buffer of a green AP: a) energy charging and discharging pro-
cesses; b) energy buffer evolution.
D
(a)
(b)
Energy
buffer
Time t
Rate
Time t
Energy
buffer
Xo
Energy charging
Energy discharging
CAI LAYOUT 10/11/11 1:25 PM Page 60
IEEE Wireless Communications • October 2011 61
sustainable with no extra expense. Instead, we
need to minimize the investment cost of APs,
because a green AP, by employing green energy
technologies, is typically more expensive than a
traditional one. That is, we need to place a mini-
mal number of APs and allocate appropriate
network resources to meet the QoS require-
ments of mobiles users, including both the band-
width and energy requirements. A mobile user is
associated with a green AP if and only if the AP
can allocate its required bandwidth and use the
harvested energy to transmit the downlink and
uplink traffic to and from the mobile user. Let
Udenote the set of mobile users and Adenote
the candidate locations where APs can be
deployed. Overall, the green AP placement
problem can be formulated as the following min-
imization problem:
(1)
where the indicator I(x) equals 1 if x> 0 and 0
otherwise. The binary variable xij equals 1 when
user jis associated with AP i, and 0 otherwise. Rj
denotes the achieved flow throughput of user j,
which should be greater than or equal to the
demanded throughput ^
Rj. Ejis the energy
demand of user j, and E+
iis the energy charging
capacity of APi.
P
(0; x0) is the energy depletion
probability of a green AP with initial energy x0,
which indicates how likely a green AP will
deplete its energy and become unavailable.
In Eq. 1, the objective function is to find a
minimal number of green APs deployed in the
network. The first constraint shows that each
user should be associated with only one AP. The
second constraint specifies the QoS demand of
every user should be satisfied. The following two
constraints stipulate that the demanded energy
consumption of all users served by a green AP
should not exceed its charging capability, and
the energy depletion probability of each AP
should be maintained at a low level, respectively.
Notice that each AP can also adjust its transmis-
sion power to achieve different transmission
rates with different coverages for a given trans-
mission bit error rate requirement. In this case,
the achieved user throughput Rj, the consumed
energy to serve each user Ej, and the network
topology will change accordingly, which results
in a different optimal placement setting {xij|i∈
A, j∈U}. The formulated problem in Eq. 1 is a
mixed integer nonlinear programming problem
(MINLP) which is known to be NP-hard. As
there is no efficient polynomial time solution for
NP-hard problems in general, we need to apply
heuristic algorithms to address the green AP
placement problem.
ADAPTIVE GREEN RESOURCE MANAGEMENT
Resource management plays a prominent role in
improving utilization efficiency of the network
resources and enabling QoS provisioning. Band-
width and energy are two important network
resources. As mentioned earlier, most previous
works on resource management have been con-
cerned with minimizing the energy consumption
on the basis that the energy supply is a fixed and
limited network resource. In a green mesh net-
work in which the energy source is inexhaustible
in the long term but dynamic and unreliable in
the short term, the objective of resource alloca-
tion should also shift to address these fundamen-
tal properties of the green network paradigm.
Similar to the energy sustainability constraint
defined in network deployment, we study how to
adaptively distribute the network traffic across
the network to ensure that the harvested energy
of green APs can sustain the traffic demands
with a minimal energy depletion probability. We
assume that an ideal medium access control pro-
tocol is in place for both inter- and intra-WLAN
communications so that all active nodes are
scheduled for data transmissions in a contention-
free manner. As the network capacity is inher-
ently bounded, we propose a distributed
admission control strategy to strike a balance
between the high resource utilization and desir-
able energy sustainability performance.
TRANSIENT EVOLUTION OF ENERGY BUFFER
To well understand the impact of dynamic ener-
gy charging and discharging processes on the
energy sustainability performance of APs, we
resort to a diffusion or Brownian motion approx-
imation to analyze the transient behavior of the
energy buffer. Diffusion approximation allows us
to approximate the discrete energy buffer size by
a continuous process such that the incremental
change in the energy over a small interval is nor-
mally distributed [10] with the mean and vari-
ance determined by the charging and discharging
processes of the energy in the energy buffer.
Provided the initial energy buffer size, x0, and
the mean and variance of the energy charging
and discharging processes, μa, va, and μs, vs,
respectively, the conditional energy depletion
probability,
P
(0; x0), can be represented as [11]
(2)
where
αand βare referred to as diffusion and drift dif-
fusion coefficients, respectively. Here, Eq. 2 indi-
cates that the energy buffer depletes with
probability 1 when the energy charging rate is
smaller than or equal to the energy consumption
rate. However, even if the mean energy charging
rate is larger than the mean energy discharging
rate, it is still possible that the energy buffer
αμ μ
βμ μ
=+
=−
⎧
⎨
⎪
⎩
⎪
vv
aa ss
as
//,
//.
33
11
P
(; )
,,
exp ,
0
10
2
00
xx
=
≤
−
⎛
⎝
⎜⎞
⎠
⎟
⎧
⎨
⎪
⎩
for
otherwise
β
β
α
⎪⎪
Minimize I x
Subject to xj
ijij
iij
:
:
∑∑
∑
()
=∀
1 ∈∈
≥∀∈
U
U
RR j
jj
ˆ,
iijj i
xE E j≤∀∈
+
∑A,
P
(; ) ,
{,
0
0
0
xj
xij
<∀∈
∈
εA
11} , ∀∈ ∀∈jjAU
To understand the
impact of dynamic
energy charging and
discharging processes
on the energy
sustainability
performance of APs,
we resort to a
diffusion or Brownian
motion approxima-
tion to analyze the
transient behavior of
the energy buffer.
CAI LAYOUT 10/11/11 1:25 PM Page 61
IEEE Wireless Communications • October 2011
62
depletes to 0 due to the variance in energy charg-
ing and discharging processes. Figure 3a plots
the energy depletion probability of an energy
buffer when the charging rate is higher than the
discharging rate. The energy charging intervals
are randomly selected from t
→
a= {1, 2, 3, 4} (in
units of time slots) with a given probability pta =
{0.15, 0.25, 0.3, 0.3}; thus, the mean and vari-
ance of the charging interval are μa= 2.75 and
va= 1.09, respectively. The inter-traffic arrivals
of a flow are exponentially distributed with mean
and variance μs= vs= 11.1. The buffer deple-
tion probability is determined by gradually
adding 12 flows and collecting the results. To
compute
P
(0; x0) in the simulation, we collect the
number of runs until the energy buffer of a node
reaches 0 when the simulation runs 6000 time
slots, and divide it by the total number of runs
(i.e., 1000), and plot the results in Fig. 3a. It can
be seen that P(0; x0) decreases with the initial
energy x0. As the simulation is only conducted
over a limited duration, the simulation results
are conservative and slightly lower than the ana-
lytical results (which converge as time goes to
infinity).
TRAFFIC LOAD DISTRIBUTION
Based on the transient energy buffer analysis, an
adaptive resource management scheme is pro-
posed in this section. To improve the network
sustainability, the network traffic should be
appropriately distributed over multiple relay
paths across the network so as to avoid over-
loading some of the mesh APs and ensure that
the probability that APs deplete their energy is
minimized. Toward this goal, in what follows we
design a relay path selection metric based on the
instantaneous energy level, energy charging
capability, and existing traffic demands at each
AP. We also present a distributed admission
control strategy to further guarantee the energy
sustainability of a green mesh network.
In a green mesh wireless network, multihop
relaying is required when a mobile user commu-
nicates with another mobile user associated with
different APs. In this case, to ensure the energy
sustainability and network connectivity, the traf-
fic should be scheduled on a relay path along
which the APs have the minimum probability of
being out of service (i.e., of depleting their ener-
gy and becoming temporarily unavailable). Based
on the analysis, we proceed with the relay path
selection as follows. A source user first floods a
request that includes the destination user and
the estimated first and second order statistics of
the traffic demands of the flow. When the desti-
nation AP receives the request, it will first calcu-
late the probability of energy depletion, P(0; x0)
in Eq. 2, based on its current energy level and
the accumulated traffic demands. Notice that
P(0; x0) is 1 when β≤0, implying that the AP
will eventually deplete its energy by relaying this
flow. Therefore, to differentiate their weights for
path selection, the APs need to further evaluate
whether their current energy level can sustain
the flow demand within a finite duration. Denote
the energy depletion duration Dand the survival
time of a traffic flow T; that is, the flow is expect-
ed to survive in the network in the subsequent T
time slots. The AP updates its weight according
to the flow request as
(3)
where the indicator Δ(⋅) equals 1 if condition (⋅)
is true and 0 otherwise, and where FD(T; x0)= ∫0
T
fD(t; x0)dt is the probability that an AP depletes
its energy before the flow survival time Texpires.
FD(T; x0) indicates how likely it is that a green
AP can sustain the traffic demand before T
expires. The closed-form density function fD(T;
x0) can also be obtained from the diffusion equa-
tion as a function f(x0, α, β) [11]. Figure 3b plots
the CDF of the energy depletion duration. Basi-
cally, for a smaller x0, an AP is more likely to
deplete its energy in the near future; thus, the
CDF curve shifts to the left.
The destination AP then attaches its weight in
the reply packet backward to the source user.
Upon receiving the reply message, each mesh AP
wx FTx
xf
v=+≤
=+≤
P
P
D
D
(; ) ( ) (; ),
(; ) ( )
00
00
00
0
Δ
Δ
β
β
((; ) ,Tx dt
T
0
0
∫
Figure 3. Energy buffer analysis (μa= 2.75, va= 1.09, μs= 11.1, vs= 11.1): a) Energy depletion probability P(0; x0); b) CDF of D.
Initial energy x0 (unit)
(a)
200
0
Energy depletion probability
0.2
0.4
0.6
0.8
1
40 60 80 100
Energy depletion duration D (slots)
(b)
200
0.1
0
CDF
0.4
0.3
0.2
0.5
0.6
0.7
0.8
0.9
1
40 60 80 100
Ana
Sim
Sim x0=10
Ana x0=10
Sim x0=20
Ana x0=20
Sim x0=40
Ana x0=40
CAI LAYOUT 10/11/11 1:25 PM Page 62
IEEE Wireless Communications • October 2011 63
also updates its weight in Eq. 3, based on the
accumulated load energy consumption over each
link. The path with the minimum energy deple-
tion probability (MEDP), Σvwv, is then selected
for the requested flow. In a multihop network,
the data buffer of the relaying AP may absorb the
traffic variance to some degree, and the output
traffic characteristics may vary. If the estimated
traffic demand changes during time duration T,
APs should update the traffic parameters and the
remaining survival time T, and repeat the afore-
mentioned path selection process. It is also possi-
ble that a mesh AP may need to retransmit a
packet after a random period if its next hop mesh
AP is currently out of service, which may change
the energy consumption statistics of the ongoing
flow. In case P(0; x0) is large, energy consumption
statistics may vary hop by hop, and mesh APs
need to update the energy consumption statistics
and recalculate P(0; x0). However, by minimizing
the energy depletion probability and ensuring a
sufficiently large depletion duration, the probabil-
ity that an AP is out of service can be reduced to
a negligible level.
Figure 4 compares the network lifetime of
three schemes, defined as the maximum dura-
tion that all APs are available until one of the
APs depletes its energy. The minimum energy
(ME) scheme selects a relay path with the mini-
mum energy consumption. The minimum path
recovery time (MPRT) chooses the path with
the minimum cumulative recovery time such that
the total consumed energy can be recovered in
the shortest duration. Thus, MPRT is more like-
ly to select the path with a higher charging rate.
The proposed MEDP distributes traffic along a
path to maintain the minimized energy depletion
probability. It can be seen that the sustainable
performance of ME is lower than those of
MPRT and MEDP as it does not consider the
energy charging capability. The proposed MEDP
outperforms MPRT as the latter considers only
the charging capability of mesh APs, neglecting
the traffic demands and variations in both charg-
ing and discharging processes.
DISTRIBUTED CALL ADMISSION CONTROL
Facing the limited network capacity, call admis-
sion control (CAC) plays a critical role in pro-
visioning satisfactory QoS to the existing users
[12]. In general, admission control is designed
to strike a trade-off between the resource uti-
lization and QoS provisioning. For instance,
when more users are admitted to the network,
they can exploit more network resources to
achieve higher network throughput and utiliza-
tion. Corresponding to that, the network
resources are consumed much faster, making
the residual energy of mesh APs deplete quick-
ly and some APs become unavailable. As a
direct result, individual users would encounter
long service delays, intensive jitter, and packet
losses. Therefore, an effective admission con-
trol strategy is necessary to ensure high
resource utilization and at the same time pro-
vide satisfactory energy sustainability perfor-
mance of the network.
In wireless networks, strict QoS provisioning
is often very difficult and not resource efficient.
Thus, we propose a stochastic QoS provisioning
method to ensure a prescribed level of service
for admitted users by ensuring that the energy
depletion duration, D, is larger than the longest
survival time of traffic flows, ^
D, with a high prob-
ability,
(4)
where 0 < ε<< 1 is an adjustable parameter
that reflects the energy sustainability condition
level. A smaller εimplies a stricter energy sus-
tainability constraint for admitting a new flow.
As such, according to the estimated flow statis-
tics in the request, each mesh AP verifies its
energy sustainability to decide whether its energy
level can sustain the requested traffic demand
based on Eq. 4, given its current residual energy
level x0. An AP only relays and responds to a
message when its energy sustainability condition
satisfies Eq. 4. If the source AP cannot establish
a valid relay path to the destination from the
received response messages, which implies that
one or more APs’ energy supply cannot sustain
the demands of the traffic flow, the source AP
will reject the flow request from the end user. By
upper bounding the energy depletion probability,
satisfactory sustainable network performance
can be achieved.
Figure 5 shows the network lifetime, which is
defined as the maximal duration that all APs are
available until one of the APs depletes its ener-
gy, with and without CAC, respectively. Without
CAC deployed, when more flows join the net-
work, the increased traffic loads will deplete the
energy of APs, and the network lifetime
degrades significantly. By using the proposed
CAC, some flow requests are rejected to guaran-
tee the QoS provisioning of the existing users as
the current energy level of green APs cannot
sustain more traffic demands. Therefore, the
existing traffic demands in the network are
maintained at a certain level to achieve a desir-
able network sustaining performance.
Pr()
DD D
D
≤=<
∫
ˆ(; ) ,
ˆ
ftx dt
0
0
ε
Figure 4. Network life time comparisons of three schemes.
Number of flows
2520
80
100
Network life time (slots)
120
140
160
180
200
220
240
260
30 35 40
MEDP
MPRT
ME
CAI LAYOUT 10/11/11 1:25 PM Page 63
IEEE Wireless Communications • October 2011
64
CONCLUSION AND FUTURE RESEARCH
In this article, we have studied network resource
deployment and management in a green mesh
network where APs are powered by renewable
green energy. We first formulate an AP place-
ment optimization problem under the energy
and QoS constraints. We then propose an adap-
tive resource management scheme to improve
the energy sustainability of the green mesh net-
work, considering variable energy charging capa-
bilities of APs.
With the advances in green energy technolo-
gies, it can be envisioned that green energy will
power more network devices and end-user elec-
tronics in the near future. The distinct character-
istics of green energy supplies (i.e., inexhaustible
in the long term and unstable in the short term)
will shift the emphasis of network design from
the requirement of energy efficiency to energy
sustainability. From this perspective, a revisit of
the existing energy efficiency and management
solutions in different areas, as shown in Fig. 1, is
therefore necessary with changes in the funda-
mental network paradigm. Along this vein, there
remain many open issues that deserve in-depth
investigations.
ENERGY SUSTAINABLE SOFTWARE DEFINED RADIO
Software defined radio (SDR) is a promising
emerging green technology that allows users to
use one hardware to adaptively and opportunis-
tically access multiple radio access networks.
Note that the network designers have incorpo-
rated SDR platforms in the BSs to provide mul-
tiple radio accesses to end users. While previous
works mainly focus on energy efficiency in SDR
configurations at both the user’s end and BSs,
the impact of sustainable energy sources has not
yet been seriously investigated in SDR configu-
ration. The dynamic availability and capacity of
the sustainable energy supply will result in dif-
ferent spectrum sensing and channel access
strategies as different frequency bands exhibit
different propagation characteristics and achieve
diverse power transmission efficiency. Thus,
based on the statistics of the energy supply and
the current energy storage, energy-aware SDR
design and radio access technology should be
jointly considered to ensure sustainable opera-
tion of the network and achieve high throughput
performance.
ENERGY-SUSTAINABLE NETWORK DEPLOYMENT
AND RESOURCE MANAGEMENT
The use of low-power, inexpensive, and small-
size femtocells or picocells has attracted increas-
ing attention in the wireless community. A
femtocell operating in a licensed band typically
uses low transmission power to avoid severe co-
channel interference with other licensed users,
and thus achieve enhanced energy efficiency and
capacity. While the deployment of femtocells in
a range of network scenarios is ongoing, the
development of green femtocells with a sustain-
able energy supply is still in its infancy. With dif-
ferent deployment locations, the harvested
energy would exhibit different statistics. As such,
the placement of green femtocells should not
only cater to the downloading demands of users,
but also consider the sustainability of the net-
work energy supply. While our article sheds light
on the deployment issue of green APs in a wire-
less mesh network, there remain significant chal-
lenging issues in different networks like femtocell
networks. For example, when femtocells are
powered by sustainable energy with diverse
amplitudes, it is important to determine the
appropriate femtocell size and adapt it to the
varying charging rate of energy and traffic
demands to ensure high energy sustainability in
the network. In a macrocell with multiple green
femtocells, it is also important to revisit spec-
trum allocation, power management, CAC, and
QoS management to attain high-performance
sustainable self-organized femtocells.
CROSS-LAYER APPROACH FOR A
SUSTAINABLE SYSTEM
Improving the overall system performance of
future green communication networks is inher-
ently a cross-layer design problem that should be
addressed by applying techniques across the pro-
tocol stack, ranging from hardware implementa-
tion, software design, and energy-efficient signal
processing to communication techniques, includ-
ing link-layer scheduling, medium access control,
network layer routing, transport layer flow con-
trol, and the upper-layer applications, as shown
in Fig. 1. It is important not only to optimize the
parameters residing in different networking com-
ponents and protocol layers, but also to study
the interactions among different functions. For
example, energy audit software provides energy
usage statistics, which can be utilized for energy
management, resource allocation, power control,
and sleep scheduling of green stations, and so
on. On the other hand, when a set of green sta-
tions are powered off or use different power lev-
Figure 5. Network lifetime with and without CAC.
Number of flows
2520
0
50
Network life time (slots)
100
150
200
250
300
350
30 35 40
x0=20, CAC, ε=0.1
x0=20, Non CAC
x0=100, CAC, ε=0.1
x0=100, Non CAC
CAI LAYOUT 10/11/11 1:25 PM Page 64
IEEE Wireless Communications • October 2011 65
els for communication, the network topology
changes, which results in different user energy
demands, multihop routing, and end-to-end QoS
performance of mobile users. In this case, a
cross-layer solution to improve overall system
reliability and sustainability is an interesting yet
challenging issue.
ACKNOWLEDGEMENT
This work was supported in part by research
grants from the Natural Science and Engineer-
ing Research Council (NSERC) of Canada and
in part by an NSERC post-doctoral fellowship.
REFERENCES
[1] P. Grant and S. Fletcher, “Mobile Basestations: Reduc-
ing Energy,” Eng. & Tech. Mag., vol. 6, no, 2, Feb.
2011.
[2] W. Tuttlebee et al., “ Saving the plannet — The Ratio-
nale, Realities and Research of Green Radio,” Int’l.
Transfer Pricing J., vol. 4, no, 3, Sept. 2010.
[3] Z. Niu et al., “Cell Zooming for Cost-Efficient Green Cel-
lular Networks,” IEEE Commun. Mag., vol. 48, no, 11,
Nov. 2010, pp. 74-79.
[4] R. L. Cruz and A. V. Santhanam, “Optimal Routing, Link
Scheduling and Power Control in Multi-hop Wireless
Networks,” Proc. IEEE INFOCOM, vol. 1, San Francisco,
CA, Mar. 2003, pp. 702–11.
[5] A. Sayegh, T. D. Todd, and M. Smadi, “Resource Alloca-
tion and Cost In Hybrid Solar/Wind Powered WLAN
Mesh Nodes,” Wireless Mesh Networks: Architectures
and Protocols, Springer, 2008, pp. 167–89.
[6] L. Xiang and K. Yeung, “Joint Access Point Placement
and Channel Assignment for 802.11 Wireless LANs,”
IEEE Trans. Wireless Commun., vol. 5, no. 10, 2006, pp.
2705–11.
[7] J. Pan et al., “Optimal Base-Station Locations in Two-
Tiered Wireless Sensor Networks,” IEEE Trans. Mobile
Computing, vol. 4, issue 5, Sept. 2005, pp. 458–73.
[8] P. Gonzalez-Brevis et al., “Base Station Location Opti-
mization for Minimal Energy Consumption in Wireless
Networks,” Proc. 2nd Green Wireless Commun. & Net-
works Wksp., Budapest, Hungary, May 2011.
[9] Z. Zheng et al., “Constrained AP placement with Rate
Adaptation in Green WLAN Mesh Networks,” Proc. IEEE
GLOBECOM, Houston, TX, Dec. 2011.
[10] D. R. Cox and H. D. Miller, The Theory of Stochastic
Processes, Chapman and Hall, 1965.
[11] L. X. Cai et al., “Adaptive Resource Management in
Green Mesh Networks with Sustainable Energy Supply,”
Proc. IEEE GLOBECOM, Houston, TX, Dec. 2011.
[12] A. Abdrabou and W. Zhuang, “Stochastic Delay Guar-
antees and Statistical Call Admission Control for IEEE
802.11 Single-Hop Ad Hoc Networks,” IEEE Trans. Wire-
less Commun., vol. 7, no. 10, Oct. 2008, pp. 3972–81.
BIOGRAPHIES
LIN X. CAI (Lincai@princeton.edu) received her M.A.Sc. and
Ph.D. degrees in electrical and computer engineering from
the University of Waterloo, Ontario, Canada, in 2005 and
2009, respectively. She is currently working as a postdoc-
toral research fellow at Princeton University. Her research
interests include green communication and networking,
resource management for broadband multimedia net-
works, and cross-layer optimization and QoS provisioning.
YONGKANG LIU (y257liu@bbcr.uwaterloo.ca) is currently
working toward a Ph.D. degree with the Department of
Electrical and Computer Engineering, University of Water-
loo, Canada. He is currently a research assistant with the
Broadband Communications Research (BBCR) Group, Uni-
versity of Waterloo. His research interests include protocol
analysis and resource management in wireless communica-
tions and networking, with special interest in cognitive
radio networks.
TOM H. LUAN (hluan@bbcr.uwaterloo.ca) received his B.E.
degree in Xian Jiaotong University, China, in 2004 and his
M.Phil. degree in electronic engineering from the Hong
Kong University of Science and Technology, Kowloon, in
2007. He is now pursuing a Ph.D. degree at the University
of Waterloo. His current research interests focus on wired
and wireless multimedia streaming, peer-to-peer streaming,
and vehicular network design.
XUEMIN (SHERMAN) SHEN [F] (xshen@bbcr.uwaterloo.ca)
received his B.Sc. (1982) degree from Dalian Maritime Uni-
versity, China, and his M.Sc. (1987) and Ph.D. (1990)
degrees from Rutgers University, New Jersey, all in electri-
cal engineering. He is a professor and University Research
chair, Department of Electrical and Computer Engineering,
University of Waterloo. His research focuses on mobility
and resource management, wireless body area networks,
wireless network security, and vehicular ad hoc and sensor
networks. He served as an Area Editor for IEEE Transactions
on Wireless Communications and Editor-in-Chief for Peer-
to-Peer Networks and Applications. He is a Fellow of the
Engineering Institute of Canada, a registered Professional
Engineer of Ontario, Canada, and a Distinguished Lecturer
of both the IEEE Vehicular Technology and Communica-
tions Societies.
JON W. MARK [LF] (jwmark@bbcr.uwaterloo.ca) received his
Ph.D. degree in electrical engineering from McMaster Uni-
versity in 1970. In September 1970 he joined the Depart-
ment of Electrical and Computer Engineering, University of
Waterloo, where he is currently a Distinguished Professor
Emeritus. He served as the Department Chairman during
the period July 1984–June 1990. In 1996 he established
the Centre for Wireless Communications (CWC) at the Uni-
versity of Waterloo and is currently serving as its founding
director. He had been on sabbatical leave at the following
places: IBM Thomas J. Watson Research Center, Yorktown
Heights, New York, as a visiting research scientist
(1976–1977); AT&T Bell Laboratories, Murray Hill, New Jer-
sey, as a resident consultant (1982–1983): Laboratoire
MASI, Université Pierre et Marie Curie, Paris France, as an
invited professor (1990–1991); and Department of Electri-
cal Engineering, National University of Singapore, as a visit-
ing professor (1994–1995). He has previously worked in
the areas of adaptive equalization, image and video cod-
ing, spread spectrum communications, computer commu-
nication networks, ATM switch design, and traffic
management. His current research interests are in broad-
band wireless communications, resource and mobility man-
agement, and cross-domain interworking. He is a co-author
of the text Wireless Communications and Networking
(Prentice Hall, 2003). A Fellow of the Canadian Academy of
Engineering, he is the recipient of the 2000 Canadian
Award for Telecommunications Research and the 2000
Award of Merit of the Education Foundation of the Federa-
tion of Chinese Canadian Professionals. He was an editor
of IEEE Transactions on Communications (1983–1990), a
member of the Inter-Society Steering Committee of
IEEE/ACM Transactions on Networking (1992–2003), a
member of the IEEE Communications Society Awards Com-
mittee (1995–1998), an editor of Wireless Networks
(1993–2004), and an associate editor of Telecommunica-
tion Systems (1994–2004).
H. VINCENT POOR [F] (poor@princeton.edu) is the Dean of
Engineering and Applied Science at Princeton University,
where he is also the Michael Henry Strater University Pro-
fessor of Electrical Engineering. His interests include the
areas of statistical signal processing and stochastic analysis,
with applications in wireless networks and related fields.
Among his publications is the recent book Quickest Detec-
tion (Cambridge, 2009). He is a member of the NAE, the
NAS, and the RAE. Recent recognition includes the 2009
Armstrong Award of the IEEE Communications Society, the
2010 IET Fleming Medal, and the 2011 IEEE Sumner
Award.
Improving the overall
system performance
of future green
communication
networks is
inherently a cross
layer design problem
that should be
addressed by
applying techniques
across the
protocol stack.
CAI LAYOUT 10/11/11 1:25 PM Page 65
