2012 9th Annual Conference on Wireless On-Demand Network Systems and Services (WONS)
The Effect of Variable Wake Up Time on the
Utilization of Sleep Modes in Femtocell Mobile
Access Networks
Willem Vereecken ∗ , Ivaylo Haratcherev† , Margot Deruyck∗ , Wout Joseph∗ , Mario Pickavet∗ ,
Luc Martens∗ and Piet Demeester∗
∗ Ghent
University - IBBT, Department of Information Technology (INTEC)
Gaston Crommenlaan 8, Bus 201, 9050 Ghent, Belgium
email: Willem.Vereecken@INTEC.ugent.be
web: www.INTEC.ugent.be
† Alcatel-Lucent Bell Labs France, Networking and Networks Domain
Centre de Villarceaux, Route de Villejust, 91260 Nozay, France
modes will become imperative. In this study we investigate
the feasibility of the introduction of sleep modes in femtocell
access networks and we evaluate the options of different sleep
modes and their associated wake up times.
Abstract—Femtocells are considered as an enabler for low
power, high bit rate future mobile access networks. In this study,
we look at them as a technology to cover an area with a high bit
rate connectivity. From the evaluation it is clear that sleep modes
are imperative to maximise the energy efficiency of the mobile
access network. We evaluate the power reduction and wake up
time of different sleep modes and apply them to the model in
order to evaluate the influence of the power consumption of
the mobile access network. We demonstrate that fast wake up
times and low power sleep modes are essential in order to make
femtocells a viable technology for mobile access networks.
Index Terms—Network Power Consumption, Carbon Footprint, Mobile Access Networks, Femtocells, Energy Reduction,
Green ICT
II. S LEEP MODES IN F EMTOCELLS
The largest part of the power consumed in a femtocell is
related to the RF front-end (45%) and the TCXO heater (7%).
Hence, switching these components off reduces the consumed
power by more than 50%. At the same time, waking up the
RF is in the order of few hundred of miliseconds. The TCXO
will indeed take some time to heat back up, but our tests
show that apart from some induced clock drift, there will be
no disruption of femtocell operation.
Based on test results and simulation, we defined some
power-save modes , ordered by ‘depth’. The deeper a sleep
mode is the more power is saved, but the more the cost of
that mode is - i.e. it takes the femtocell additional time to
wake-up.
I. I NTRODUCTION
Energy efficiency is a growing concern in the modern day
telecommunication industry. Increasing energy prices on the
one hand and growing attention for environmental aspects such
as climate change and the associated carbon emissions lead to
a trend in which the energy consumption of technologies needs
to be reduced. In the current network technologies, access
networks are largely consuming most of the energy. Of these
access networks, high bit rate mobile access networks are a
concern [1].
Also, the bit rate demand of mobile users is increasing in
order to enable applications like gaming or video streaming
[2]. This high bit rate demand will only be available to the
user at shorter distances to the base stations and assuring
these bit rates will require more dense mobile access networks.
Since it will be difficult to achieve these densities with large
base stations or so called macrocells, a new technology called
femtocells is being examined.
Femtocells are inexpensive, low RF power base stations
with only a small coverage area and associated reduced
power consumption. On the other hand, covering an area will
require much more devices. If we want to make this kind
of access network energy efficient, the introduction of sleep
978-1-4577-1722-2/12/$26.00 ©2012 IEEE
•
•
•
•
On: The femtocell is in full operation, and is consuming
maximum power. Depending on the design that power is
typically between 8 and 15 Watts.
Stand-by: The femtocell is in ‘light’ sleep and can wakeup quickly. The RF and the TCXO heater are switched
off.
Sleep: The femtocell is in ‘deep’ sleep and needs some
time to wake up. In this mode only the power supply, the
backend connection and the generic CPU core remain
active.
Offline: The femtocell is off and consumes no power.
These wake-up times are summarized in Table I. The power
consumption is expressed as a percentage of the active power
consumption.
63
Mode
On
Stand-by
Sleep
Offline
Wake-up time (s)
N/A
0.5
10
30
Power Consumption
100%
50%
15%
±0
this area, assuring that a user is able to connect to the access
network at any place in the covered area.
As we explained in section III, the highest bit rates are
available closest to a base station. This implies that the main
driving parameter for the coverage design is the bit rate we
want to guarantee for a user.
In this study, we assume the influence of the distance of the
user and the provided bit rate on the power consumption of
the base station to be marginal. Hence, an active femtocell will
have a fixed power consumption. This means the overall power
consumption of the access network (without sleep modes) is
proportional to the assumed base station density. When we
want to cover a certain area, we know that a hexagonal grid
provides the most efficient coverage. Hence, in this study, we
assume a hexagonal grid covering a certain area.
Table I: Wake up times of femtocells in different modes
Coding Scheme:
1/4 QPSK
1/2 QPSK
3/4 QPSK
1/2 16-QAM
3/4 8-QAM
3/4 16-QAM
3/4 64-QAM
Bit Rate (Mbps)
1.3
2.5
3.8
5
5.6
7.5
11.3
Range (m)
51.4
39.5
30.1
24.3
22.2
18.1
11
Table II: Ranges of HSPA
III. HSPA AND PATH L OSS
B. Introducing Sleep Modes
HSPA (High Speed Packet Access) is currently widely
introduced in operator networks [5]. It works in the 2.1 GHz
band but a higher performance is obtained by using improved
coding schemes and refined protocols for communication
between handset and base station. HSPA provides peak data
rates up to 14 Mbps in the downlink. In the downlink (DL),
HSPA supports different coding schemes each corresponding
with a certain bit rate and a certain range. Each coding
scheme consists of a modulation scheme which translates the
binary bit stream into an analogue signal, and a coding rate,
which indicates how many redundant bits will be added per
number of information bits. Based on the used modulation
scheme and coding rate it is possible to determine the bit rate
as shown in Table II. For the uplink (UL), HSPA supports
only one modulation scheme (i.e., BPSK) resulting in a bit
rate of 1.5 Mbps.
When we calculate the base station density when covering
for the highest bit rate of HSPA, we get a density of 3181 base
stations per km2 . In comparison, in most countries, an area is
considered to be densely populated as of approximately 500
people per km2 . Moreover, it is safe to assume that only a
limited number of people will require the highest bit rates.
Therefore, when one wants to use femtocells to cover a
certain area, the use of sleep modes will be indispensible. In
[7] we derived a heuristic to establish the least number of
active base stations required for a given user distribution with
a certain bit rate demand in a given mobile access network.
The principle of the heuristic is to gradually switch on base
stations satisfying as many users as possible in each step. The
effectiveness of the heuristic is based on the fact that, although
the network is designed to cover for a high bit rate, in reality
there will be many users requiring lower bit rates, thus being
able to connect to a base station further away than the nearest
base station. It is important to note that the heuristic is intended
as a means to establish a lower bound for the potential of sleep
modes. The usability in practical situations is impaired by the
requirement of exact knowlegde of a users location.
An example distribution is displayed in Fig. 1. In an access
network with 1951 base stations users were distributed with
a user density of 1000 users per km2 . For the bit rate
requirements, we assumed an exponential distribution based
on the available bit rates:
Each coding scheme also corresponds with a certain
receiver SNR (Signal-to-Noise Ratio) which represents the
SNR at the receiver for a certain BER (Bit Error Rate). This
receiver SNR is taken into account when determining the
maximum allowable path loss P Lmax to which a transmitted
signal can be subjected while still being detectable at the
receiver. The path loss is the ratio of the transmitted power to
the received power. Once P Lmax is known, the corresponding
range can be calculated by using a propagation model. Here,
the ITU-R P.1238 model for a residential environment is
used as this is the most appropriate model for femtocell base
stations and the environment considered. [6].
1
(1)
BRα
The factor α is determined so that 10 % of the users require
a bit rate higher than 5 Mbps. This corresponds with the
assumption that most users require lower bit rates.
We can see in the example that 118 of the 1951 base stations
are active. We denote this as the active base station fraction
FA = 6.0%. Typically the base stations closes to the users
with high demand are switched on and already provide a good
coverage for the users with lower requirements. Additionally,
some coverage holes for low bit rates need to be eliminated,
leading to the optimal active base station distribution.
φ(BR) ∝
It is thus clear that the coding scheme influences both the
bit rate and the range.
IV. M ODEL OF A F EMTOCELL ACCESS N ETWORK
A. Covering an area for a certain bit rate
When designing an access network, the principal question
is how to provide the user with a certain access bit rate. In
mobile access networks, the user can be located at any place
in a certain area. The base stations are then used to cover
64
20
Switched on BS
Switched off BS
BR=1.3 Mbps
BR=2.5 Mbps
BR=3.8 Mbps
BR=5.0 Mbps
BR=5.6 Mbps
BR=7.5 Mbps
BR=11.3 Mbps
400
300
18
16
14
12
200
10
8
100
y(m)
6
0
4
2
−100
0
-2,66E-15 0,3
0,6
0,9
−200
1,2
1,5
1,8
2,1
2,4
2,7
3
vmax (m/s)
Off/On
Sl/On
Sb/On
P/P[on]
−300
Figure 2: Influence of the Maximal Allowed User Speed on the
Number of Base Stations in Offline, Stand-by or Sleep Mode
and the Power Consumption of the Mobile Access Network
−400
−400
−300
−200
−100
0
x(m)
100
200
300
400
Figure 1: Example of the Introduction of Sleep Modes in an
HSPA Femtocell Access Network with Static users. (DU =
1000, P>5M bps = 10%)
bit rate. If we assume that at any point in time any user might
need a high bit rate and needs to obtain it instantaniously, we
need to treat him like a high bit rate requiring user.
On the other hand, the case where a user switches from the
lowest bit rate to the highest available is the most extreme
case and will occur only exceptionally. Moreover, with other
users being active in the network, there is a high probability
the user will be able to connect to a base station with a higher
bit rate than the previously required. Combining this with the
possibility to wake up femtocells within a reasonable amount
of time, it is likely that this problem can be tackled without
having to guarantee immediate access to the highest bit rates
at all times. On the other hand, we consider this out of the
scope of the study performed in this paper.
C. The Influence of Wake Up Times
The calculation of the optimal distribution of active femtocells does not take into account any time constraints. If
waking up and handing over to a different cell would happen
instantaniously, this would not matter. The optimal distribution
can than be applied at any point in time and will be closely
correlated to the one at the previous point in time. However,
as we already demonstrated in section II, there are different
possible sleep modes, each requiring a certain wake up time.
A first factor to take into account is the moving speed of
the users. If a user moves out of range of a cell before the
next cell can wake up, there is a problem. As such, we need
to make sure the the neighbouring cells are in the appropriate
sleep mode. Hence, we define a maximal speed of a user in
the network vmax . In this study, we vary this vmax between
0 and 3 m/s, corresponding to the speed of a running person.
Based on this speed, first of all, we need a denser grid as
the user cannot move out of range of a base station before
the next cell can wake up from stand-by. The distance he can
travel during this time of a few hundreds of miliseconds needs
to be taken from the cell size.
Additionally, based on the distance a user can travel in order
to wake up from sleep or off mode, the right sleep mode needs
to be defined for neighbouring cells.
A different reason why a femtocell could have to change
its operating mode is when a user shifts from a low bit rate
requirement to a high bit rate requirement. Again, this can be
related to the speed with which a user changes its behavior.
This change is however much more difficult to cover for. The
main empowering factor of the sleep mode introduction is the
fact that there is a limited amount of users requiring a high
V. E VALUATION OF THE MODEL
A. Influence of the Maximal Allowed User Speed (vmax )
We simulated active base station distributions vmax
varying between 0 and 3 m/s with a user density
DU = 1000users/km2 . The results are displayed in Fig. 2.
It is immediately clear that, due to the variation of active base
stations and the variation in the distribution, the exact number
of base stations in a certain mode will be difficult to predict.
We can see from the fraction of offline base stations compared
to online base stations, there are almost no offline base stations
at vmax > 1. Hence, the entire mobile access network is filled
with base stations in sleep mode.
On
)
B. Influence of Online The Base Station Density (DB
On
The density of online base stations (DB
) has a large impact
on the number of femtocells in stand-by or sleep mode. We
simulated situations with user densities varying between 25
and 2000 users per km2 . The result is displayed in Fig. 3. We
displayed the density of offline, sleeping and stand-by base
stations in function of the online base station density. Note that
in case of 1000 users per km2 (cfr. section V-A) the online
65
12
2500
10
2500
10
2000
8
2000
8
6
1000
500
0
0
50
100
150
200
250
300
DB
1500
1500
6
4
1000
4
2
500
2
0
0
350
0
0
50
100
DBOn
Off
Sleep
Stand-by
Stand-by / Sleep Linear
P/P[On]
3000
P/P[On]
12
DB
3000
150
200
250
300
350
DBOn
P/P[On]
Off
(a) vmax = 0.5m/s
Sleep
Stand-by
Sleep Linear
Stand-By Linear
P/P[On]
(b) vmax = 1.5m/s
On
DB
Figure 3: Influence of the Online Base Station Density
on the Base Station Density in Offline, Stand-by or Sleep Mode
and the Power Consumption of the Mobile Access Network
base station density is about 215 base stations per km2 , which
is also indicated. Additionally, we displayed the relative power
consumption.
In Fig. 3a we displayed the result when assuming a maximal
on
user speed vmax = 0.5m/s. As of DB
> 75, the number of
sleeping base stations begins to decrease from the linear trend
due to overlaps of sleep mode zones in the network. As of
on
> 125, the number of base stations in stand-by mode also
DB
tends to decrease from the linear trend for the same reason. At
these slow speeds, the power consumption of the total network
is about 4 - 5 times the power consumption of the active base
station.
For higher speeds, for example vmax = 1.5m/s (Fig. 3b), a
higher number of base stations in sleep is required. Hence, the
on
influence of the base station density starts at about DB
≈ 25
on
. As of DB ≈ 125, the network is so dense that the overlap
causes the number of sleeping base stations to be reducing.
When evaluating the power consumption, one sees that for
low densities the power consumption is up to thirteen times
the power consumption of an online base station. For higher
densities this reduces to about five.
power consumption. It is important to both reduce the power
consumption of the sleep modes as well as increase the speed
of taking a femtocell out of sleep mode in order to maximize
the energy saving effect of sleep modes.
ACKNOWLEDGEMENTS
The research leading to these results has received funding
from the European Union Seventh Framework Programme
(FP7/2007-2013) under grant agreement n. 257740 (Network
of Excellence “TREND”) and the IBBT-project Green ICT.
W. Joseph is a Post-Doctoral Fellow of FWO-V (Research
Foundation Flanders).
R EFERENCES
[1] W. Vereecken, W. Van Heddeghem, M. Deruyck, B. Puype, B. Lannoo,
W. Joseph, D. Colle, L. Martens, and P. Demeester, “Power consumption in telecommunication networks: overview and reduction strategies,”
Communications Magazine, IEEE, vol. 49, no. 6, pp. 62 –69, june 2011.
[2] Cisco VNI Mobile, “Global mobile data traffic forecast update 2010 2015,” Cisco, Tech. Rep., 2011.
[3] G. Korinthios, E. Theodoropoulou, N. Marouda, I. Mesogiti, E. Nikolitsa,
and G. Lyberopoulos, “Early experiences and lessons learned from
femtocells,” Communications Magazine, IEEE, vol. 47, no. 9, pp. 124
–130, september 2009.
[4] 3rd Generation Partnership Project: Basestation (BS) radio transmission
and reception (FDD) (Release 6), TS 25.104, 3GPP.
[5] 3rd Generation Partnership Project: Physical layer aspects of UTRA High
Speed Downlink Packet Access (Release 4), TR 25.848 v4.0.0., 3GPP,
March 2001.
[6] ITU-R, “Propagation data and prediction methods for the planning of
indoor radiocommunication systems and radio local area networks in the
frequency range 900 MHz to 100 GHz,” Recommendation P.1238-6, 2009.
[7] W. Vereecken, M. Deruyck, D. Colle, W. Joseph, M. Pickavet, L. Martens,
and P. Demeester, “Evaluation of the potential for energy saving in
macrocell and femtocell networks using a heuristic introducing sleep
modes in base stations,” submitted to EURASIP journal on wireless
communications and networking, august 2011.
VI. C ONCLUSIONS
Femtocells are a promising emerging technology for mobile
access networks. Due to their small scale they consume less
than a macrocell station but also cover a smaller area. As a
consequence, it will be required to be able to introduce sleep
modes in femtocell base stations in order to make a sustainable
deployment possible.
Multiple sleep modes are available. We made a distinction
between offline, sleeping and stand-by mode. The deeper the
sleep mode, the lower the power consumption of the station,
but also the longer it takes to bring the base station online.
It is demonstrated that the power consumption of these
sleep modes has a large impact on the mobile access network
66