A New Modelling Approach Is Required to Help Mobile Network
Operators Handle the Growing Demand for Data Traffic
Leonardo Lamorgese, Tomas Eric Nordlander and Carlo Mannino
SINTEF ICT, Department of Applied Mathematics, Oslo, Norway
Keywords: Optimisation, Mobile Network Planning.
Abstract: Last year, global mobile data traffic grew by 69%, and similar growth rates are expected in the coming
years. This growth affects the quality of service, and mobile network operators are finding it increasingly
difficult to manage mobile data traffic. To this end, they are drastically increasing the number of sites and
antennae, as well as modernising existing networks. This requires selecting the best antenna locations in
terms of service area coverage, spectrum availability, installation costs, demographics, etc. In addition,
when extending the wireless network with new antennae, the radio-electrical parameter settings of new and
neighbouring antennae require (re)calibration to minimise interference—a process that in principle may
affect the entire network. Moreover, the antennae must connect to the core network and influence it. This
complex optimisation planning problem does not lend itself well to a manual solution approach. Still, these
plans are developed “manually”, with the support of IT tools, through a time-consuming and inefficient
trial-and-error process. Applied optimisation is needed to tackle this problem effectively, but this requires
advancing the state-of-the-art: Most papers focus on solving the different sub-problems independently.
However, these affect each other heavily and they must be considered simultaneously to maximise the
offered service: optimising the location and configuration of new antennae and the configuration of wireless
network radio-electrical parameters, while taking into account access to the core network.
1 INTRODUCTION
Last year, global mobile data traffic grew by 69%
(CISCO, 2015), and similar growth rates are
expected in the coming years
1
. This growth affects
the quality of service, and mobile network operators
(MNOs) are finding it increasingly difficult to
manage mobile data traffic during peak times
(Rivanda, 2015). The telecom community expects
that MNOs will need about 1,000 times today’s
capacity to handle the demand in 2020 (NGMN,
2015). Consequently, MNOs are trying to access
additional bandwidth, which is a scarce and
congested resource. Moreover, MNOs are drastically
increasing the number of sites with antennae, as well
as modernising part of the existing network.
These are challenging planning tasks. To satisfy
evermore-common increases in demand, planners
are required to select appropriate locations for
installing new base stations and to configure radio-
1
Mobile data traffic is expected to grow at a compound annual
growth rate of 57% from 2014 to 2019 (CISCO, 2015).
electrical parameters
2
both of the new and pre-
existing antennae. This involves finding efficient
locations among a large number of candidate sites,
while considering service area coverage, spectrum
availability, installation costs, demographics, etc. In
Figure 1: Wireless and backhaul network.
2
E.g. power emission, transmission frequency, tilt, height, antenna
diagram or type and many more.
402
Lamorgese, L., Nordlander, T. and Mannino, C.
A New Modelling Approach Is Required to Help Mobile Network Operators Handle the Growing Demand for Data Traffic.
DOI: 10.5220/0005814904020407
In Proceedings of 5th the International Conference on Operations Research and Enterprise Systems (ICORES 2016), pages 402-407
ISBN: 978-989-758-171-7
Copyright
c
2016 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
addition, when extending the wireless network
3
with
new antennae, the antennae may interfere with each
other, and the parameter settings of neighbouring
antennae might need recalibration to increase
coverage while trying to minimise interference.
Furthermore, the wireless network must connect
to the core network, via backhaul connections
(wireless, copper or through fibre optic cables) (see
Figure 1). The set of connections in the backhaul
must be adjusted whenever changes in the wireless
network take place or when traffic volumes increase.
Clearly, all above planning decisions affect each
other and they should be taken jointly to maximise
the offered service. Access to the backhaul network
in particular is expected to become a major
bottleneck in the coming years (NGMN, 2012). With
the drastic increase in the number of sites/antennae,
the problem of choosing the best sites, calibrating
the radio-electrical parameters and establishing low-
cost backhaul connections might need to be solved
several times per month for a given area. This is a
complex combinatorial optimisation problem where,
for even a small version with few antennae, the
amount of possible combinations of parameters and
decisions quickly exceeds what a human can
evaluate manually. Infrastructural investments,
whether installing antennae or new connections to
the backhaul network, are generally very costly, so
planners are put under great pressure by MNOs to
minimise overall capital and operational
expenditures. Today, this planning task is carried out
by specialised personnel (mobile network planners)
in a time-consuming and inefficient trial-and-error
process. Most existing planning tools only present
functionalities to assist in the manual planning
process. To our knowledge, the few tools that do use
optimisation techniques for automated planning only
handle simplified versions of the problem (e.g. fewer
parameters).
The paper is organised as follows: Section 2
provides a brief literature background on the
optimisation of mobile network planning. Section 3
describes the current approach used by MNOs and
its limitations, our arguments for using optimization
in the planning process and the research challenges
that have to be overcome to do so. Section 4
concludes the position paper.
3
Normally, this part of the mobile network is called the wireless
access network. For the sake of simplicity, we drop the term
“access” in this paper.
2 BACKGROUND
The literature on optimisation applied to the design
of wireless and mobile networks is very wide. It
dates back to the late 60s/early 70s, when the first
works on frequency assignment appeared. The merit
for introducing mathematical optimisation and graph
colouring techniques for this type of problem is
usually credited to (Metzger, 1970). Since these
early works, devoted to frequency assignment, the
focus has widened to encompass all the
geographical, physical and radio-electrical
parameters of wireless networks. These may include,
e.g. antenna location, tilt, height, transmission
frequencies, power emission, polarisation, diagram,
etc.
Traditionally, the design of a mobile network has
been decomposed into two major sub-problems:
Wireless Network Design (WND). This can
be summarised as the problem of 1) selecting
appropriate locations for new base stations of a
wireless network and 2) establishing radio-
electrical parameters of the old and new
antennae and assigning radio channels so the
demand of service is satisfied and the overall
installation and operational costs are minimised
(e.g., Capone et al. 2006).
Backhaul Network Design (BND). This
consists of choosing the cheapest way to link
the antennae to aggregation nodes (assumed
already available), which provides the interface
to the core network. Connections must meet
quality of service and survivability
requirements (Charnsripinyo and Tipper,
2005). Choosing the most adequate links will
allow MNOs to sustain traffic and avoid the
risk of seeing the backhaul become the
bottleneck of the network (NGMN, 2012).
This conceptual subdivision is reflected both in the
practice and in the literature of mobile network
planning. The literature abounds with models and
solution techniques for WND and BND (treated
independently), which are both very difficult
combinatorial optimisation problems. Since the early
1980s, several optimisation models have been
developed to tackle WND (see Capone et al., 2006),
at the time considered the bottleneck problem. The
benefits of bringing these developments into practice
were already pointed out in (Ceria et al., 1999). In
2005, Dehghan stated that the use of automatic- and
optimisation-oriented planning techniques might
lead to a cost reduction of up to 30%. Even if very
limited, experiences show that the exploitation of
optimisation techniques may produce significant
A New Modelling Approach Is Required to Help Mobile Network Operators Handle the Growing Demand for Data Traffic
403
increases in coverage (see e.g. Atesio, 2000 and
Mannino et al., 2006). Typically, WND is further
decomposed into two sub-problems (Capone et al.,
2006): coverage planning, devoted to the choice of
antenna localisation and radio-electrical parameters,
and capacity planning, often denoted by frequency
assignment (for a survey, see Aardal et al., 2007),
where one wants to find an optimal assignment of
radio channels to minimise overall interference. A
successful attempt to combine the two sub-problems
into a unique optimisation task was carried out in
Mannino et al. (2006) for broadcasting networks and
in D’Andreagiovanni and Mannino (2009) for
WIMAX (Worldwide Interoperability for
Microwave Access) mobile networks. From the
existing literature, the successful approaches to
WND often combine heuristic frameworks with
mathematical optimisation, with emission powers
represented by a continuous variable (as in Amaldi
et al, 2006) or, more recently, by binary variables (as
in D’Andreagiovanni et al., 2013). BND is typically
modelled with mathematical programmes (e.g.
Charnsripinyo and Tipper, 2005; Cox and Sanchez,
2000; Islam et al., 2015; Wu and Pierre, 2003).
These models are then solved either heuristically or
by using exact methods for small instances, as in
Grøndalen et al. (2015). Few authors have attempted
to solve WND and BND as a joint optimisation
problem (e.g. St-Hilaire and Liu, 2011). In general,
such attempts are limited to so-called metaheuristic
approaches, without guarantee on the quality of the
solutions produced.
3 TIME TO MODEL THINGS
DIFFERENTLY
Current Approach and its Limitations. Today,
mobile network planners perform their planning
tasks ‘manually’, usually assisted by the available
commercial tools; however, these only provide very
basic support. From our experience, our literature
research, and discussion with Teleplan Globe AS
4
and Telenor
5
the planners’ current workflow can be
summarised as follows: they iteratively identify
possible locations for antennae, simulate their
instalment and adjust the parameters of the new and
pre-existing antennae accordingly. While choosing
4
Provider of planning software for MNOs (http://
teleplanglobe.no/)
5
Norway's main mobile network operator (http://www.
telenor.com/about-us/global-presence/norway/)
locations, they also check for possible backhaul
connections. This time-consuming task is repeated
until an acceptable result is reached. As mentioned,
finding the best combination of location and
parameter values for all of the network antennae is
actually a very hard optimisation problem.
Furthermore, the locations chosen should allow a
(possibly least-cost) backhaul connection. The sheer
number of possible decisions to explore makes a
‘manual’ approach clearly inefficient. Moreover,
such tasks will in any case soon become impossible
for the current workforce to carry out, given that the
rapid increase in smaller cells (antenna coverage)
constantly requires the installation of new antennae
and the re-configuration of network parameters.
Need for Applied Optimisation. Some planning
software providers claim that they embed
optimisation in their tools. To our knowledge
however, the tools used by practitioners have limited
or no actual optimisation functionality. In addition,
such tools never address the planning of the
backhaul and wireless network jointly.
Optimisation-based planning tools could
substantially reduce the planning time and provide
solutions that are more efficient, allowing MNOs to
keep up with market demand and to enforce their
customer policy at minimal costs, despite the sharp
increase in demand. Such tools would also be
beneficial for the public in general: their use would
result in better coverage, increased capacity, reduced
interference and reduced outages of the selected
sites. Moreover, improved service could be crucial
for critical/emergency services that require a stable
connection. Optimising the layout of network
elements and minimising interference will also lead
to cleaner signals, where the energy usage is
minimal for a given quality of service. Additionally,
the predictability of the network will lead to reduced
operational costs for the MNOs. The savings could
be reinvested into installing new antennae and, in
general, into improving and modernising the
network’s infrastructure. This positive externality
will in turn contribute to improve further the quality
of service and to connect a higher number of users.
However, the limitations in existing systems are
rooted in the status quo of related research. Indeed,
the state-of-the-art in optimisation applied to
telecommunications must be advanced to overcome
these limitations and to achieve the above
improvements.
Applied Research Challenges. In the optimisation
literature (section 2), most approaches have either
tackled somewhat stylised problems or focused on
ICORES 2016 - 5th International Conference on Operations Research and Enterprise Systems
404
specific sub-problems. To have a substantial impact
on the current practice requires designing richer
models and algorithms that are more effective for
the overall problem faced by mobile network
planners. To interact with the planning process and
to be used in practice, such algorithms should be
capable to provide good or optimal solutions to the
overall problem very quickly. More specifically, the
following parts need to be planned jointly:
The location and configuration of new
antennae (WND).
The configuration of wireless network
radio-electrical parameters.
The connection to the core network (BND).
Addressing these effectively requires facing critical
modelling and algorithmic challenges. A unifying
mathematical model to combine WND and BND
must be devised, along with effective algorithmic
schemes to solve the models in a reasonable amount
of time for real-life scenarios. This is a clear
challenge, as, to our knowledge, none of the
methods presented in the literature has ever actually
been applied in a real-life planning process. In
summary, the following advancement in the state-of-
the-art for WND/BND will be needed:
Novel unified models for the joint
optimisation of WND and BND.
A fast and effective algorithmic framework
for solving the unified model in a
reasonable amount of time and producing
good or optimal solutions for real-life
instances.
Modelling to tackling potential planning
scenarios, including different wireless and
backhaul techniques, topologies and a large
numbers of parameters.
Our community must focus on designing such
effective integrated approach. First, this requires
defining a strong mathematical formulation for the
joint WND and BND problems. To this end, a
possible approach is to build upon the pure 0,1
formulations for WND introduced in
D’Andreagiovanni et al. (2013) and to extend them
to cope with joint WND and BND. This allows for
the computation of bounds on the optimal solution
values and helps to assess the quality of solutions at
hand. Second, effective algorithms to solve the
unifying model must be developed. Due to the well-
known computational difficulty of the problem, new
decomposition techniques in conjunction with
classical row-and-column generation schemes must
be considered. These algorithms could include
approximate methods, such as meta-heuristics, to
produce quickly good quality solutions that could be
used as initial feasible solutions to the problem or,
indeed, as final solutions when other methods fail to
improve on these. The ability of optimisation
algorithms has increased dramatically during the last
decades, mainly due to the improved methods and
the increased processing powers of personal
computers (PCs). (Bixby, 2002) argued that
algorithmic and software improvements have played
as large a part as processing power
6
when it comes
to solving large linear programmes faster. During
the period 1987 to 2000, Bixby estimated a speedup
increase of six orders of magnitude in solving
power, where processing power and memory
contributed by three orders of magnitude. The
remaining three orders of magnitude is due to an
improved algorithm: “A model that might have
taken a year to solve ten years ago, can now solve in
less than 30 seconds”. Recent developments in
hybrid, parallel and heterogeneous computing could
allow us to overcome the computational challenges
encountered when moving towards the integration of
the planning problems. For instance, Graphics
Processing Units (GPUs) can be exploited as
computational power. In addition, most PC-based
optimisation algorithms use sequential optimisation
methods, which was not an issue while we had an
exponential increase in processor clock frequency.
However, the modern PC architecture is parallel and
heterogeneous—its multiple cores, programmable
GPUs open up for parallel computing and
heterogeneous computing. Hybrid optimisation
methods, as mentioned above, lend themselves well
to parallelisation, and these have been successfully
applied to solve a number of large scale optimization
problems (Brodtkorb et al., 2013). A possible line of
research to investigate is the use of heterogeneous
computing to plan mobile networks by solving
WND and BND jointly.
A final research question concerns the so-called
self-optimising network (SON). SONs are functions
that allow the network to react to fluctuations in
traffic demand by adjusting antenna parameters in
real time to hand over capacity where needed.
Thanks to this, the network can adapt to serve
several different demand scenarios. This feature is
neglected by current planning models. However,
considering SON capabilities in the optimisation
model would allow the design of better performing
and less costly networks. We believe that one way to
incorporate SON capabilities into planning models is
by means of recoverable robust optimisation
models. Recoverable robust optimisation has been
6
The increase in memory is also beneficial for the algorithms.
A New Modelling Approach Is Required to Help Mobile Network Operators Handle the Growing Demand for Data Traffic
405
recently introduced in (Liebchen et al., 2009), where
they consider several input scenarios along with an
input algorithm capable of partially recovering
deviations from the nominal input. SON capabilities
can thus be interpreted as recovering algorithms.
4 CONCLUSIONS
The rapid growth rates in global mobile data traffic
impacts the quality of service, and MNOs are
finding it increasingly difficult to manage this
traffic. They are drastically increasing the number of
sites, antennae and modernising existing networks—
this is a combinatorial planning problem, where
optimisation-based tools could provide substantial
assistance. Today however, the planning process is
still “manual”. Existing optimisation literature have
either tackled stylised problems or focused on
specific sub-problems. To provide an optimisation-
based planning tool to effectively assist planners, the
optimisation community needs to work with richer
models and algorithms that tackle the overall
problem faced by mobile network planners. This
requires rethinking current approaches. We need to
optimise jointly the location and configuration of
new antennae, the configuration of wireless network
radio-electrical parameters and the connection to the
core network. New models have to be designed,
supported by effective algorithms that fully exploit
recent improvements both in methods and hardware.
The potential benefits of using such optimisation-
based planning tools reach further than just MNOs—
it will have a positive impact on society as a whole.
ACKNOWLEDGEMENTS
We would like to thank Ola Aanstads at Teleplan
Globe AS and Vegard Tingstad at Telenor Norge
AS, and Isabelle Catherine Rebecca Tardy at
SINTEF’s department of Communication Systems
for the discussion and their opinion on this topic.
REFERENCES
Aardal K., S.P.M van Hoesel, A. Koster, C. Mannino, A.
Sassano (2007), ”Models and Solution Techniques for
Frequency Assignment Problems”, Annals of
Operations Research, 153 (1), pp. 79-129.
Amaldi E., A. Capone, F. Malucelli (2003), ”Planning
UMTS base station location: optimization models with
power control and algorithms”, IEEE Transactions on
Wireless Communications, Vol. 2, No. 5 939-952.
Amaldi E., A. Capone, F. Malucelli, C. Mannino (2006),
“Optimization problems and models for planning
cellular networks”, Handbook of Optimization in
Telecommunication, Eds. M. Resende and P. Pardalos,
Springer Science.
Atesio (2000) “Atesio GmbH”. Available from:
http://www.atesio.de/technology/index.html. Accessed
2015-11-02.
Bixby Robert E. (2002). Solving Real-World Linear
Programs: A Decade and More of Progress.
Operations Research 50(1), pp. 3-15.
Brodtkorb, T. Hagen, G.Hasle and C. Schulz. (2013).
GPU Computing in Discrete Optimization. Part I:
Introduction to the GPU. EURO journal on
Transportation and Logistics, 159-186.
Ceria S., C. Mannino, A. Sassano, (1999) Planning Tools
Help Designers Optimize Cellular Network, Wireless
Design, Available from: http://wirelessdesignonline.
com/ doc/planning-tools-help-designers-optimize-
cellul-0001. Accessed 2015-11-01.
Charnsripinyo C., D. Tipper (2005), “Topological design
of 3G wireless backhaul networks for service
assurance”, in Design of Reliable Communication
Networks, 2005, Proceedings. 5th International
Workshop on. IEEE.
Cisco (2015) Visual Networking Index: “Global Mobile
Data Traffic Forecast Update 2014–2019” Available
from: http://www.cisco.com/c/en/us/solutions/
collateral/service-provider/visual-networking-index-
vni/white_paper_c11-520862.html Accessed 2015-10-
06.
Cox L. A., J. R. Sanchez (2000), “Designing least cost
survivable wireless backhaul networks”, Journal of
Heuristics, vol. 6, pp. 525-540.
D'Andreagiovanni F., C. Mannino (2009), “An
optimization model for WiMAX Network Planning”,
WiMAX Network Planning and Optimization, Eds.
Yan ZHANG, Auerbach Publications.
Dehghan S. (2005), “A new approach”, 3GSM Daily 1
(44).
Grøndalen O., O. Østerbø, G. Millstein, T. Tjelta (2015),
“On planning small cell backhaul networks”,
European Conference on Networks and
Communications (EuCNC).
Islam M., A. Sampath, A. Maharshi, O. Koymen, N.B.
Mandayam (2014), “Wireless Backhaul Node
Placement for Small Cell Networks”, Annual
Conference on Information Sciences and Systems
(CISS), 2014.
Liebchen C., M. Lübbecke, R. Möhring, S. Stiller (2009),
“The Concept of Recoverable Robustness, Linear
Programming Recovery, and Railway Applications”,
in Lecture Notes in Computer Science, Vol. 5868, pp.
1-27.
Mannino C., F. Rossi F, S. Smriglio (2006), “The network
packing problem in terrestrial broadcasting.”
Operations Research 54(6), pp. 611–626.
ICORES 2016 - 5th International Conference on Operations Research and Enterprise Systems
406
Metzger, B. H. (1970), “Spectrum management
technique”, presentation at 38th National ORSA
meeting, Detroit, MI.
NGMN Alliance (2012), “Small cell backhaul
requirements”, White Paper. Available from:
https://www.ngmn.org/uploads/media/NGMN_Whitep
aper_Small_Cell_Backhaul_Requirements.pdf.
Accessed 2015-10-26.
Rivanda, (2015), “Addresses Ever Increasing Demand”
Available from: http://rivada.com/addresses-ever-
increasing-demand/. Accessed 2015-10-06.
St-Hilaire M., Shangyun L. (2011), “Comparison of
different meta-heuristics to solve the global planning
problem of UMTS networks.” Computer Networks 55
12, pp. 2705-2716.
Y. Wu and S. Pierre (2003), “Optimization of access
network design in 3g networks”, in Proc. Canadian
Conference on Electrical and Computer Engineering
IEEE CCECE 2003, vol. 2, May 4–7, pp. 781–784.
A New Modelling Approach Is Required to Help Mobile Network Operators Handle the Growing Demand for Data Traffic
407