See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/261390030

A novel traffic capacity planning methodology for LTE radio network

dimensioning

Conference Paper · January 2011

DOI: 10.1049/cp.2011.0711

CITATIONS READS

8 1,436

4 authors, including:

Chaowei Wang
Beijing University of Posts and Telecommunications
137 PUBLICATIONS   2,183 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Interference suppression in telecommunication systems View project

Coverage and optimization of Mobile Croudsensing View project

All content following this page was uploaded by Chaowei Wang on 23 November 2016.

The user has requested enhancement of the downloaded file.

Proceedings of ICCTA2011

A NOVEL TRAFFIC CAPACITY PLANNING

METHODOLOGY FOR LTE RADIO NETWORK
DIMENSIONING
Jun Gu, Yufeng Ruan, Xi Chen, Chaowei Wang
ZTE Corporation Shanghai 201203, China
[email protected], [email protected], [email protected], [email protected]

Abstract the network has fixed resource (number of

channels in
Traffic capacity planning is a challenging task in
multiple input multiple output (MIMO) & GSM network, etc). In order to cope with the
orthogonal frequency division (OFDM) based long dynamic requirement in more advanced wireless
term evolution (LTE) cellular networks, resulted system, such as LTE, WiMAX, stochastic
from emerging diverse multimedia traffic knapsack is proposed, without considering the
requirement, together with highly open & flexible influence of packet level behavior[3][4]. In [5], a
air interface design in LTE. In this paper, a new systematic capacity estimation methodology for
methodology for dynamic real-time capacity system beyond IMT-2000 provides a way to
planning is proposed for LTE radio network calculate resource requirement in packet switch
dimensioning, based on unified traffic process based wireless network, with the packet level
mechanism, fresh simulation methodology for air characteristics into consideration. In this method,
interface, and smart self-evaluation and the main drawback comes from the fact that it is
optimization. By corresponding software design under the assumption of fixed spectrum efficiency
and implementation, it provides powerful tool for for given scenario, which in fact is not the case due
LTE network planners to get efficient, accurate to the interplay among traffic characteristics,
and professional capacity planning outcome algorithm behavior and network performance.
without much manual effort.
In LTE, due to the introduction of various
Keywords: Long-term evolution (LTE);capacity advanced link level and system level techniques,
planning;traffic; simulation such as flexible bandwidth, OFDM, MIMO, inter-
cell interference coordination (ICIC), inter-cell
1 Introduction power control (PC), frequency domain fast
scheduling (FDFS), the network enables great
Since 2009, LTE technology has attracted great flexibility in higher date rate provision and better
interests from top operators around the world. quality of service (QoS) guarantee[1][6][7], but at
With the enhanced technical flexibility and the same time, it is more complex to gain precisely
improved network capability, LTE shows to be the quantitative insight into system capability under
great momentum for the convergence of cellular different service provision scenarios, especially for
network and internet, which will bring practical LTE network deployment, which requires
revolutionary transform of traffic pattern in professional capacity planning for improved user
cellular networks. Traffic diversity, along with the experience and reduced cost.
flexibility and complexity of LTE air interface,
bring the capacity planning for cellular network Generally, capacity planning is the process to
into new dilemma [1]. determine network topology and configuration
(number of site, MIMO configuration, basic radio
In traditional 2G/3G cellular network, circuit parameter determination, etc), under the
switched voice traffic is the dominant service and constraints of service requirement based on traffic
Erlang formula is the most popular and useful growth prediction; in addition, operator’s strategy
methodology to calculate network capacity [2], should be taken into consideration with highest
given specific network configuration information priority before concrete network design. According
and call blocking probability. When coming into to previous illustration, due to the strong inter-
mixed traffic dimensioning, with diverse resource action of different aspect in LTE cellular networks,
requirement for different traffic type, knapsack traditional methodology for capacity could be
model for multiple traffic capacity planning taken to dynamic LTE system, even if with some
became more popular under the assumption that enhancement.
Therefore, for LTE network, system level network should be provided, scenario definition
simulation became more and more important due (Dense urban, urban, suburban, rural, and so on),
to its capability in investigating systematic population distribution (population density in each
performance, with specific modeling method under scenario), service penetration ratio, traffic type,
certain evaluation target. Among the three popular traffic arrival density, traffic QoS requirement, etc.
simulation methods (static, semi-static, dynamic), In addition, the definition should be taken
dynamic system simulation is the most powerful in operator’s specific strategy into consideration with
revealing accurate system performance. But due to highest priority, which is of vital importance to the
highly complex modeling and low efficiency, it is quantitative definition of above traffic related
prohibitively difficult for network planners to use requirement. 2) Parameter input: two types of
such platform to search for optimum capacity parameters are defined. First class is the basic
planning result. radio and engineer parameters, which includes
System bandwidth, transmission power (total
In this paper, a novel system simulation design power, power allocation/control parameter),
based LTE radio network capacity planning antenna type/pattern, MIMO configuration, and so
methodology is proposed for intelligent optimum on. The second class is optimization parameters
result search, given specific traffic information. It and tuned to search for optimum capacity planning
includes key components of unified traffic solution. Typically, such class mainly refers to
processing, flexible air interface adaptation and inter-site distance and antenna downtilt, and for
simulation and Smart evaluation & optimization. more advanced planning, some radio parameters
And specially, for air interface simulation, (power parameter, bandwidth parameters, etc)
probabilistic interference modeling and user could be chosen as optional optimization
mapping transform multi-cell simulation into parameters.
single cell simulation while keeping system 
characteristics, with dramatic reduction on Before iterated optimization, a unified traffic
simulation complexity. process module is design to transform
aforementioned diverse traffic requirement into
2 Methodology framework illustration uniform form format for convenient air interface
simulation. In this procedure, the complex service
The overall framework of the proposed LTE radio requirement from operator is translated into
network capacity planning methodology is shown simultaneous online user number and QoS
in Figure 1. Generally, this methodology is based requirement for each traffic type and then such
on innovative iterated system simulation to find information is put into the iterated simulation
optimum capacity planning solution, based on process.
specific traffic requirement.
Iterated dynamic simulation based optimization
process is the key of proposed LTE capacity
planning methodology. After determining the
unified traffic requirement and basic engineering
and radio parameters, the optimization parameters
should be initiated. Then optimization target is set
and the whole iteration process starts, which is
followed by dynamic simulation and smart
optimization.

In the dynamic simulation part, firstly, real-time

traffic data is generated based on the information
received from unified traffic processing module,
and the output traffic packet is storage in data
buffer for scheduling. Then CINR modeling and
user mapping the two core components are carried
out. By multi-cell topology and wrap around
relationship construction, together with large
amount of randomly dropped CINR collection
Figure 1 LTE capacity planning methodology users (CCU), CINR distribution is collected in
network wide basis. The CINR collection process
For the input part, it consists of two main elements: is aided by random inter-cell interference selection
1) traffic requirement input: in order to achieve technique, and provides near-to-fact CINR
instructive capacity planning outcome, a complete distribution under predetermined network topology
set of information about traffic prediction in future and system parameters, without the need of
scheduling. Based on the network CINR During LTE network capacity planning process,
distribution, and traffic user number in one cell traffic model varies greatly. The simplest and
calculated from total online user number & site convenient way is to get complete information
number in current iteration, CINR mapped traffic about user density, traffic type/QoS requirement,
users (CMTU) are linked to specific CINR values traffic mix pattern, call arrival density, etc. A
by probabilistic mapping according to CINR sample service requirement table is provided in
cumulative density function (CDF). By such novel Table 2. And such set of information could be
processing, the simulation of one cell could be a translated into simultaneous online user number
representative of the whole network in average and related QoS requirement for subsequent.
sense, hence it is reasonable to execute single cell
simulation during iteration. Such simplification Table 2 Traffic requirement example
brings remarkable computation burden without
much performance loss. After the CINR collection
and user mapping, fast scheduling is execute on
transmission time interval (TTI) basis, and then
link throughput is determined from the relationship
with resource block (RB) number and CINR
through link level simulation under specific
channel model. However, in many network planning cases, we
could only get rather rough information for service
In order to obtain optimum capacity planning requirement, and certain degree enhancement and
solution, above dynamic system level simulation is remodeling should be done before used for
performed in iterated way, with automated simulation process:
evaluation and optimization parameters setting 1) If only aggregate throughput requirement is
before next iteration. The self-evaluation process provided, set of virtual full buffer traffic users
decides whether or not the previous optimization should be modeled. In this case, no explicit
parameters fulfill capacity planning requirement QoS requirement is imposed on user and
under predetermined conditions; the automated during iteration, the data rate constraint is on
search is in charge of optimization setting for next cell basis, the modeled user number and
round simulation and evaluation. In the self- attached full buffer property would be
evaluation process, when the optimum solution is delivered to air interface simulation.
found, iteration stops and capacity planning results 2) If very rough traffic properties (user density,
are delivered. traffic type, etc) are provided, typical QoS
settings for each type of traffic could be used
3 Unified traffic processing as default input. After such remedy and
simultaneous online user number calculation,
In LTE, two class of traffic are defined as shown in the traffic information is transfer to simulation.
Table 1: guaranteed bit rate (GBR) and non-GBR, 3) If complete traffic model information is
and classified into 9 QCIs, with diverse provided, after simultaneous online user
requirement on priority, packet delay and packet number calculation, the traffic information is
loss rate, and impose different requirement on delivered to simulation.
radio resource management during resource
allocation. Some traffic examples for each type are After the processing, the traffic requirement is
listed in the table [1]. input into iterated simulation with uniform pattern:
simultaneous online user and marked QoS
Table 1 QoS Class Identifiers (QCIs) for LTE requirement for each type of traffic. Therefore,
when carried out iteration, no matter what of traffic
(even non-type-service is included), the simulation
could be executed in a unified way without any
change to adapt to diverse input.

4 Flexible air interface and simulation

After finished unified traffic processing,
simultaneous online user number and associated
QoS requirement are delivered for dynamic
iterated system simulation, with uniform format.
Then simulation starts for optimum capacity
planning solution search.
4.1 Simulation parameter setup 4.3 CINR modeling
Before integrated simulation, two types of Carrier to interference and noise ratio (CINR)
parameters should be setup: modeling is the key process for determining
accurate statistic signal quality distribution in
1) Basic parameter
network wide sense, which requires a certain
This class parameter mainly consists of the
number of randomly dropped users in each for
basic network configuration such as system
covered area, especially for uplink because the
bandwidth, transmission power, antenna
interference for uplink comes from terminal, where
configuration, MIMO mode, ICIC mode, etc.
the power is generated. Traditionally, the same
When entering into integration, such
user set if generated for CINR collection and
parameters keep constant.
traffic simulation, which require large number of
2) Optimization parameter
snapshots based simulation to get averaged results
This type of parameter typically includes
under specific scenario and parameter setting.
network topology related configuration: inter-
site distance (ISD), antenna downtilt, etc. For
In this paper, a novel simulation process
advanced capacity planning process, other
independent CINR modeling methodology is
parameters could be chosen optionally for get
proposed to get reliable CINR distribution and
more intensive planning results. Such
simplify subsequent dynamic simulation, with only
parameters are determined automatically by
one round simulation is necessary for each time of
Automated Search module according to last
iteration.
round simulation results evaluation, based on
smart search mechanism.
In the propose scheme, large number of traffic
dependent user are randomly distributed over
4.2 Traffic model adaptation
modeled area, and then:
The aim of traffic model adaptation module is to
1) For downlink
adjust simultaneous online user number and the
In each cell, for every user, reference signal
behavior of generated traffic packet.
received power is calculated based on
1) User number adaptation propagation model, then interference is
For each round of system simulation, a typical derived base on system load set, with wrap
network topology configuration is the ISD and around technique, and finally CINR for each
when different ISD is set, according to given user could be calculated. Based on the user
user density in predetermined planning area, level CINR value, network level CINR
user number in each cell could be calculated distribution is formed. Generally, CINR
based on number of cell number derived from collection process is similar to traditional art,
single cell covered area. CMTU generation is except for the principle for user number
in accordance with the calculated user number. setting over the concerned area.
Specifically, such user number is exclusively 2) For uplink
used for dynamic traffic simulation and has no In the propose method, due to the
relationship with the collection of CINR independence of CINR collection and
distribution. dynamic simulation process, there is no prior
2) Traffic model adaptation information on frequency resource occupation
When number of simultaneous online cell status, which is compulsory to decide whether
users is derived in 1), based on which CMTUs or not specific neighboring cell generates
are generated. According to previous interference upon serving cell, and from which
illustration, different types of traffic user, and hence randomized interferer
requirement would be delivered. In this part, selection (RIS) method is designed to solve
randomly distributed packet size and arrival this problem.
process is explicitly modeled based on traffic
type and associated QoS parameters. The RIS methodology design complies with
generation of data packets is carried in a real- network interference characteristic in statistic
time fashion and the formed packets are put sense, resulting from the random frequency
into radio link layer (RLC) buffer for hopping behavior by dynamic scheduling. In
scheduling. With effectively transmitted data this scheme, for every power controlled user
mount/data rate/delay/jitter detection and in each serving cell, after received power
collection, the statistics of traffic requirement spectrum density (PSD) is calculated, an
satisfaction could be provided and used as interferer is randomly selected in neighboring
indicators for optimization parameter update cell, with interference strength scaled by
and judgment of capacity planning solution system load. Then all interferences are
quality. summed up for derivation of user level CINR.
 The RIS methodology could reveal ICIC fashion to search for optimum capacity planning
impact by imposing constraints on flexibility results, with updated network parameter
of interferer selection. All CCUs’ CINR configuration in each round of simulation.
values are collected to form network CINR Specifically, the independent modeling for CINR
distribution. distribution and traffic simulation enable network
level averaged single cell simulation, which
4.4 User mapping dramatically reduces system complexity, without
much loss for simulation capability.
User mapping is the process of CINR value
assignment for each terminal. In the CINR
modeling module, the network level CINR 6 Conclusions
distribution is collected through CCU behavior
modeling. Based on the collected CINR CDF, the In this paper, a novel LTE network capacity
simultaneous online users are mapped to vertical planning methodology is proposed for flexible,
axis in uniform distributed manner. Specially, such accurate, and efficient network dimensioning,
mapping process is only needed in cell basis and based on unified traffic processing and iterated
the subsequent single cell simulation would be simulation. For the simulation part, new network
effectively represent system level behavior while modeling method dramatically improves system
largely reduce simulation burden compared with efficiency and enables fast implementation
traditional simulation methodology. engineering usage. In addition, the intensively
modeled simulation platform could also be utilized
4.5 Traffic simulation for theoretical research for LTE capacity planning
related topics.
In this part, traffic simulation is carried out in TTI
basis. In each TTI, users are ordered based on
References
associated CINR value, QoS requirement and
buffer status and scheduled according to specific [1] S.Sesia, I.Toufik, M.Baker. LTE-The UMTS
scheduling principle and criteria. Then certain Long Term Evolution-From Theory to
number of RBs is allocated to each user for data Practice. John Wiley & Sons Press, 2009.
transmission. Link level performance is modeling [2] Ajay R Mishra, Advanced Cellular Network
through throughput vs. SINR mapping based on Planning and Optimization. John Wiley & Sons
scenario based link level simulation. Press, 2007.
[3] H.Maral, G.Joshi, A.Karandikar, Downlink
5 Smart evaluation and optimization Erlang Capacity of Cellular OFDMA, IEEE,
2011.
During the iterated simulation, smart evaluation [4] M.Jaber, S.A.Hussain, A.Rouz, Downlink
and optimization update behave as two key Erlang Capacity of Cellular OFDMA,
elements for planning result quality and process FUJITSU Sci.TECH, 2002.
efficiency. Smart evaluation is implemented to [5] J.Huschke, T.Irnich, A.Lappeteläinen,
investigate simulated performance after each round Methodology for estimating the spectrum
simulation, and decide whether or not fulfilling requirements for further developments of
capacity planning target under allowable deviation. IMT-2000 and systems beyond IMT-2000,
If predetermined target is achieved, iteration would WINNER project, 2005.
be stopped and capacity planning results are [6] G.D.A.Monghal, Downlink Radio Resource
delivered for engineering usage. Management for QoS Provisioning in
OFDMA systems, Aalborg University, PHD
If capacity planning target is not satisfied under thesis, 2009.
current optimization parameter configuration, [7] Naizheng Zheng, Multi-Cell Uplink Radio
another round of simulation would be executed, Resource Management- A LTE Case Study,
before which new set of optimization parameter Aalborg University, PHD thesis, 2011.
should be automatically configured. Determination [8] G.Piro, L.A.Grieco , G.Boggia, Simulating
of new parameters is highly dependent on previous LTE Cellular Systems: An Open-Source
performance evaluation. Based on the gap between Framework, IEEE TRANSACTIONS ON
simulated performance and input requirement, VEHICULAR TECHNOLOGY, VOL. 60, NO.
parameters for next round simulation are set 2,2011.
automatically with assistance of fast search
algorithm.

Based on the flexible air interface modeling,

dynamic traffic simulation is carried in iterated

View publication stats