International Journal of Computing & Information Sciences

Performance Evaluation for Wi-Fi Offloading

Schemes in LTE Networks

Thamary Alruhaili1 , Ghadah Aldabbagh1,3, Fatma Bouabdallah2 , Nikos Dimitriou4 , Moe Win3
Computer Science Department, Faculty of Computing and Information Technology, King Abdulaziz
University, Jeddah, Saudi Arabia1,3
Information Technology Department, Faculty of Computing and Information Technology, King Abdulaziz
University, Jeddah, Saudi Arabia2
Laboratory for Information & Decision Systems, Massachusetts Institute of Technology, Cambridge, MA,
USA3
Institute of Informatics and Telecommunications, NCSR Demokritos, Athens, Greece 4
Emails: talruhaili@stu.kau.edu.sa1, fothman1@kau.edu.sa2, galdabbagh@kau.edu.sa1,3,
ghadah@mit.edu1,3, nikodim@iit.demokritos.gr4, moewin@mit.edu3

Abstract: The exponential data traffic growth cannot be satisfied anymore with traditional ways by building new
Base Stations (BS) or leasing new spectrum bands, which is costly from the operator’s point of view. In order to
accommodate with this high data proliferation, the Third Generation Partner Ship Project (3GPP) has introduced
Long Term Evolution (LTE). An LTE network can handle more traffic with reduced cost and complexity, by adding
new spectrum and wider bands. Researchers are more interested in coping with increased capacity by involving a
heterogeneous network along with LTE as a better solution. The focus of this research is to discuss a specific
variation of heterogeneous networks, which is Multi Radio Access Technology (RAT), as multiple RATs can be
operating jointly with an LTE network. The target scenario is to evaluate the performance of incorporating Wi-Fi
Access Points (AP) along with an LTE network. The performance of two Wi-Fi offloading algorithms is compared
with the Optimized Wi-Fi Offloading algorithm proposed by the authors. The first offloading algorithm is Wi-Fi if
coverage and the second is Fixed SNR Threshold.

Keywords: LTE; Wi-Fi Offloading; Network Selection; Capacity; Performance; QoS

1. Introduction laptops and smart phones as in [1], the data traffic

The future cellular network is facing a huge
challenge with the data traffic explosion, as the number
of users is increasing exponentially. The operators need
to find a low cost solution to solve the network
capacity issues, and hence this field has gained
international research attention to study benefits of new
paradigms. In 2008, the LTE revolution was introduced
by 3GPP to the cellular communication field, which
was ratified then in 2010 as an IMT Advanced
technology. International Mobile Telecommunication
Advanced (IMT Advanced) was established by the
International Telecommunication Union (ITU). IMT
systems provide service coverage for wide ranges.
An LTE network is widely deployed in cellular
systems for its reliability in providing a good balance
between wireless coverage and capacity. According to
the uptake of mobile data services driven by tablets,
growth will be tenfold until 2016. Although LTE
macro-cells can provide high speed data, they can’t
fulfill the increased subscribers' demands in high
density places. Heterogeneous networks (Het-Net)
seem to be a promising low cost solution from
operators' point of view. Het-Net incorporates the
coexistence of low power nodes along with a macro
base station to improve coverage in high demand
areas. The Het-Net variations are categorized into a
single RAT multi-tier network and a multi RAT
multi-tier network. The possible single RAT
scenarios are : micro-cell, pico-cell, femto-cell or
fixed relays. For the multi RAT scenario, the possible
network components are: Wi-Fi offload, mobile
hotspots and virtual carrier.
In Het-Nets using another network
technology to deliver data traffic originally supposed
to be delivered by LTE BS is known as data
offloading [1]. The offloading can be more beneficial
when it can balance the load between LTE macro and
the Radio Access Technology (RAT) used. According
to research field, offloading can be categorized into
two categories based on the RAT type: either intra-
RAT Offloading in the single-RAT or inter-RAT
offloading in the multi-RAT. In the intra-RAT
offloading, the data traffic can be offloaded to smaller
cells like pico or femto cells, which they are mostly
operating on the same LTE macro band. On the other
hand, inter-RAT offloading indicates an LTE macro
cooperating with Wi-Fi small cells. The majority of the
research community has focused on intra-RAT
offloading. Inter-RAT offloading attracts a lot of
attention due to its low complexity and cost
deployments, but still it’s in its infancy. Most studies
are heading towards the exploration of possible
benefits and enhancements that can be done through
exploiting Wi-Fi RAT, which is the interest of this
study.
The main focus of this research is to address the
capacity issue in dense LTE networks by offloading
data traffic to Wi-Fi as a complementary network. The
objective is to enhance the performance in terms of
network capacity and reduce the number of outage
users. This would be achieved by serving users
according to their required quality of service (QoS),
based on the traffic type. The authors propose an
algorithm to optimize balancing the traffic load
between the two tiers, LTE tier and Wi-Fi tier. The
load balancing takes into account the maximum
throughput, minimum path loss and network load for
each tier. This optimized offloading algorithm is
compared with two different offloading algorithms,
Wi-Fi if coverage and Fixed SINR threshold. This is
done through simulation in order to evaluate the
performance of the three algorithms.
The rest of the paper is organized as follows:
Section 2 addresses the state of the art in the Wi-Fi
offloading field, and presents the effective capacity
Wi-Fi offloading algorithms in the literature. Section 3
explains the system model and the target scenario of
the study. Section 4 describes Network Selection
Schemes (NSS), which are the three compared
algorithms: Wi-Fi if coverage, Fixed SNR Threshold,
and Optimized Wi-Fi Offloading algorithm proposed
by the authors. Section 5 presents the simulation setup
along with numerical results. Section 6 concludes the
study and presents ideas for future work. Finally
acknowledgement is in section 7.
2. Related Work & State of the Art
In this section we review the state of the art in Wi-
Fi traffic offloading in two parts. The first part will
present LTE and Wi-Fi integration levels, while the
second part will demonstrate effective Wi-Fi
offloading algorithms in the research field.
2.1 Wi-Fi Offloading Integration Levels
The tremendous growth of data traffic in
mobile networks is an important driver for employing
the data offloading concept. Using offloading
scenarios in congested areas may help in relieving the
mobile macro base stations and improving the
customers QoS. This is achieved by enhancing the
capacity for each user within the coverage area [2, 3].
Utilizing Wi-Fi is considered one of the possible
scenarios in inter-RAT offloading. Wi-Fi stands for
Wireless Fidelity in the research community, which
is an IEEE 802.11 standard. Unlike mobile stations,
Wi-Fi AP has the ability to provide high data rates in
indoor environments with limited mobility. There is a
tendency to shift Wi-Fi networks toward ubiquity and
use it for offloading mobile data in outdoor [4].
Wi-Fi IEEE 802.11 can use either a 2.4 or 5 GHz
frequency band depending on the standard type used.
There are 14 channels designated in the 2.4 GHz
range spaced 5 MHz apart, with the exception of a
12 MHz spacing before channel 14. There are 3
channels called the non-overlapping channels, which
they are channels 1,6,11. The Wi-Fi IEEE 802.11
family has different standers: a, b, g, n, used with
different bands: 20 MHz or 40 MHz. IEEE 802.11n
is a new standard added to the WLAN family with a
better throughput as it supports MIMO technology.
Wi-Fi uses Carrier Sense Multiple Access with
Collision Avoidance (CSMA/CA) method, as all
mobile stations demands a connection to an AP, are
contending to gain access to a common channel [5].
There are three main levels of integration
between Wi-Fi network and cellular networks, which
are: un-managed data offloading, managed data
offloading and integrated offloading as presented in
Table 1. The first approach is un-managed data
offloading or bypass offloading, considered the
easiest type of offloading, where data is directed to
Wi-Fi whenever the coverage is found with no need
for equipment installment. The voice services in this
type of offloading will remain on the mobile core
network. This immediate offloading solution suffers
from different issues such as: 1) the operators will
lose visibility, control on their own subscribers. 2) the
operators will not be able to send subscribed content,
which leads to lost revenue [2, 4].
The second approach, managed data
offloading, is used by the operators who don’t want
to lose control of their subscribers, however they are
not allowed to send subscribed content. The third
offloading approach is integrated data offloading,
which empowers the operators with full control over
their subscribers along with the ability to send
subscribed content. The drawback of the latter
offloading, in order to establish the data flow forming
a bridge between a cellular network and a Wi-Fi
network is needed. Integrated offloading raises the
coupling architecture concept for Wi-Fi with cellular
systems, which can be divided into two coupling
 architectures: loose coupling and tight coupling [2, 4]. operating on different bands they don’t suffer from
co-channel interference [2, 4, 7, 8].
Offloading Adaption Coupling
Advantage Disadvantage Network access selection or offloading decisions,
type approach architecture
1)whenever are made based on different parameters with
there is Wi- different objectives as well, such as boosting the
Fi coverage capacity or user QoS [2, 6, 9]. Different algorithms
move data to
it.
1) operators have been tackled this in research; the first one is Wi-
lose control Fi if coverage, which considers the SNR value to
Unmanaged 2)bypasses no need
on
data the core
subscribers.
for
----
trigger the offloading decision to Wi-Fi [1]. This
offloading network . network algorithm can end with overloaded Wi-Fi APs with a
[2, 4] 2)operators
3)QoS for equipment
losses poor user throughput at high traffic loads. A similar
voice
revenue. offloading algorithm, usually adapted by mobile
services
delivered by operators is presented in [10], which is a Wi-Fi First
core algorithm (WF). It depends on connecting the users
network first to Wi-Fi, whenever there is Wi-Fi coverage, and
Through
only joins LTE if there is no availability of Wi-Fi.
Managed 1) operators placing
1)operators The authors studied WF performance under non-
data have control intelligent
offloading
losses ---- uniform and a uniform Wi-Fi backhaul capacity
on session a
[2, 4] revenue
subscribers ware gate distribution, and the result shows that WF is worse in
way terms of fairness and average throughput per user.
1) operators 1) need to
have control form a Form a Another offloading algorithm using WLAN is
Integrated on coupling bridge 1)Loose mentioned in [1], which is Fixed SNR Threshold. In
data subscribers bridge between Coupling
offloading 2)operators between Wi- networks 2)Tight
this algorithm the network offloading is based on
[2, 4, 6] don’t lose Fi and to allow Coupling choosing the best SNRmin for WLAN APs, which is
revenue cellular data flow close to Wi-Fi if coverage concept only at low traffic
system loads. The SNRmin in Fixed SNR Threshold is a
Table 1. Characteristics of Wi-Fi integration levels function of the WLAN load. In every WLAN AP the
In loose coupling architecture, there is no need for SNRmin increases individually whenever the load
major cooperation between a Wi-Fi network and a increases as well; otherwise the user will connect to
cellular network, which both can be independent. The LTE instead of WLAN. Fixed SNR algorithm is
service in this architecture can be provided by using an considered better than Wi-Fi if coverage in terms of
external IP network and roaming between the two balancing the load between a Wi-Fi AP tier and an
networks. In tight coupling architecture, there has to LTE tier.
be a cooperation between the two networks through a In [11] the authors studied another traffic steering
common core and there will be a central administration policy for network selection, which is called Best-
for the majority of network functions, such as billing, Server algorithm. This algorithm behaves similar to
resource management and vertical handover. This is Wi-Fi if coverage and WF algorithms, as the user
provided through I-WLAN standard from 3GPP, which connects first to Wi-Fi. This happens if SINR
allows transferring data through Wi-Fi between users’ perceived by the user is greater than a certain
devices and cellular systems [2, 4, 6]. threshold; otherwise he will connect to an LTE
2.2 Effective Wi-Fi Offloading Algorithms network. The difference between Best-Server
algorithm and the previous algorithms mentioned is
that it depends on SINR while in WF and Wi-Fi if
A Wi-Fi network would be the natural solution for
coverage they depend on SNR by assuming using
data offloading due to several reasons. One reason is
non-overlapping channels. All three algorithms do
that Wi-Fi capabilities are already installed in most
not guarantee a quality of service to connected users
new devices like laptops, iPads, tablets and mobile
to Wi-Fi; they only guarantee the connection.
phones. Users on the other hand will prefer to move on
Another offloading scheme presented in [10] is
Wi-Fi unlicensed free band rather than using an
Physical Data Rate Based algorithm (PDR), which is
expensive licensed cellular band, which is much
purely based on the PDR provided by different
cheaper for them. Wi-Fi also enhances the QoS as it
available RATs. This algorithm compares the PDR
can support high data rate, so streaming videos,
for Wi-Fi and LTE, and chooses the highest value a
downloading music or transferring big file data would
user can get. A similar algorithm presented in [11],
be much better using a Wi-Fi network. From an
called the SMART algorithm, triggers the offloading
operator's point of view, using Wi-Fi offloading can be
based on the minimum data rate perceived by the user
beneficial from different angles, such as deploying
as long his experienced SINR exceeds the threshold.
their own Wi-Fi access points in dense areas to support
cellular coverage and capacity. Wi-Fi uses a wider A performance investigation presented in [12] to
band compared to cellular and because they are down link in a realistic indoor Wi-Fi and femto-cells,
as a solution to offload LTE macro networks with
the 50x growth assumption for data traffic. The Wi-Fi 3. System Model
deployment can meet the target network (KPI) Key
Performance Indicator 90% ,with a minimum data rate 3.1 Target Offloading Scenario
of 1 Mbps at 230AP/Km2 access point density. More
AP density of 1200 AP/Km2 is required to meet the
The integrated data offloading approach
KPI in the out band LTE femto-cell deployment
specified in [2, 4, 6] empowers operators to have full
scenario. On the other hand, the least favorable
control over their subscribers; they connect to either
scenario is in-band femto-cell, where the femto-cell
an LTE or a Wi-Fi network to send their subscribed
shares one carrier with the macro layer. The authors
content. The coupling architecture for Wi-Fi with
have shown that for the same access point density out-
cellular systems either is loose coupling or tight
band femto-cells and Wi-Fi APs can offload a similar
coupling. In this research, we assume having a loose
number of users, while in-band femto-cells can be
coupling architecture, as it doesn’t need a major
improved by interference mitigation.
cooperation between Wi-Fi and LTE networks. The
authors assume a controlled scenario by the 3GPP
In [13], the authors have developed a generic
network operator as the network operator in charge of
framework called (MFW), which stands for mobile
both LTE and WLAN networks. This helps the
femto cells utilizing Wi-Fi, that can exploit both Wi-
network operator to have control over Wi-Fi traffic
Fi and femto-cell networks at the same time. The
with a guarantee of better experience for customers
proposed framework allows operators in a public
across the available networks.
transportation system to offload the traffic generated
The system is composed of a single cell LTE
by the users onto femto base stations (mob FBS), and
eNodeB (BS) operating on 20MHz band and 4 Wi-Fi
the mob FBS utilizes the Wi-Fi as a backhaul to route
AP located inside the LTE cell boundaries operating
the traffic to the cellular network instead of the loaded
on IEEE 802.11n standard with a 20MHz band. The
macro-cell. A numerical experiment has been
location of the 4 open access APs is fixed, while
conducted on MFW with practical applications such
users are randomly distributed inside the macro LTE
as YouTube and Skype, and the efficiency was
and Wi-Fi cells. This architecture is mainly proposed
demonstrated in terms of data traffic offloading. As a
to be tested in congested areas with the objective of
conclusion, when the mobFBS is saturated, MFW can
improving the capacity for the whole system in terms
offload up to 50% of macro-BS data traffic. Table 2 of total throughput, number of connected, and
shows the most effective Wi-Fi offloading algorithms. blocking probability. An offloading approach is used
to overcome the congestion in the cellular network,
No Offloading Trigger Wi-Fi Wi-Fi Wi-Fi AP
Algorithm Criteria Coverage Deployment
by offloading the users to Wi-Fi Aps to relieve the
Model LTE network. Users can connect to either LTE or
Wi-Fi based on specific parameters in the offloading
1 Wi-Fi if Best AP Outdoor IEEE Planned algorithm tested to meet their traffic demand, as both
coverage SNR 802.11 Deployment are controlled by the same operator.
a,b
2 Traffic SINR + Indoor IEEE Traffic
Steering Min DR 802.11 g Centric 3.2 Environment Effect and Link Budget
Calculations
3 Wi-Fi Connect Outdoor IEEE Planned
First to 802.11 g Deployment
Several factors can affect the radio channel in an
Wi-Fi
First LTE or Wi-Fi such as environmental effects, which
4 PHY Data Higher Outdoor IEEE Planned consider the path loss model used in the system,
Rate PDR 802.11 g Deployment shadowing and link budget.
Based
5 Fixed Best AP Outdoor IEEE Planned
SINR SNR 802.11 Deployment
(1) Path Loss:
threshold + Load a,b Three main factors affect the radio channel:
6 MFW RSSI Outdoor IEEE Planned
path loss, shadowing and fast fading [14]. The
802.11 Deployment authors in this research will consider two factors,
r,e path loss and shadowing, which is known as fast
7 Best Best Indoor IEEE Traffic fading. Slow fading is neglected for simplicity
Server SNR 802.11ac Centric reasons. Path loss mainly depends on the distance
LTE-Wi- between a sender and a receiver. It measures the
Fi
power reduction of a transmitted signal when it
Table 2: Effective Capacity Wi-Fi Offloading Algorithms
propagates through space from sender to receiver.
The log-distance path loss model is used to calculate
path loss when a user is connected to either eNodeB
or Wi-Fi AP as shown in equation 1.
(1)
 Where r is the distance between two nodes
from user to either eNodeB or Wi-Fi AP, which is
calculated by the Euclidian distance. The path loss
exponent is e, and its value depends on the type of
environment as in [15]. Fast fading or shadowing is
the power reduction caused by obstacles or huge
objects in the propagation path such as buildings.
Adding the shadowing effect to log-distance path loss
model, path loss can be calculated as in equation 2:
(2)
For eNodeB shadowing is assumed to take a
value of 3 dB, while in Wi-Fi the shadowing is
neglected with the assumption of indoor environment.
(2) Interference:
Interference is assumed to be neglected as the
simulation proposed area is only a single LTE cell with
no boundary effect from surrounding LTE cells. On the
other hand, the authors assumed allocating each user
with one Resource Block (RB) in the frequency
domain, and thus there is no interference between
allocated users as they are using different RBs in one
sub-frame. For the Wi-Fi APs interference between
APs is neglected as well, due to conducting a non-
overlapping scenario, by using channels 1,6,11.
Interference can be calculated according to
equation 3 as follows:
(3)
Interference mainly is the summation of all
transmitted powers and path losses for all users
using the same RB of useri in LTE tier, or all APs using
the same channel in WLAN tier as in DL case.
According to the above assumption the interference for
both LTE and Wi-Fi is equal to zero, I=0.
(3) Link Budget
The link budget is used to define and calculate
the Signal to Interference and Noise Ratio (SINR). For
downlink traffic, SINR is used to estimate the channel
quality or link budget for each user, either from
eNodeB or Wi-Fi AP. SINR is calculated as follows in
equation 4:
(4)
where is the eNodeB transmission power or
WLAN AP, while L is the total path loss between the
user and eNodeB or WLAN AP, I is the total value of
interference, and N represents the thermal noise.
Knowing that I=0 in eNodeB and WLAN AP, as no
user is allowed to use the same RB allocated to another
user, and WLAN APs are using the non-overlapping
channels, SINR can be calculated as follows in
equation 5.
By substitution:
(5)
(4) Noise
Noise power in a receiver usually is a thermal
noise generated from the amplifier, and it is
unwanted energy from natural and man-made
sources. It is calculated as follows:
(6)
From equation 6, N is the noise in linear scale;
K is Boltzmann's constant, which is equal to 1.381 ×
10-23 , T is the reference receiver temperature in
Kelvin, equals to 290; and B is the noise bandwidth
in hertz. According to equation (6) the noise value in
LTE differs from Wi-Fi depending on the bandwidth
allocated to the user in each. Authers assume to
perform a Frequency-Division Duplexing (FDD)
simulation base, thus LTE allocates each user 1 RB
for a 1 ms sub-frame in the time domain. In Wi-Fi
the connected users allocates the whole bandwidth
for a specific time as it uses the CSMA/CA approach
for all users contending to use the same channel.
(5)Throughput and SNR Requirements
Throughput can be calculated originally by
using the Shannon Capacity equation as in equation
(7).
(7)
This is used to calculate throughput as in equation 8:
, ) (8)
Where,
 B is the bandwidth.
 is determined by air interface
capabilities.
Hence throughput can be varied due to several
effects: bandwidth, cyclic prefix, modulation scheme
, antenna configuration, and Mac overhead for both
LTE and Wi-Fi.
The maximum throughput of LTE under a
given assumptions is 300 Mbps in the downlink.
Assuming to have LTE bandwidth of 20MHz with
10% reserved guard band, thus the effective
bandwidth is 18MHz. knowing that RB bandwidth
is equal to 180 Kbps, the total number of available
RBs, can be estimated as in equation 9:
RBs=LTE_TotalBW/RB_BW (9)
Based on equation 9 the total RBs in a sub-
frame with 1 ms duration is equal to 100 RBs , then
the maximum throughput for each user will be 3 Mbps
in the downlink as shown in equation 10.
LTE_Tmaxuser=LTEDL_Tmaxframe/RBs (10)
For the Wi-Fi case, with conducting IEEE
802.11n with a bandwidth of 20 MHz short guard
interval, 64 QAM modulation scheme, 4 × 4 MIMO,
and 85% MAC overhead, then the maximum
throughput is calculated as follows:
2.077 (factor of 20 MHz) × 65 Mbps × 4 antennas ×
1.11 (factor of short guard interval) = 600 Mbps.
By reducing MAC overhead, the total Wi-Fi
IEEE 802.11n stander approximately supports
throughput equal to 100 Mbps, as mentioned in [5, 16].
SNR Requirement Calculation:
Assuming web browsing service with a required
throughput of at least 500 Kbps, and using the Shannon
Capacity equation (7). In order to calculate the required
SNR for a user connected either to LTE or Wi-Fi using
the equation (11) as follows:
 Use Shannon's equation 7:
◦
◦ 4.2 Fixed SNR Threshold Algorithm
◦
◦ (11)
Now by substituting all these values in
equation (11), the required SNR for a single connection
will be:
(12)
5.858
(13)
0.0174
4. Network Access Selection & Fixed SNR Threshold Algorithm:
Offloading Schemes
From the literature, there are several schemes to
perform the offloading from a cellular network to a Wi-
Fi network to relieve a congested 3GPP cellular
system. The authors in this research choose two well-
known offloading algorithms to compare with the
proposed Optimized Offloading algorithm. The first
algorithm is Wi-Fi if coverage and the second one is
Fixed SNR threshold. In this section we present a brief
on each algorithm simulated in this performance study.
4.1 Wi-Fi If Coverage Algorithm
In this algorithm the user will connect to the
best Wi-Fi AP based on the best SNR value he can
get from all APs. The best SNR should be greater
than , and connect to a Wi-Fi whenever the
user finds a WLAN coverage instead of connecting to
a cellular Network.
This algorithm considers only SNR
information to make the network selection decision;
hence it’s a traffic load independent access as in (14).
(14)
Wi-Fi if Coverage Algorithm:
1. Useri requires a connection.
2. Set SNRmin =0.
3. First check the availability for Wi-Fi.
4. Choose the best Wi-Fi AP based on SNRwifi.
5. If SNRwifi >= SNRmin, connect Useri to best
Wi-Fi AP.
6. Otherwise, connect Useri to LTE.
In Fixed SNR threshold algorithm, WLAN gains
the first priority in connection, as in Wi-Fi if
coverage algorithm. A user connects to the best Wi-
Fi AP based on the measured SNRwifi, if it is greater
than or equal to the threshold SNRmin. Otherwise,
the user is directed to a cellular network (i.e. LTE).
The main parameter needed for network selection is
the SNR information from WLAN. The threshold
SNRmin is traffic load-dependent, which means
when the load increases at each individual AP, the
SNRmin is increased as well, as shown in the
following expression in (15).
(15)
1. Useri requires a connection.
2. Set SNRmin =0 & Max_Load forAP.
3. First check the availability for Wi-Fi.
4. Choose the best Wi-Fi AP based on SNRwifi.
5. If SNRwifi >= SNRmin &Load<= Max_Load,
connect Useri to best Wi-Fi AP.
6. Increase SNRmin & Load.
7. Otherwise, connect Useri to LTE.
4.3 Optimized Wi-Fi Offloading Algorithm 5. Simulation Setup & Results
This optimized algorithm is proposed by the In this section, a performance evaluation is
authors, and it’s a cost-function-based algorithm along performed for the two algorithms presented
with QoS. The main three factors taken into account in previously compared to the proposed Optimized
the cost function optimization are: 1) network load, 2) Offloading algorithm. The basic simulation setup
path loss, and 3) throughput. The cost function's main conducted in this study is presented along with
optimizing objective is to maximize the throughput and further assumptions taken into account. The
minimize the network load and path loss. Those three comparative results between the three algorithms
factors have to be taken with an equal weight for both schemes are demonstrated as well with the
tiers, either LTE or WLAN. For each user needs a discussion.
connection, two cost functions will be calculated, one
for LTE and the second for Wi-Fi. After calculating the
5.1 Simulation Setup
cost, the minimum value will be chosen; hence the user
is connected with the minimum cost. This is considered
A total of 100 RBs are used at a single sub-
as the first stage in the Optimized Offloading
frame, and the scheduling requirement is fixed to 1
algorithm, by optimizing the three factors at the same
RB per user in LTE tier. For the WLAN tier, there
time. The second phase of this algorithm is to
are 4 APs deployed in the circulation area of LTE BS
guarantee QoS to the connected user either LTE or Wi-
in fixed locations. Each APs is assumed to be
Fi, by achiving the minimum cost along with the
operating with a non-overlapping channels, 1, 6, and
required throughput. The offloading is triggered if
11 respectively. The eNodeB is fixed in the center of
SNRWiFi or SNRLTE is greater than or equal to SNRmin.
the cell with coverage of 100 meters, while the Wi-Fi
The SNRmin is calculated according to the required
AP's radius is around 20m each. A pair of 20 MHz
throughput to achieve a specific service, which is
bands is assumed for both uplink and downlink
assumed to be web browsing with 500Kbps
transmissions, as it is FDD system based. Different
throughput. If the user is not in the range of any
densities of users are tested with each different
WLAN AP then he will be directed to the cellular
simulation scenario. All simulations of the three
network (LTE). The cost function is calculated as
offloading algorithms are done using MATLAB.
shown in equation (16):
Table 3 displays the main simulation parameters.
(16) Parameter LTE Wi-Fi
knowing that : Cell radius 100 m 20 m
Base Frequency 2 GHz 2.4GHz
The proposed Optimized Offloading algorithm Effective 18 MHz 20MHz
aims to minimize two criteria, the load for each Wi-Fi Bandwidth
AP and LTE BS, and the path loss for each. This will Required 500Kbps web 500Kbps web
happen while maximizing the third criterion, Throughput (R) browsing browsing
throughput for users connected either to Wi-Fi or LTE. No. of Resource 100 1,6,11 channels
Blocks
At the same time the connected users in the Optimized
Number of users 100,200,300, 100,200,300,
algorithm will be served with a guarantee of QoS based 400,500 400,500
on the service type, which is assumed to be web Path loss exponent 4 4
browsing in this research. The load in equation 16 can Shadowing 3 dB 0dB
be expressed by the total number of users connected, Position of BS (0,0) X[-50 50 50 -50]
while path loss is calculated based on equation 2, and (x,y) Y[50 50 -50 -50]
throughput is calculated based on equation 8. Noise -121 dBm -130.9dBm
SNRth/connection 7.2 dB 8 dB
Optimized Wi-Fi Offloading Algorithm: Duplexing FDD CSMA/CA
Modulating 64 QAM 64 QAM
1. Useri requires a connection. scheme
2. Set SNRmin & Max_Load for AP. BS, AP Ptx 21 dBm 23dBm
3. If Useri in the range of any Wi-Fi AP otherwise Max-Throughput 300 Mbps 100 Mbps
check LTE. Table 3 : Parameters Used in Simulation
4. Calculate Cost_WiFi & Cost_LTE.
5. Choose the Min_Cost.
6. If Min_Cost= Cost_WiFi. 5.2 Simulation Assumptions
7. If SNRwifi >= SNRmin & Load<= Max_Load,
connect Useri to best Wi-Fi AP. Moreover, below are some assumptions and
considerations:
8. If Min_Cost= Cost_LTE.
9. Connect Useri to LTE.  The locations of user nodes are randomly
distributed in the cell area.
  The BS can gather information about users.
 Users have limited movement due to density.
 Simulation is done in 1 ms in time domain (single
sub-frame) acquiring 1/10 of all active users.
 The total number of available RBs per TTI
Transmission Time Interval (sub-frame) is equal
to 100.
 The Wi-Fi AP placement is done manually in the
system.
 Wi-Fi Access points are assumed to be open
access for all users.
 Wi-Fi IEEE 802.11n uses the MIMO with 4*4. Figure 1: LTE Standered Case with 500 UEs
 Wi-Fi IEEE 802.11n is assumed to use normal
transmission power with 23 dBm, as it’s the legal
power for most of the available channels.

5.3 Single LTE Cell with Multi APs Results &

Discussion

In this study a performance evaluation is conducted

for two offloading algorithms from literature compared
with the proposed one. The performance is evaluated
with three performance criteria: 1) number of
connected users, 2) blocking ratio as shown in equation
17, and 3) network throughput which expresses the Figure 2: Wi-Fi if Coverage with 500 UEs
total effective throughput of the network.

(17)

The simulation is performed for different numbers

of users generated randomly inside the LTE single cell.
The number of users tested ranges from 100 to 500
users with a step value equal to 100. Figure 1 shows
the simulation for the standard LTE case with no
offloading deployed. With the given assumptions in
this research, a maximum number of users reaches 100
even with a total of 500 users in an LTE cell, as there
are only 100 RBs that can be allocated for users in each Figure 3: Fixed SNR Threshold with 500 UEs
sub-frame with 1ms duration. The number of
connected users is changed when the offloading is
conducted through WLAN APs.

Figures 2 through 4 shows the simulation scenario,

the created connections for 500 UEs with each
offloading algorithm: Wi-Fi if Coverage, Fixed SNR
Threshold, and Optimized Wi-Fi Offloading. The
connected users based on any offloading algorithm are
green, either connected to the LTE BS or Wi-Fi APs,
while red ones represent blocked users. It’s clearly
shown that the number of blocked users is decreased
with the offloading, especially with the optimized Wi- Figure 4: Optimized Wi-Fi Offloading with 500 UEs
Fi offloading. Deploying the Wi-Fi APs near to the
edge of an LTE cell is used to connect more users far After running the simulation for different
from the eNodeB. With Fixed SNR the number of numbers of users with 100 iterations, it is observed
connected users on LTE is not enhanced as much as in that offloading has a major effect in increasing the
Wi-Fi if coverage and Optimized Wi-Fi Offloading. number of connected users compared with the
standard LTE case. In Figure 5, the total number of
connected users is successfully enhanced with all
three offloading algorithms compared to the LTE
standard case. Fixed SNR Threshold performance in
terms of the number of connected users is lower than
Wi-Fi if coverage and the proposed Optimized Wi-Fi
Offloading algorithm, as they seem to have a similar
performance.
Blocking probability is enhanced with the deployed
Wi-Fi offloading scenarios compared to LTE standered
case as shown in Figure 6. With 500 users available in
an LTE cell, depending only on LTE resource we may
end with a blocking ratio of 80%. This is successfully
enhanced through Wi-Fi offloading deployed with
LTE, as the blocking ratio can drop to 62% with 500
users available in the cell with the proposed Optimized
Wi-Fi Offloading. Fixed SNR Threshold seems to have
the worst performance in blocking users; however Wi-
Fi if coverage performs better with a 64% blocking
ratio.
In terms of total network throughput, LTE standard
can serve users with the available RBs; hence with the
previous assumptions the total throughput is fixed even
when the number of users is increased as shown in
Figure 7. Optimized Wi-Fi Offloading and Wi-Fi if
Coverage seems to have a similar performance by
increasing the total network throughput as the number
of users is increased. Fixed SNR has the worst
performance in enhancing the network throughput as
the load threshold and SNR threshold for each AP is
increased with every user connected to the AP. This
results in having a fixed number of connected users for
each AP, and the blocking ratio will be increased for
that.
Figure 5: No. of Connected UEs vs. Total No. of UE
Figure 6: Blocking Probability vs. Total No. of UEs
Figure 7: Total Network Throughput vs. Total No. of UEs
6. Conclusion and Future Work
This research adapted two offloading algorithms
from the literature to compare their performance with
the proposed offloading algorithm by the authors
along with the standard LTE case. The three
offloading algorithms tested in this study are: Wi-Fi
if Coverage, Fixed SNR Threshold, and Optimized
Wi-Fi Offloading. The first two algorithms do not
guarantee a QoS to connected users. This is enhanced
with the Optimized Offloading algorithm, as all
connected users are served with the required QoS
based on service type, which is assumed to be a web
browsing with a 500Kbps required throughput.
Results show that under given assumptions
deploying Wi-Fi offloading can successfully enhance
the performance in terms of connected users,
blocking ratio and total network throughput
compared with the LTE standard case with any
offloading algorithm. Optimized Wi-Fi offloading
algorithms outperform Fixed SNR threshold in all
three performance metrics, while it has a close
performance to Wi-Fi if Coverage. Optimized
Offloading is better than Wi-Fi if coverage as it
guarantees QoS to all connected users, and it
enhances the overall system capacity by optimizing
three criteria at the same time. Optimized Wi-Fi
Offloading improves the network capacity by 55.6%
in terms of connected users with QoS. Future work
includes deploying different numbers of Wi-Fi APs
in LTE cells, to find the optimum number of APs that
enhance the capacity and overall network
performance.
7. Acknowledgments
This paper was funded by the Deanship of
Scientific Research (DSR), King Abdulaziz
University, under grant No. (11-15-1432 HiCi). The
authors acknowledge with thanks DSR technical and
financial support. The thanks extend to Mrs.Maryam
Hajjar for the technical support.
References
[1] D. H. Hagos and R. Kapitza, "Study on
performance-centric offload strategies for
LTE networks," in Wireless and Mobile
Networking Conference (WMNC), 2013
6th Joint IFIP, 2013, pp. 1-10.
[2] R. R. P. ME and G. B. HOD, "Mobile Data
Offloading the Growing Need with Its
Solutions and Challenges."
[3] S. P. Thiagarajah, A. Ting, D. Chieng, M.
Y. Alias, and T. S. Wei, "User data rate
enhancement using heterogeneous LTE-
802.11n offloading in urban area," in 2013
IEEE Symposium on Wireless Technology
& Applications (ISWTA), 2013, pp. 11-16.
[4] A. Aijaz, H. Aghvami, and M. Amani, "A
survey on mobile data offloading: technical
and business perspectives," IEEE Wireless
Communications, vol. 20, pp. 104-112,
2013.
[5] T. Ali-Yahiya, Understanding LTE and its
Performance, 2011.
[6] A. Aijaz, O. Holland, P. Pangalos, and H.
Aghvami, "Energy savings for cellular
access network through Wi-Fi offloading,"
in 2012 IEEE International Conference on
Communications (ICC), 2012, pp. 4369-
4373.
[7] S. Singh, H. S. Dhillon, and J. G. Andrews,
"Offloading in Heterogeneous Networks:
Modeling, Analysis, and Design Insights,"
IEEE Transactions on Wireless
Communications, vol. 12, pp. 2484-2497,
2013.
[8] E. Bulut and B. K. Szymanski, "WiFi
access point deployment for efficient
mobile data offloading," ACM
SIGMOBILE Mobile Computing and
Communications Review, vol. 17, pp. 71-
78, 2013.
[9] J. Kou, J. Miao, Y. Xiao, Y. Wang, and D.
A. Saikrishna, "An offloading algorithm
based on channel quality in mobile
integration network," in 2014 12th
International Conference on Signal
Processing (ICSP), 2014, pp. 1735-1738.
[10] A. Ting, D. Chieng, K. H. Kwong, I.
Andonovic, and K. D. Wong, "Dynamic
backhaul sensitive Network Selection
Scheme in LTE-WiFi wireless HetNet," in
2013 IEEE 24th Annual International
Symposium on Personal, Indoor, and
Mobile Radio Communications (PIMRC),
2013, pp. 3061-3065.
[11] L. Hu, L. L. Sanchez, M. Maternia, I. Z.
Kovacs, B. Vejlgaard, P. Mogensen, et
al., "Modeling of Wi-Fi IEEE 802.11ac
Offloading Performance for 1000x
Capacity Expansion of LTE-Advanced,"
in Vehicular Technology Conference
(VTC Fall), 2013 IEEE 78th, 2013, pp. 1-
6.
[12] L. Hu, C. Coletti, N. Huan, I. Z. Kovács,
B. Vejlgaard, R. Irmer, et al., "Realistic
indoor wi-fi and femto deployment study
as the offloading solution to lte macro
networks," in Vehicular Technology
Conference (VTC Fall), 2012 IEEE, 2012,
pp. 1-6.
[13] M. H. Qutqut, F. M. Al-Turjman, and H.
S. Hassanein, "MFW: Mobile femtocells
utilizing WiFi: A data offloading
framework for cellular networks using
mobile femtocells," in Communications
(ICC), 2013 IEEE International
Conference on, 2013, pp. 6427-6431.
[14] M. K. Hajjar, "Using Clustering
Techniques to Improve Capacity of LTE
Networks," Master of Science, Computer
Science, KING ABDUAZIZ
UNIVERSITY, 2016.
[15] I. S. Misra, Wireless Communications
and Networks: 3G and Beyond: McGraw
Hill Education (India) Pvt Ltd, 2013.
[16] T. Vanhatupa, "Wi-Fi Capacity Analysis
for 802.11 ac and 802.11 n: Theory &
Practice," Ekahau Inc, 2013.
Thamary Abdullah Al-ruhaili was born in
Khobar, Saudi Arabia. She received a BS with an
honors degree in computer science from Al-Dammam
University, Saudi Arabia, in 2003. She is currently
pursuing the Master degree in computer science at
King Abdulaziz University, Jeddah, Saudi Arabia.
Since 2004 she has been a computer science
instructor at Jubail Industrial City, Saudi Arabia.
From 2012 to 2014, she was a member of Deanship
Graduate Studies on behalf of FCIT master students
at King Abdulaziz University, Jeddah, Saudi Arabia.
She is a member of the IEEE Young Professional and
IEEE Computational Intelligence Society. Her
research interests include communication, cellular
networks, artificial intelligence, and network
selection in Multi-RAT and offloading schemes.
Ms. Al-Ruhaili's awards include having the
first place in the Seventh Scientific Forum for King
Abdulaziz University Students, Research in
Engineering and Technology Sciences held in
December 2015.