RESOURCE SHARING IN A WIFI-WIMAX INTEGRATED NETWORK
Except where reference is made to the work of others, the work described in this thesis is
my own or was done in collaboration with my advisory committee. This thesis does not
include proprietary or classified information.
Nirmal Andrews
Certificate of Approval:
Shiwen Mao
Assistant Professor
Electrical and Computer Engineering
Prathima Agrawal,
Samuel Ginn Distinguished Professor
Director, WEREC
Electrical and Computer Engineering
Thaddeus A. Roppel
Associate Professor
Electrical and Computer Science Engi-
neering
George T. Flowers
Dean
Graduate School
RESOURCE SHARING IN A WIFI-WIMAX INTEGRATED NETWORK
Nirmal Andrews
A Thesis
Submitted to
the Graduate Faculty of
Auburn University
in Partial Fulfillment of the
Requirements for the
Degree of
Master of Science
Auburn, Alabama
December 18, 2009
RESOURCE SHARING IN A WIFI-WIMAX INTEGRATED NETWORK
Nirmal Andrews
Permission is granted to Auburn University to make copies of this thesis at its
discretion, upon the request of individuals or institutions and at
their expense. The author reserves all publication rights.
Signature of Author
Date of Graduation
iii
Vita
Nirmal Andrews, son of Philip Andrews and Usha Philip, was born on March 6, 1984 in
Dubai, United Arab Emirates. He completed his Bachelor of Engineering at K.G.G College
of Technology, Anna University, in Chennai, India. He entered the Electrical and Computer
Engineering graduate program at Auburn University in August of 2006.
iv
Thesis Abstract
RESOURCE SHARING IN A WIFI-WIMAX INTEGRATED NETWORK
Nirmal Andrews
Master of Science, December 18, 2009
(B.E., Anna University, 2005)
83 Typed Pages
Directed by Prathima Agrawal
Resource allocation is a challenging issue in integrated wireless environments due to
factors like difference in participating network protocols, design of hierarchical network
architecture, user mobility and multiple traffic types. In order to ensure fair access and
efficiency of bandwidth usage among the component networks in an integrated network,
resource allocation algorithms should be properly designed. The resource considered in
this thesis refers to the wireless channel bandwidth. The design of such resource allocation
algorithms becomes complex in wireless networks when compared to wired networks because
of the error prone nature of the wireless medium. However, when a widespread Wi-Fi
network is integrated with a next generation network such as WiMAX, both the subscribers
and the service providers benefit significantly.
This thesis describes two different resource allocation algorithms in an integrated net-
work of Wi-Fi and WiMAX. The first algorithm is an improvement over an existing two
threshold mechanism suggested by Yahiya et. al. for an integrated Wi-Fi and WiMAX net-
work. In the existing model, the channels are classified as reserved and shared. Reserved
channels are to be accessed only by specific users to which they are assigned and shared
channels can be accessed by users of the component networks i.e. Wi-Fi or WiMAX based
on availability. The proposed algorithm assigns priorities to the users and employs hybrid
resource sharing.
v
Completesharing, completepartitioningandhybridsharingapproachesandtheirshort-
comings in WiMAX-Wi-Fi integrated networks are described. To overcome some of the
shortcomings, a second algorithm called Prioritized Resource Sharing algorithm is pro-
posed. In this algorithm, the channels are prioritized for different traffic classes rather than
strict reservation or open access. The prioritized resource sharing algorithm?s objective is
to improve channel utilization. Channel utilization is improved because prioritized sharing
allows the users to access any of the channels if they are not being used, unlike in complete
partitioning where users are restricted to use only allocated channels even when channels
restricted to other users in the integrated network are not being used. Prioritized sharing
method is an improvement over complete sharing as it guarantees a minimum number of
channels due to the prioritization, unlike in complete sharing where no reservation is pro-
vided which could lead to users of one network capturing all the channels instead of fairly
sharing them.
Finally, the proposed algorithms are analyzed using two dimensional continuous time
Markov chains. It is observed via simulations that using the forced termination model, in
unbalanced loads both Wi-Fi and WiMAX component networks get unbiased service and
neither network is starved. The simulation results also indicate that the Prioritized Sharing
model achieves the best system utilization compared to other algorithms.
vi
Acknowledgments
Firstly, I would like to thank my advisor, Dr. Prathima Agrawal, for her continued
support, encouragement and valuable advise. I convey my sincere gratitude to my advisor
for supervising me in every step in my research and for providing flexibility and freedom in
my work.
I also thank Professors Shiwen Mao and Thaddeus A. Roppel for serving on my advisory
committee. My gratitude also goes out to my lab-mates Nida Fatima Bano, Yogesh Reddy
Kondareddy, Pratap Prasad, Santosh Kulkarni, Srivatsan Soundararajan, Harish Kongara,
Santosh Pandey, Sreekanth Boga, Indraneel Gokhale and Gopalakrishnan Iyer for their
useful suggestions and encouragement. They make Lab405, a great place to work with a
friendly atmosphere. I was blessed with great state-of-art simulation tools in my lab.
I thank my parents for their love, support and encouragement throughout my stay in
Auburn. Above all I thank Jesus Christ for His greatest gift which is salvation and for His
grace and mercy over all my life.
vii
Style manual or journal used Journal of Approximation Theory (together with the style
known as ?aums?). Bibliograpy follows van Leunen?s A Handbook for Scholars.
Computer software used The document preparation package TEX (specifically LATEX)
together with the departmental style-file aums.sty.
viii
Table of Contents
List of Figures x
1 Introduction 1
1.1 Existing Models Of Integrated Networks . . . . . . . . . . . . . . . . . . . . 3
1.1.1 The Necessity For Introducing An Integrated Network With WiMAX
and Wi-Fi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 WiMAX and Wi-Fi Basic Operation . . . . . . . . . . . . . . . . . . . . . . 5
1.2.1 WiMAX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.2 Wi-Fi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3 Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2 Co-existence and Integration of Wi-Fi and WiMAX Networks 11
2.1 Deployment models of integrated 802.11 and 802.16 . . . . . . . . . . . . . 12
2.2 Related work in integrated Wi-Fi and WiMAX networks . . . . . . . . . . . 17
2.2.1 Previous works in integrated Wi-Fi and WiMAX resource allocation 21
2.3 Summary of the technical challenges in coexistence and integration of WiFi
and WiMAX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3 A Prioritized sharing and Forced Sharing Model for Sharing Re-
sources in Wi-Fi - WiMAX Integrated Networks 26
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2 Resource Sharing Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.3 Channel Allocation in Wi-Fi-WiMAX Network Based On Previous Models . 27
3.3.1 System model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.3.2 Channel Allocation Problem . . . . . . . . . . . . . . . . . . . . . . 30
3.3.3 Forced Sharing Model . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.3.4 Prioritized Sharing Model . . . . . . . . . . . . . . . . . . . . . . . . 36
4 Study of the Improvement in Channel Utilization and Network Star-
vation Using the New Sharing Models 39
4.1 Markov Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.2 The Forced Termination Model . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.2.1 Simulation And Results Of Forced Termination Sharing Model . . . 46
4.3 Prioritized Sharing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5 Conclusion 68
Bibliography 70
ix
List of Figures
1.1 Representation of Wireless networks represented based on Radio Access Tech-
nologies and number of hops required to communicate [1] . . . . . . . . . . 2
1.2 Difference between operation of 802.11 and 802.16 except for 802.11e . . . . 6
1.3 Frame format of WiMAX . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4 Frame format of WiMAX . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1 A deployment scenario where 802.16 (WiMAX)/ 802.11 (Wi-Fi) is used ?on
the go?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Dense Urban area coverage using 802.11 - 802.16. . . . . . . . . . . . . . . . 15
2.3 802.11 and 802.16 interworking network serving less populated areas. . . . . 15
2.4 City wide model for deployment of integrated 802.11/802.16. Larger cells are
WiMax cells and the cells inside them are WiFi cells. . . . . . . . . . . . . . 16
2.5 The detailed structure of 802.11 and 802.16 interworking network. . . . . . 19
3.1 System Model of WiMax and WiFi integration. . . . . . . . . . . . . . . . . 29
3.2 a) Complete Sharing, b) Complete Partitioning, c) Hybrid Sharing. . . . . . 31
3.3 Threshold based channel distribution between 802.11 and 802.16 users. N=
Total number of channels. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.4 Direct comparison of blocking probability of 802.11 and 802.16 users with
variation in arrival rate of 802.11 users. . . . . . . . . . . . . . . . . . . . . 34
3.5 Forced sharing mechanism where WiMax user has more priority in the shared
channels and is replacing a WiFi user . . . . . . . . . . . . . . . . . . . . . 35
3.6 Prioritized channel allocation Model. . . . . . . . . . . . . . . . . . . . . . . 37
4.1 Forced sharing mechanism where WiMax user has more priority in the shared
channels and is replacing a WiFi user . . . . . . . . . . . . . . . . . . . . . 42
x
4.2 Markov Model representing Forced Termination Sharing . . . . . . . . . . . 43
4.3 Balance equations representing first come first serve model . . . . . . . . . . 45
4.4 Equations to calculate Blocking probability of both WiFi and WiMAX users 46
4.5 802.11 blocking probability variation with increase in number of reserved
channels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.6 Direct comparison of blocking probability of 802.11 and 802.16 users with
variation in arrival rate of 802.11 users. . . . . . . . . . . . . . . . . . . . . 48
4.7 802.16 blocking probability variation with increase in number of channels . 49
4.8 (a)Blocking probability of WiMax with change in WiMax load as well as
change in channels reserved for WiFi users.(b)Blocking probability of WiFi
with change in WiFi load as well as change in channels reserved for WiFi users. 50
4.9 Comparison of WiMAX throughput between first come first serve model and
forced termination model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.10 Comparison of Wi-Fi blocking probability between first come first serve
model and forced termination model . . . . . . . . . . . . . . . . . . . . . . 52
4.11 (a)Variation of WiMax throughput with change in channel reservation.
(b)Variation of WiFi throughput with change in channel reservation. . . . . 53
4.12 802.11 dropping probability variation with increase in number of reserved
channels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.13 Threshold based channel distribution between 802.11 and 802.16 users. N=
Total number of channels. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.14 Prioritized channel allocation Model. . . . . . . . . . . . . . . . . . . . . . . 56
4.15 Prioritized channel allocation Model. . . . . . . . . . . . . . . . . . . . . . . 57
4.16 Markov Model representing Prioritized Sharing . . . . . . . . . . . . . . . . 59
4.17 Blocking probability of WiMax Users with varying WiFi user traffic. . . . . 63
4.18 Blocking probability of WiFi Users with varying WiFi user traffic. . . . . . 64
4.19 Blocking probability of WiFi Users with varying WiMax user traffic. . . . . 64
4.20 Total Percentage Channel Utilization with varying WiMax user traffic. . . . 66
4.21 Total Percentage Channel Utilization with varying WiMax user traffic. . . . 67
xi
Chapter 1
Introduction
Development of Internet and wireless networks has been the two most significant contri-
butions in the information revolution. An ever increasing user demand for ubiquitous data
and voice access at lower cost has instigated the necessity for a next generation network.
The aim of these networks would be to provide multimedia services, such as voice, video
and continuous data streams at high data-rates to mobile users in large coverage areas at
all times. The architecture for the next generation of wireless networks aims to integrate
multiple networks and benefit from the resulting synergy. There are various standard-
ization bodies working towards this vision. Examples include the 3GPP (3rd Generation
Partnership Project), 3GPP2, IEEE 802.21 Media Independent Handover Working Group
and Network Working Group (NWG). However, to achieve a seamless integration there
are several technical challenges that have to be addressed such as mobility management,
resource allocation, admission control, protocol adaptation, security, and pricing. On the
brighter side, the advances in integrated circuit design and software radio make it possible
to implement multiple network interfaces in a single mobile terminal. Such terminals can
access different types of wireless and mobile networks, which provide more versatile and
flexible access options, making the integration feasible.
It is envisioned that beyond 3G networks, systems will integrate to provide overlapping
coverage for mobile users, due to the complementary characteristics of various networks
especially for wider coverage and higher data rates. Integration involves various challenges
as mentioned earlier such as choosing the most efficient architecture for integration based
on the standards that are being integrated and how to share resources (such as bandwidth,
power,etc.) among the integrated networks. For example, the architecture of the integrated
networks could be centralized, decentralized or it could be a combination of both depending
1
Figure 1.1: Representation of Wireless networks represented based on Radio Access Tech-
nologies and number of hops required to communicate [1]
on the different features that are expected from the integration. If it is centralized, a
base station handles all computation or it could also be distributed within different base
stations. Each of the architectures has its own advantages and disadvantages, which should
be weighed and analyzed to make a choice on the best architecture for integration. The
decision takes into consideration various features that the subscribers and service providers
are expecting from the integrated network as well as the compatibility of the standards.
Figure 1.1 from [1], shows a possible design space for integrated wireless networks and
establishes a difference between integrated networks and other architectures such as WLAN
hotspots, relay based networks, multi-hop networks, mesh networks etc. The difference
is based on two factors which are represented as the two axis in Figure 1.1, the number
of hops required to connect to the Internet and the number of radio access technologies
(standards) involved. Integrated networks have multiple standards working together and
use single hop communication. Wide area networks like 3G can be integrated with a local
area network such as WLAN. Users can seamlessly roam between these networks using
2
single hop communication to their respective base stations. On the contrary mesh networks
use multiple hops to access a wide area network.
1.1 Existing Models Of Integrated Networks
Integration of multiple wireless access technologies is not a new concept. Multi-homing
is an example of integration of various access technologies in wired networks which allowed
users to access Internet through dial up, Ethernet, cable, DSL (Digital Subscriber Link),
etc. However, within wireless networks the early adopter of the integration concept was
wireless voice networks. The dual mode DECT/GSM (Digital Enhanced Cordless Telecom-
munication/ Global System for Mobile communications)phones indicate the same. This
dual mode phone connects to the home telephone networks when indoors and the GSM
network for outdoors when the user is not in the range of the DECT base station. The
development of new devices such as the dual mode phone and routers that bridge different
technologies have brought us closer to realizing integration as a reality.
Although the advantages of integration of networks are numerous, designing the ar-
chitecture and efficiently integrating the various wireless standards to obtain the numerous
advantages is a challenging task. For example, on analyzing the interworking between
WLAN and 3G, it could be observed that WLAN was not targeted for public environments
as a result of which it doesnt have universal AAA (authorization, authentication and ac-
counting) mechanism, support of roaming agreements, strong mobility management with
QoS etc. Similarly, even though 3G has most of these mechanisms inbuilt, vertical handoffs
always pose a challenge in mobility management as 3G is built for horizontal handoff.
1.1.1 The Necessity For Introducing An Integrated Network With WiMAX
and Wi-Fi
Most of the existing work in integrating wireless networks for next generation commu-
nication standards is focused on combing cellular standards along with WLAN. Cellular
networks have better mobility management and larger coverage which WLAN do not have
3
whereas, WLANs have better data rates which are much higher and provide faster data
access compared to cellular standards. If they are integrated, the complementary nature of
both the cellular and WLAN networks would benefit both the networks by covering up the
drawbacks of the individual standards. However, recently the 4G working group defined
the following as objectives of the next generation wireless communication standard in [2],
[3] and [4]. They are:
1. A spectrally efficient system (in bits/s/Hz and bits/s/Hz/site),
2. High network capacity: more simultaneous users per cell,[2]
3. A nominal data rate of 100 Mbit/s while the client physically moves at high speeds
relative to the station, and 1 Gbit/s while client and station are in relatively fixed
positions as defined by the ITU-R,[3]
4. A data rate of at least 100 Mbit/s between any two points in the world,[3]
5. Seamless connectivity and global roaming across multiple networks [4],
6. High quality of service for next generation multimedia support (real time audio, high
speed data, HDTV video content, mobile TV, etc)
7. Interoperability with existing wireless standards, and
8. An all IP, packet switched network.
In summary, the next generation systems should dynamically share and utilize net-
work resources to meet the minimal requirements of all the 4G enabled users. With this
objective set for the next generation networks, it becomes obvious that a combination in-
volving cellular standards would not suffice. Although 2.5 and 3G cellular data services
offer wide area Internet connectivity, these services do not provide the broadband speeds
of Wi-Fi. WiMAX on the other hand, can provide high speeds in service along with the
quality of service. By combining WiMAX and Wi-Fi technologies, service providers can
4
offer their subscribers a more complete suite of broadband services over a wider variety of
user environments. The service providers can leverage both types of spectrum; for example,
license exempt for best effort local area traffic and licensed for wide area and QoS sensitive
traffic. Similarly, the subscribers get a full range of services at home and office, as well as
on the road and economical coverage over large areas. These advantages make integration
of WiMAX and WiFi desirable.
1.2 WiMAX and Wi-Fi Basic Operation
The study of integration and coexistence of 802.11 (used interchangeably as Wi-Fi
hereafter. Wi-Fi is based on IEEE 802.11 standard) and 802.16 (used interchangeably as
WiMAX hereafter. WiMAX is based on IEEE 802.16 standard) networks is impossible
without the knowledge of the differences in operation and access schemes of both of these
technologies. Table in Figure 1.2 gives a brief overview of each technology and the differences
between them. Based on this table it could easily seen that interworking between both
networks needs a lot of coordination at various levels of the network namely physical, MAC
, routing and transportation layers.
The table in Figure 1.2, specifically does not mention 802.11e (Wi-Fi with Quality of
Service(QoS)) since it is different from the rest of the 802.11 systems and its features and
structure make it highly adaptable to co-existence and interworking. The above mentioned
802.11 systems do not focus on QoS architecture and so they usually have only one type
of service flow which is Best Effort. Considering the fact that QoS plays an important
role while integrating 802.11 and 802.16, it should be understood that 802.16 have different
service flows for guaranteeing Quality of Service. The details of the service flows will be
discussed later. The mapping of quality of service classes between 802.11 systems and 802.16
systems is quite challenging. Therefore, 802.11e which was developed for providing better
Quality of Service standards for the 802.11 systems have categorized service flow which
makes it more adaptable to co-existence and integration than its peers. The table shows
5
Figure 1.2: Difference between operation of 802.11 and 802.16 except for 802.11e
the difference between 802.11 and 802.16 networks at different levels. The next subsections
will explain the working of WiMAX and WiFi standards.
1.2.1 WiMAX
IEEE 802.16 [5] is a radio standard for WMANs (Wireless Metropolitan Area Networks)
operating in the frequencies between 2 and 11 GHz often referred to as WiMAX. It specifies
four different physical layers (PHYs), as shown in Figure 1.2. IEEE 802.16 has a centralized
architecture provided by a central Base Station (BS) with associated Subscriber Stations
(SS). Typically, a BS is connected either directly or via additional BSs to the core network.
Therefore, 802.16 offers an optional mesh deployment that introduces multi-hop connections
via relaying BSs.
In 802.16, the BS estimates the operational power level for each subscriber and broad-
casts a ranging code. The initial ranging process is described in [6]. The 802.16 subscribers
are allowed to contend in the ranging channel for the initial ranging process and only those
6
successful subscribers can establish communication with the BS [6]. In [7], the design of
a WiMAX based last mile network solution provides a practical example of the various
network deployment structures for WiMAX, which includes the working of 802.16 BS and
SS too.
Figure 1.3: Frame format of WiMAX
A detailed analysis over the 802.16 MAC is provided in [8].With a centrally controlled,
frame based MAC approach 802.16 offers guaranteed multimedia quality of service. 802.16
supports non line-of-sight operation and large coverage areas, which enables a rapidly de-
ployable infrastructure. The frame structure of the OFDM (Orthogonal Frequency Division
Multiplexing) PHY layer operating in Time Division Duplex (TDD) mode is illustrated in
Figure 1.3. Each frame consists of a Downlink (DL) subframe always followed by an Uplink
(UL) subframe. The DL subframe starts with a long preamble used for synchronization
followed by the Frame Control Header (FCH). The DL subframe consists of one or multiple
DL bursts containing MAC Packet Data Units (PDUs) scheduled for DL transmission. The
UL subframe starts with contention intervals scheduled for initial ranging and bandwidth
request purposes. This is followed by one or multiple UL-bursts follow, each transmitted
from a different SS. An UL-burst is initiated with a short preamble and contains one or
7
several MAC PDUs. DL and UL subframe are separated by the Receive/transmit Transi-
tion Gap (RTG) and the Transmit/receive Transition Gap (TTG). An (optional) extension
of the DL-MAP with the duration of a burst enables the BS to flexibly arrange concurrent
DL bursts. The knowledge of start time and the duration overcomes the restriction of the
sequential nature of bursts. This benefits the Space Division Multiple Access operation
of IEEE 802.16. The next section gives a similar description of the operation of 802.11
networks.
1.2.2 Wi-Fi
This section provides a brief description of the working of Wi-Fi MAC layer. WiFi
technology is based on IEEE 802.11 standard. The IEEE 802.11 standard includes a medium
access control (MAC) layer that uses carrier sense multiple access with collision avoidance
(CSMA/CA) and an optional handshaking mechanism termed RTS-CTS. The use of these
two mechanisms in the distributed coordination function (DCF) mode allows fair sharing of
the medium between users in close proximity. The CSMA/CA protocol requires that when
the medium is sensed to contain power over a certain threshold, a prospective transmitter
must defer to the next available transmission opportunity. If the sensed transmission is
decodable (i.e. a transmission from a node using a compatible standard) the deferral uses
header information to discover the start of the next available slot. To allow contention
between a number of nodes competing for the same transmission slot, a random backoff is
selected by each node according to rules defined in the standard. Only the node with the
shortest backoff may transmit.
Use of RTS-CTS allows a data transmission to be protected (via CSMA/CA) at both
transmitter and receiver ends. The main drawback of the IEEE 802.11 standard is its
poor efficiency. There are significant overheads to each data exchange - transmission slot
boundaries (termed DIFS periods in DCF mode), backoff intervals, RTS-CTS delays if
used, and acknowledgment (ACK) overhead if used (the IEEE 802.11e standard allows for
different quality of service (QoS) types which may or may not require ACKs). A typical
8
Figure 1.4: Frame format of WiMAX
exchange is shown in Figure 1.4. The actual data transmission occupies less than half of the
exchange length. It should also be mentioned that the CSMA/CA mechanism is inefficient
since often a node will defer unnecessarily, i.e. when it could have transmitted and not
disrupted the transmission that it sensed.
There is also a point coordination function (PCF) mode specified in the standard that
allows one node to control the scheduling of packets. This has advantages in improved
efficiency (through reduced overheads, backoff time and collisions) and also allowing the
controlling node (which typically is an access point to the wired infrastructure and through
which most traffic is routed) to o?oad its queued data with a higher priority. However, the
PCF is less compatible with co-existence scenarios.
Integrating 802.11 standard with 802.16 standards would be easier if both the chosen
standards had an inbuilt QoS feature rather than implement a new one. In case of 802.11e
(802.11 with QoS) and 802.16 the difficulty lies in matching the already existing service
flow or QoS class of both the networks to form a common service class. The related work
section in the next chapter will describe various methods used to make a common service
class for smooth functioning of the integrated networks.
9
The next subsection explains how a new 802.11 standard was designed to include QoS
features which were not included in legacy 802.11 or 802.11a/b and g standard. The 802.11
standard with QoS is named IEEE 802.11e. The description of this standard follows in the
next subsection.
1.3 Organization
The rest of this thesis is organized into four chapters. In Chapter 2, a case for the
integration of Wi-Fi and WiMAX networks is built by studying the necessity, the advantages
and feasibility of such an integration. It also summarizes all the existing work in the area of
Wi-Fi and WiMAX integration methods and categorizes the technical challenges involved
in the integration of the two networks. Chapter 3 deals with the technical challenges in
the integration of WiFi and WiMAX and narrow the focus to resource allocation problem
in the network. Resource allocation methods are summarized along with the drawbacks
of existing models for integrating WiFi and WiMAX networks. Then two new algorithms
to improve over the drawbacks of the existing models, in terms of network utilization and
call blocking probability, are proposed. The proposed algorithms for resource sharing are
analyzed using MATLAB in Chapter 4. The models are based on 2D Markov chains. The
results of the simulations are discussed and the improvement over existing models is shown.
Chapter 5 summarizes the results and mentions some future work.
10
Chapter 2
Co-existence and Integration of Wi-Fi and WiMAX Networks
WiMAX which is WMAN (Wireless Metropolitan Area Network) standard comple-
ments Wi-Fi which is a WLAN (Wireless Local Area Network) standard, by providing
support for high speed applications requiring data rates of up to 100 Mbps (802.11 a and
d) over a wider coverage area with a radius of up to 1 mile as compared to lower coverage
radius of 802.11 standards. The complementary nature of both the networks makes the
integration of Wi-Fi and WiMAX favorable to offer several advantages such as ubiquitous
high data rate provision, high quality of service (QoS) from the inbuilt QoS provision of
WiMAX, etc. The service operators get larger network coverage with a small investment
and the users get ubiquitous network access with guaranteed service [9]. According to [10],
some of the advantages of integrating Wi-Fi and WiMAX are listed below as:
1. High connectivity, where users can connect to either of the networks based on their
location, Quality of Service requirements and coverage.
2. Efficient global roaming.
3. Provision of broader range of services both at home as well as on the road.
4. Usage of both licensed and license exempt spectrum based on service requirements.
5. Economical coverage of large areas.
The above mentioned reasons make the integration of Wi-Fi and WiMAX highly de-
sirable. Concisely, the integration results in wide area broadband communication with high
QoS. Gakhar et al., in [11] states that efficient interworking of Wi-Fi and WiMAX is required
for providing quality of service for delay sensitive and bandwidth intensive applications such
11
as VoIP, video transmission, large volume FTP etc. According to Thomas et al. in [12], the
possibility of integration between Wi-Fi and WiMAX becomes more realistic in the future
generations because Wi-Fi is widely established and WiMAX is still evolving providing all
the data rate and QoS features that are not available in Wi-Fi.
The standardization of both Wi-Fi and WiMAX technology by Wi-Fi Alliance and
WiMAX Forum makes the interoperability within each network and between the two net-
works feasible. Similarities in both technologies such as the use of IP (Internet protocol)
- based technologies to connect to the Internet also favors integration. There are a few
requirements for integration Wi-Fi and WiMAX that were suggested by Motorola and Intel
in [10]. The most important ones are multi-mode subscriber devices that can communicate
with both Wi-Fi and WiMAX networks and maintaining session continuity. Network Work-
ing Group (NWG) [10] gives the specifications for allowing transition of users from 802.16
networks to other networks. An ASN (Access Service Network) gateway has been designed
by NWG for managing access to services such as AAA (Authentication, Authorization and
Accounting), DHCP, session and mobility management [10]. This allows all calls or data
transfers to be smoothly handled when Wi-Fi network users or WiMAX network users en-
ter WiMAX or Wi-Fi networks, respectively. These developments will facilitate in making
integration of 802.11 and 802.16 feasible.
The next section explains how the WiMAX (802.16) and Wi-Fi (802.11) integrated
network will be deployed and what economic benefits will such a deployment have.
2.1 Deployment models of integrated 802.11 and 802.16
Motorola and Intel in [10] recommended four different deployment models for the in-
tegrated 802.11 and 802.16 architecture. The different deployment models were:
1. Broadband ?on the go?
Deployment of 802.11 and 802.16 in the user devices, allows service providers to
offer transparent service between 802.11 in hotspots and 802.16 beyond the hotspots.
12
Deployment of 802.16 in areas with high density of Internet users extends broadband
services beyond hotspots providing both users and service providers the utility and
value of service.
Figure 2.1: A deployment scenario where 802.16 (WiMAX)/ 802.11 (Wi-Fi) is used ?on the
go?.
2. Last mile broadband
In rural areas where the population density is scarce the expansion of DSL will be very
expensive and in urban areas where it could be difficult to add wired connections to
existing multiple dwelling units (MDU) the integrated network can be used to extend
broadband connectivity in the last mile. 802.16 will be used as backhaul which will
make the extension less costly.
3. Broadband Campus coverage
802.16 can cover larger area. It can provide connectivity beyond the individual build-
ings to an entire campus as blanket coverage. This allows service providers to offer
choice of connectivity to users enabling them to connect to the 802.11 (Wi-Fi) network
in the building or the 802.16 network which covers the campus. This usage of 802.16
13
can reduce the number of 802.11 access points needed to cover the entire campus,
thereby reducing maintenance cost.
Gunasekaran V. et al., in [13] suggested a few deployment models that can be used
by service providers based on the density of population in the area of deployment. They
classified the areas as: dense urban area and low density area. The authors build the
integrated network model on top of the underlying 802.11 network which is already deployed
widely. They identified the widely deployed 802.11 models as: 1) Location specific models
or hotspots covering specific locations where Internet usage could be high such as airports,
hotels, cafes?, etc. and 2) Extensive or outdoor model which could be an example of areas
with low population density where 802.11 is used to provide outdoor coverage. The different
deployment scenarios and the integrated Wi-Fi and WiMAX models that suit the scenarios
are mentioned briefly as:
1. In the dense urban areas majority of the SOHO (small offices, home offices) and
households are in multi dwelling units or apartment complexes as shown in Figure
2.2. The authors suggest that 802.16 be used as backhaul in these dense urban area
deployment scenarios instead of the existing wired backhaul for individual 802.11
hotspots. 802.16 backhaul can deliver megabits of data to the multi dwelling units
from where 802.11 can be used to distribute services to individual users or rooms.
2. The second case is where the population density is low. This scenario differs in
the way that it requires a larger number of 802.11 access points to cover the area.
Gunasekaran et al. in [13] suggests a single private owner model for operating the
services and selling it directly to the customers. The other possibility is that the
local government can make deals with multiple owners to grant license to operate,
providing the service providers the city?s existing infrastructure like street poles and
lamp posts to deploy the network as shown in Figure 2.3. This helps in reducing
the capital cost and operating cost, which enables the operators to offer lower cost of
service to customers.
14
Figure 2.2: Dense Urban area coverage using 802.11 - 802.16.
Figure 2.3: 802.11 and 802.16 interworking network serving less populated areas.
15
3. Gunasekaran et al., provide yet another infrastructure for the integrated 802.11/802.16
network shown in Figure 2.4. They use 802.16 mesh as a backhaul. In a thirty six
square mile radius assuming that nine 802.16 cells of diameter coverage two miles
can be fitted, it would require nine wired connections for each of the BS in the cell,
but in case of a 802.16 mesh one wired connection could substitute the nine wired
connections instead. Once the 802.16 cells are in place, the next step is to pack these
802.16 coverage cells with 802.11 access points (AP) and using 802.11 as the last mile
connection. The authors suggest the usage of 802.11 a/b and g to provide an extensive
option for users as well as operators to provide best service.
Figure 2.4: City wide model for deployment of integrated 802.11/802.16. Larger cells are
WiMax cells and the cells inside them are WiFi cells.
Economic benefits of the integration
Gunasekaran et al., study the most optimal deployment of 802.11/802.16 network based
on financial aspects. Using the model shown in Figure 2.4, a cost analysis is performed
and an approximate cost of service that the service providers can charge the end users is
established. Users are classified as light, heavy and business users. The average bandwidth
16
for light users is 250kbps and for heavy users is about 500 Kbps. The benefit for the end
user is lower cost of service and multiple service provisions. The monthly subscription cost
of $15 for light users and $30 for heavy users. For business users the service is provided
for $60 per month. On demand users will be charged $5 per day of usage. $20 per month
will provide voice over WiFi (VoWiFi) which allows users to make unlimited local and long
distance calls for an entire month. The model designed based on all the above mentioned
assumptions will enable for the operators to recover cost of network deployment within two
years of operation.
2.2 Related work in integrated Wi-Fi and WiMAX networks
Figure 2.5, shared the details of integrating WiFi and WiMAXnetworks and integration
have been classified based on the different layers of the protocol stack. It categorizes the
most relevant functions performed in physical, data link and network layers of the integrated
Wi-Fi and WiMAX network. In this section the previous related work is described in detail
and the most prominent issues in an integrated WiFi and WiMAX network are summarized.
Thomas, Nicholas J. et al in [12] studied the effects of coexistence of 802.11 and 802.16
on the performance of both 802.11 and 801.16 and found that the total throughput of the
coexisting system drops to about 10% of the combined offered load. This is because of
interference between 802.11 and 802.16 networks and the absence of coordination or call
admission protocols to eliminate such interferences.
Since it has been proved that the coexistence of 802.11 and 802.16 without any control
scheme would lead to severe interference and performance degradation, Jeng Xiangpeng
et al in [14] designed three reactive schemes; dynamic frequency selection (DFS), power
control (PC) and time agility (TA) to avoid interference. Using time agility throughput of
the hotspot could be increased by 30% and using power control scheme the throughput at
the 802.16 SS can be increased by 4 times while reducing the throughput of 802.11 hotspots
by a very small amount. Jing, Xiangpeng and Raychaudhuri, Dipankar also designed a
17
proactive scheme in [15] named as common spectrum coordination channel (CSCC) eti-
quette protocol to consider the hidden station problems which were not accounted for in
the reactive schemes. Using CSCC it was observed that in the case of a single BS with
single hotspot the DL throughput of 802.16 is improved by 35% with the change in distance
between the SS and hotspot. In the multiple BS and hotspot scenario different 802.16 SS
geographic distributions were studied. When the SS are distributed uniformly there is an
802.11 throughput increase of 15% and when the SS are clustered around the hotspot the
throughput increases by 160% for any larger clusters the throughput degrades.
Berlemann, Lars et al in [16] studied the coexistence of 802.11(a) and 802.16 within
the license exempt frequency band and proposed a solution for interference avoidance in
such a coexistence scenario. In the proposed solution, full control of the radio resource is
provided to 802.16. The solution targets the avoidance of an idle medium for duration of
DIFS or longer before and during the 802.16 MAC frame transmission and thereby avoiding
interference and protecting the 802.16 MAC frame. This solution proves highly unfair to
802.11a users since they have no control. Unnecessary handoffs in a network also leads to
drastic reduction in throughput and call handling efficiency. Nie, Jing et al., in [17] designed
a vertical handoff technique for integrated 802.16/ 802.11 networks. In the case of 802.11/
802.16, both the networks are high bandwidth networks. In the proposed technique, a user
in the 802.16 network is handed off to 802.16 network if the bandwidth provided by 802.11
is higher than that offered to 802.16. Now if the user is in 802.11 network, then the handoff
decision is made based on average RSS (Received signal strength) and bandwidth. The
bandwidth of 802.11 was calculated using NAV packets.
Coexistence just studies the effect that the coexisting networks have on each other.
However, interworking goes deeper since it has to consider MAC as well as Quality of
Service mapping issues for the combined working of both the interworking networks. For
information exchange within an 802.11 and 802.16 integrated network, Berlemann, Lars et
al in [18] designed a common frame structure capable of operating in both 802.11e (802.11
18
Figure 2.5: The detailed structure of 802.11 and 802.16 interworking network.
19
with QoS) and 802.16 network according to [18]. The authors propose a single common com-
munication device named as Base Station Hybrid Coordinator (BSHC). The BSHC embeds
the 802.11e (802.11 with QoS) transmissions into 802.16 frame structure. Simulation results
showed that with increase in the number of 802.11e end users the maximum throughput
decreases because of more contention in contention period of 802.16 transmission. Later,
Tang-Lin et al suggested the model of an integrated access point named W2- AP which
acts as either a bridge in translating frames between 802.11 and an Ethernet interface or
as a relay node transferring frames between 802.11 and 802.16 in [19]. The Wi-Fi protocol
operation is modified to match WiMAX system. It was observed that the proposed scheme
provided significant improvement in end to end delay due to the connection oriented unified
approach.
Since the MAC protocols of both 802.11 and 802.16 networks vary, the direct mapping
of Quality of Service frameworks of 802.11e and 802.16 are not possible. Gakhar, Kamal
et al proposed a Quality of Service framework named Radio Gateway (RG) in [11] for
802.11e/802.16 integration. A common service flow structure was designed which combined
the service flows of both 802.16 and 802.11e which made mapping possible. However a
drawback of this solution is that initial setup time is required since it involves buffering,
mapping and setting up new connection. So, Fantacci and Tarchi in [20] proposed a bridging
device capable of transparently interconnecting both 802.11a and 802.16 keeping in mind
QoS sensitive applications. Simulation results using the proposed model showed that end
to end delay for QoS sensitive applications reduced significantly and thus QoS sensitive
applications were served in a timely manner. Ali Yahiya et al [21] concentrated on the
quality of service maintenance during mobility or vertical handoff between the WiFi and
WiMAX integrated networks. A bandwidth reservation policy was used in 802.16 which
set aside a dedicated amount of resource for the handoff users in such a manner that the
existing 802.16 SS users do not have to suffer loss of quality of service. It was observed
that within a set limit of maximum tolerable latency, 802.16 could support 24 voice sessions
20
and 21 video sessions transferred from 802.11 without any packet drop using the proposed
model.
For routing in 802.11 and 802.16 integrated networks, Ibanez et al in [22], modified the
existing AODV protocol to accommodate the integration of 802.11 and 802.16. The algo-
rithm sends out a request packet until it can reach a destination node or the 802.11/802.16
gateway. If the destination is outside the coverage of 802.11 network, the AP retransmits
the request packet. Once the 802.11/802.16 gateway or the destination node receives the
packet it sends a reply packet to the source which helps in route discovery.
2.2.1 Previous works in integrated Wi-Fi and WiMAX resource allocation
From the literature on resource allocation in integrated 802.11/802.16, it can be seen
that mathematical modeling of resource allocation problem is achieved using Markov chain
models or game theory. Niyato and Hossain [23] in their study aim at providing good Quality
of Service to end users by designing efficient bandwidth management and admission control
algorithms. The authors use a hierarchical model for the bandwidth allocation. The hier-
archical bandwidth allocation is done in two levels; the first level uses a bargaining game to
allocate resources to the SS (subscriber stations) and the WLAN APs? and then the second
level distribution of the given bandwidth among different WLAN end users. The solutions
obtained focus on providing end user Quality of Service based fair bandwidth allocation
among the connections in the integrated network. Niyato and Hossoin in [24] extended the
idea discussed above to a different scenario of 802.11 in a multihop 802.16 network where
802.16 mesh is used as a backhaul for 802.11 hotspots. The resource allocation in this case
was not just between 802.11 APs and 802.16 SS, but it also includes resource sharing with
traffic from relay connections. The results in this method indicate that the mesh routers are
given fair bandwidth allocation. Niyato and Hossoin [25] also modeled a pricing strategy for
the integrated networks. The model uses the Stackelberg equilibrium where the 802.16 BS
uses the 802.11 users best response information to provide optimal supply of bandwidth so
as to gain highest profit. Numerical results indicate that at equilibrium 802.16 BS charges
21
the same price for every 802.11 router even though the bandwidth demands differ and when
traffic arrival increases the 802.16 BS increases the price on 802.11 APs to compensate for
the loss in revenue by degradation of quality of service to SSs.
In [26] and [9] Yahiya Ali et al., study resource allocation in an integrated 802.11/802.16
network. The authors also propose a resource allocation scheme for the efficient performance
of the interworking network. The policy based on channel bandwidth threshold is used for
managing resources between traffic of both types. Two thresholds are used corresponding
to direct and indirect traffic to 802.16 BS. The direct traffic is the traffic from 802.16 users
and indirect traffic is 802.11 users traffic. The architecture in this interworking scenario is
tight coupling which makes 802.11 traffic dependent on 802.11 BS for reaching the Internet
as shown in Figure 3.1. Therefore in this proposed solution for efficient management of
resources, priority is given to 802.11 users over 802.16 users. There are several methods of
bandwidth allocationwhicharecompletesharing, completepartitioningandamethodwhich
integrates both. The authors use the integrated method where the resource is partitioned
partially and the rest of the available resource is left for sharing. The resource in this
discussion is bandwidth in terms of channel. The users are blocked only when the respective
partitioned channels as well as the shared channels are depleted or assigned. The model is
analyzed using Markov chains. The results indicate that with variation in partitioning and
load the blocking probability also varies.
2.3 Summary of the technical challenges in coexistence and integration of WiFi
and WiMAX
Based on the previous work, the different problems in the area of integrated Wi-Fi and
WiMAX network are classified as:
1. Handoff
In interworking and coexisting networks there are two kinds of handoff namely vertical
and horizontal handoff.When the handoff takes place between similar type of networks
22
it is called horizontal handoff and when the handoff is in a hybrid network it is called
vertical handoff. In an 802.11 and 802.16 integrated network, there are doubly covered
areas service by 802.11 as well as 802.16. In such areas, decision is to be made
as to how the handoff will be made and on what network criteria it will be made.
Unnecessary handoffs can reduce the throughput of the network drastically and also
cause unnecessary call droppings which will reduce the Quality of Service provided to
the users.
2. Resource sharing
Resource sharing becomes an important issue in interworking networks since users of
two different networks are to be served and both of them require Quality of Service
with the limited resources available. Network resources should also be fairly shared
between the two different networks. Researchers working in this area have adopted
various mechanisms for resource sharing such as admission control and channel allo-
cation schemes.
3. Quality of Service
In integrated networks, Quality of service is a major issue since the quality of service
provisions for the component networks are different. In fact, it could be said that
802.11 does not inherently have any Quality of Service provisioning where as 802.16
has a very structured quality plan. This mismatch in quality provisions in both these
networks will require to be bridged for efficient performance using service mappings
or tight coupling where the service provisions of one network are exactly duplicated
and used by the other network.
4. Protocol adaptation
The MAC protocol, frame structure and operation mechanism of 802.16 and 802.11
are both different. 802.16 as we have seen earlier uses TDMA access and 802.11 uses
CSMA/CA. They use dissimilar frame formats. The incompatibility has to be bridged
23
in order for communication to be established between the two networks. Solutions
that exist for this problems, redesign the protocol and suggest embedding the 802.11
frame in the 802.16 frame. Such solutions lead to throughput loss and hence efficient
design becomes a challenge.
5. Routing
Routing protocols for interworking networks have to take into consideration factors
such as vertical handoff and mobility issues. The traditional routing protocols such as
AODV (Ad hoc On-Demand Distance Vector), DSR (Dynamic Source Routing) do not
take into account routes between different networks. Interworking brings about new
dynamics and more routes to select from and energy efficiency should be considered
too.
6. Price
Pricing is always an issue in situations where there is interworking where one network
uses the other network for data transfer. The most commonly suggested architecture
in the 802.11/802.16 interworking network employs 802.16 as a backhaul for the 802.11
hotspots. Depending on how the networks are owned pricing becomes an issue. If both
the networks are owned by separate individual owners, then a service level agreement
has to be reached to decide on a pricing model. The basis of the pricing could be
based on usage, traffic load, bandwidth etc. Another issue related to pricing is the
fairness issue.
The existing work in [9], [26], [25] and [24] do not take into consideration cases where
one network, which has an aggressive contention based channel access scheme, takes up most
of the resources and the other network is starved due to this. Such a scenario could occur
under conditions of heavy network traffic. Under such situations the network should be
able to fairly support users from both the networks that are integrated without drastically
blocking one another. Designing a model for the above mentioned situation of high network
24
traffic in integrated network has motivated the work in this thesis to focus specifically on
resource sharing.
The next chapter will discuss the existing resource sharing models, their drawbacks
and also suggest the new models designed to improve over the drawbacks in the existing
models.
25
Chapter 3
A Prioritized sharing and Forced Sharing Model for Sharing Resources in
Wi-Fi - WiMAX Integrated Networks
3.1 Introduction
Resource allocation in a resource starved environment like wireless networks is impor-
tant to obtain better network performance. Resource allocation challenges are based on fac-
tors like channel conditions, bandwidth requirement and mobility that change with location
and time. Designs which do not consider such diversities lead to inefficient use of the re-
sources. There is no single optimization tool that can be used to solve all resource-allocation
problems simultaneously. Examples of resource allocation strategies are antenna-array pro-
cessing, dynamic resource allocation and game theory. Based on earlier work on resource
allocation over integrated Wi-Fi and WiMAX networks, it is observed that scheduling and
channel assignment are the widely studied resource allocation strategies in the specific area
of integrated Wi-Fi-WiMAX networks. This can be observed in Figure 2.5 in chapter two,
where most of the functions in the MAC layer are related to channel allocation and resource
sharing using scheduling. Designing an efficient resource management scheme in integrated
network is challenging due to various network requirements such as avoiding drastic vari-
ation in data rate, obtaining high channel utilization and avoiding unnecessary handoffs
which reduce the throughput of the system. So choosing the right resource sharing model
or modifying the existing ones to fit new networking standards poses a challenge to network
designers.
26
3.2 Resource Sharing Techniques
This thesis, focuses on resource allocation techniques in the MAC layer of the integrated
Wi-Fi and WiMAX networks. Some MAC layer resource allocation techniques mentioned
earlier are briefly described below.
1. Scheduling: The resource under consideration is bandwidth and scheduling is related
to allocation of channels to the users at specific time intervals by either a centralized
or decentralized controller. This helps users to occupy the limited available bandwidth
for different lengths and periods of time depending on their call requirement, channel
condition and fairness. It helps avoiding collision occurring between users trying to
occupy same channel at a given time. The controller maintains the information of
how to schedule the channels to avoid such collisions.
2. Channel allocation: Limited spectrum availability leads to dividing available band-
width or spectrum into channels and allocating the channels in such a way that mul-
tiple users can use different parts of the spectrum without interference. Co-channel
interference is an issue that is to be confronted while allocating channels as it can
degrade the quality of service, delay and throughput seen by the applications. Chan-
nel allocation can be done based on a central controller or a distributed controller.
In centralized scenario, a central controller makes sure that the channels and they
are appropriately allocated to all the users served. In a distributed scenario, users
communicate among each other to coordinate and achieve proper resource allocation.
The following section will describe the previous resource allocation methods ([9] and
[26]) used in integrated Wi-Fi and WiMAX networks specifically. The various issues related
to the models used earlier in [9] and [26] are also discussed in detail.
3.3 Channel Allocation in Wi-Fi-WiMAX Network Based On Previous Models
To understand the specific channel allocation methods used in this thesis, it is necessary
to have a clear understanding of the system model that is used in Wi-Fi-WiMAX integrated
27
networks. In [9] and [26], as well as in [23] and [24] the integrated network model that is
used is known as tightly coupled model. To understand tightly coupled networks a brief
description of two keywords, core network and access network, is required. Core network
is the central part of a network which provides various services to customers connected
by access network. Core network includes the AAA (Authentication, Authorization and
Accounting) server which does the billing, security control etc. Access network is a part of
the network which connects a subscriber to their immediate service provider. Now if there
are two networks N1 and N2 being tightly coupled, a tightly coupled network is defined as a
network in which both N1 and N2 are integrated in such a way that N1 directly connects to
the Internet through its wired backbone, whereas N2 connects to the Internet through the
core network of N1. For example, if a 3G network and WLAN network are tightly coupled,
then WLAN is integrated with 3G network such that WLAN connects to the Internet
through the core network of 3G network. It could be coupled in way that 3G connects to
the Internet through WLAN core network too. In this case, the billing and authentication
services will be provided by WLAN. This indirectly indicates that N1 network should expose
its core network functionalities such as authentication, authorization and accounting to the
N2 network. Therefore most of the time tightly coupled architecture is adopted by service
providers when both the integrated networks are owned by one service provider. The system
model used in this thesis is an extension of tightly coupled networks over Wi-Fi and WiMAX
networks since the new models proposed in this chapter are an improvement over the existing
models ( complete partitioning, complete sharing, etc. explained in section 3.3.1 and first
come first serve model explained in 3.3.2) which use tightly coupled architecture in [9] and
[26].
3.3.1 System model
The system model used in this thesis considers a tightly coupled architecture based on
WiMax and Wi-Fi networks as shown in Fig. 3.1. In this architecture there is a WiMax base
station encompassing multiple WiFi hotspots. The WiMax BS connects to the Internet and
28
acts as the backbone to the WiFi network. The WiMax subscribers referred to as subscriber
stations, communicate directly to the WiMax Base Station (BS) whereas the WiFi users
have to use a special access point (AP) to communicate with the WiMax BS. This special
access point is named as WiFi -WiMax Bridge (WWB) in this thesis. The WiMax interface
is used for communicating with the BS, and WiFi interface for communicating with WLAN
stations. Since the range of WiFi users are much smaller than the WiMax counterparts
and the access mechanism of WiFi (random) and WiMax (time slotted) are different, the
WWB acts as a link for WiFi users to reach the WiMax BS. The WWB aggregates all the
WiFi user traffic and requests the WiMax base station for a service. The traffic from WiFi-
Wimax bridges are just referred to as WiFi users/traffic since they forward the aggregated
WiFi requests and the traffic from WiMax subscriber stations is referred to as WiMax
users/traffic. The important issue focused on is the prioritization of these two different
traffics and the corresponding changes in channel utilization and blocking of traffic. We
assume that each cell has a total of C channels. The admission control residing in the base
station manages the resource C between the two types of traffics.
Figure 3.1: System Model of WiMax and WiFi integration.
29
General Resource Sharing Models
Before describing the resource sharing models in the Wi-Fi and WiMAX architecture
described above we briefly explain the already existing allocation algorithms in previous
works ([9] and [26]) which are known Complete sharing (CS), Complete Partitioning (CP)
and Hybrid Sharing (HS) schemes.
In CS allocation (Fig. 3.2a), WiFi and WiMax have their own queues and both of them
share all the available channels. There is no prioritization and the users are served on first
come first serve basis. If quality of service is defined as the minimum number of channels
assured for a particular class of users, then there is no QoS guarantee in this scheme.
The CP allocation scheme (Fig. 3.2b) is a variant of CS scheme, where different resource
utilization thresholds are assigned to WiFi and WiMax users based on each traffic priority.
Higher number of channels are reserved for higher priority users, in this case WiMax users.
QoS is achieved by reserving channels for each class of users.
HS scheme (Fig. 3.2c) is a combination of CS and CP schemes. There are a set of
common channels which can be shared by both the class of users and the rest of the channels
are reserved according to the priority of the class of users. In Fig. 3.2c, WiMax and WiFi
are allocated three and two channels respectively and the rest three channels are common
and are allocated on first come first serve basis. In this type of allocation, there is a trade-off
between channel utilization and QoS.
3.3.2 Channel Allocation Problem
Channel allocation has been used as a means of resource sharing in many network
solutions. From the second chapter, considering the work of the two main researchers
Tara Ali Yahiya et al. and Niyato et al., in [26], [9], [23] and [24] it is observed that
there exists a problem in the previous resource sharing model used. The users of one of
the networks involved in integration are blocked unfairly in the existing models. Briefly,
this problem occurs when resources allocation is random without any pre-defined criteria,
where one network captures most of the bandwidth and the other one starves for resource.
30
Figure 3.2: a) Complete Sharing, b) Complete Partitioning, c) Hybrid Sharing.
31
This problem is named as starvation problem in this thesis. The detailed description of th
starvation problem will be provided in the subsection below.
First Come First Serve Model
First come first serve model, is based on the hybrid sharing model which is explained
in section 3.3.1. As seen in Figure 3.3, in the first come first serve model a set of P channels
from a total of N channels are reserved for users from the networks that are integrated.
There can be several networks that are being integrated. However, this thesis focuses on
the only two networks being integrated which are Wi-Fi and WiMAX. The remaining N-P
channels are free to be accessed by users of either Wi-Fi or WiMAX network based on a first
come first serve policy. Now, the P reserved channels between users of the two networks
can be divided in two ways. In the first way, the P channels can be divided equally between
users of Wi-Fi and WiMAX networks, where each network gets P/2 reserved channels for
their users. In the second way, the P channels are divided unequally among the users of
both the networks.
The authors in the previous work [9] use the second way of division so that users of
one network can be given more priority. They give users of Wi-Fi network more channels
in the division than the channel given to WiMAX. The reason being that Wi-Fi users are
not directly connected to the Internet. They have to go through WWB, the integrated
bridge, to reach to WiMAX and the WWB supports multiple Wi-Fi users. They assume
that at any given time, the traffic on WWB would be high and the bandwidth request for
WWB would be higher since it has to serve many 802.11 end users. WWB operates as a
intermeadeate base station in itself in contrast to the WiMAX SS which can be an just an
individual end user.
From Figure 3.3, it can be seen that out of a total of seven channels represented by N,
three channels are given to 802.11 users as compared to only two for WiMAX users. 802.11
users are given more channels under the assumption that WWB which supports multiple
802.11 users have higher traffic. There are two shared channels that can be used by users of
32
Figure 3.3: Threshold based channel distribution between 802.11 and 802.16 users. N=
Total number of channels.
either network on a first come first serve basis. Since the two shared channels can be used
by users from either networks, in a worst case scenario, WiMAX users end up getting only
two channels always and Wi-Fi users get five channels. This division would prove unfair
because with the increase in arrival rate of 802.11 users, 802.16 users are blocked significantly
compared to 802.11 users which indirectly shows that 802.16 users are starved of resources.
The first come first serve model described above is analyzed using Markov Models. The
description of the parameters and the equations used to calculate the parameters in the
analysis is mentioned in Chapter 4 of this thesis.
For the analysis of first come first serve model, the arrival rate of 802.11 users is
kept fixed and the arrival rate of 802.16 users is varied. The fixed arrival rate of 802.11
users are kept high in order to incorporate the assumption that the authors make where
WWB representing 802.11 users traffic (defined in system model) will always be experiencing
high load. The graph shown in Figure 3.4 compares the variation of 802.11 with 802.16
blocking probability when the arrival rate of 802.16 users (represented as ?WiMAX in Figure
3.4)increase.(In this thesis, blocking probability is used as the statistical probability that a
user does not get a channel for transmission due to insufficient transmission resources)
From the graph in Figure 3.4, it can be observed that blocking probabilities of both
802.16 and 802.11 increases when the arrival rate of 802.16 users is increased. This is
33
because when the rate of arrival of 802.16 users increase, the probability of more channels
being occupied increases which in turn reduces the number of channels available to serve
the newly arriving users. This causes the newly arriving 802.16 users to be blocked due to
unavailability of free channels for transmission.
Now, when the blocking probability of 802.11 users is compared with 802.16 users, it
is observed that the blocking probability of 802.16 is higher than that of 802.11. This is
because more number of channels are already reserved for 802.11 and then as the traffic
of 802.11 increases they start using all the shared channels too which starves the 802.16
user. The study of performance of first come first serve model is a repetition of the existing
work, but it is analyzed in the light of the failure of existing models on the basis of resource
starvation of 802.16 users. This result helps in conforming the theory that if the existing
models are used under high load conditions the 802.16 users will experience unfair and
significant blocking compared to 802.11 users and thus 802.16 users indirectly experience
starvation.
Figure 3.4: Direct comparison of blocking probability of 802.11 and 802.16 users with
variation in arrival rate of 802.11 users.
34
The authors assumed that providing more reserved channels to the Wi-Fi users from the
P channels, will offer better connectivity or lesser blocked requests for Wi-Fi networks. The
results of the authors simulation of the first come first serve model confirm their assumption.
However, the authors do not consider a scenario of congested network with heavy traffic.
Under these conditions the resource sharing policy becomes heavily biased to Wi-Fi which
increases the number of WiMAX requests being blocked by a significant amount. So, this
thesis tries to provide a solution by suggesting additions to the already existing models of
resource sharing so that under condition of heavy traffic the resource sharing model provides
unbiased service to both the Wi-Fi and WiMAX users.
3.3.3 Forced Sharing Model
Figure 3.5: Forced sharing mechanism where WiMax user has more priority in the shared
channels and is replacing a WiFi user
The previous model appear biased to 802.11 users under heavy traffic conditions be-
cause the resource sharing is not well defined. The previous model explained the the last
35
section also does not take into consideration a sharing policy for high network load condi-
tion. Therefore, this model focuses on designing a sharing mechanism within the common
channels such that the model shares the available resource in an unbiased manner between
the different users. In this model ,if one of the networks gets more priority through more
reserved channels the other network will adopt an aggressive sharing mechanism in the
common channel region. The aggressive sharing mechanism will force the users with higher
priority to be dropped when the traffic of the network experiencing starvation increases .
As an example from Figure 3.5 if WiFi network is given three channels out of total seven
channels, according to earlier distribution the WiFi users get a minimum of five channels
including the common channels in the worst case. This is because all the reserved as well
as shared channels were distributed on a FIFS (First In First Serve) basis. According to
the FIFS mechanism if the arrival rate of users from WiFi network is more than that of
WiMax then WiMax users get only two channels in the worst case. In our model,WiMax
which has lower priority, even in the worst case would obtain four channels rather than two
in the previous model since they force the WiFi users to be dropped when they experience
starvation i.e when the user traffic of WiMAX increases and no channels are available in
shared space. So finally WiMax will have four channels and WiFi will have three channels
according to the diagram below, which ensures that the users of WiMAX network are not
starved when the WiFi users get unfair advantage. Figure 3.5 given above will represent
the above discussed forced termination model. This theoretically shows that channels are
shared in an unbiased manner between the users even under a situation of highly congested
network. This will be proved analytically using simulation results in Chapter 4.
3.3.4 Prioritized Sharing Model
The problem with the complete partitioning (CP), complete sharing (CS) and Hy-
brid Sharing (HS) schemes that we mentioned earlier are that they solve specifically either
channel utilization or blocking probability issues. They do not tackle both the problems
simultaneously. In CP scheme the un-utilized channels of one class can not be used by
36
other class of users and therefore the capacity is wasted. In CS, during unbalanced loads,
the class of users with low traffic suffer resource starvation. In HS scheme the QoS for each
class is guaranteed but this is achieved at the expense of inefficient total system utilization.
Since the forced termination model also is an extension of a hybrid sharing model, the
channel utilization problem still exists even though the blocking probability of the involved
networks have improved. Our proposed Prioritized Sharing (PS) takes these two problems
into account.
It is important to note the difference between the terms shared, reserved and prioritized.
Shared means that the channels can be accessed by anyone on a first come first serve basis.
reserved means that the channels are allocated to a particular class of users and no users of
any other class can access them. prioritized means that the channels can be accessed when
they are free but a user can be dropped and queued to accommodate other class of users if
certain pre-defined criteria is met.
Figure 3.6: Prioritized channel allocation Model.
In PS scheme as shown in Fig. 3.6, both the WiMax and Wi-Fi users are allowed to
access all the channels if they are free on first come first serve basis like in CS scheme.
This allows the users to use the full capacity of the system. In PS scheme, Quality of
Service is achieved by prioritizing the channels to WiMax and Wi-Fi users according to
their respective QoS requirements rather than strict reservation. In Fig. 3.6, WiMax has
37
five prioritized channels and WiFi has three, but all the eight channels can be accessed by
any user if they are available.
The criteria for call admission and forced termination for WiMax users in Prioritized
resource sharing algorithm is stated as follows:
1. A WiMax call is admitted if the at least one channel is free.
2. A WiMax call is admitted if the at least one channel is free.
3. If all the channels are full and if the number of channels occupied by WiMax users is
already greater than or equal to the number of prioritized channels the call is blocked.
The same procedure holds for Wi-Fi traffic also. Since, the WiMax and Wi-Fi are
allocated a predefined number of prioritized channels, those many channels are assured at
the least in any traffic conditions. Hence, QoS of service is achieved as in CP case as will
be observed in the simulation results.
In the next chapter we design analytical models for the different resource allocation
schemes that we describe above. The results of the simulation of the analytical models
will be described and then we discuss how it improves the channel utilization and blocking
probabilities.
38
Chapter 4
Study of the Improvement in Channel Utilization and Network Starvation
Using the New Sharing Models
As discussed in Chapter 3, most commonly used resource sharing models in homoge-
neous networks (networks in which all components use same protocol or IEEE standard)
are complete sharing, complete partitioning and hybrid sharing shown in Figure ??. From
Chapter 3, it is also observed that a direct adoption of the models used in homogeneous
networks will not work efficiently in integrated networks. In section 3.3.1, the inefficiency
of direct adoption of resource sharing models from homogeneous networks into integrated
networks is discussed using the existing first come first serve model suggested by Ali Yahiya
et al. in [9] and [26]. The first come first serve model is based on the hybrid sharing concept
as mentioned in Chapter 3. It has a total of N channels divided into P reserved channels
and (N-P) shared channels. The P channels can be divided equally between users of Wi-Fi
and WiMAX network or such that more reserved channels are given to a preferred network.
In [9] and [26], the authors assumed that providing more reserved channels to the Wi-Fi
users will offer better connectivity by reducing the number of blocked requests from the Wi-
Fi network. To confirm their assumption, the authors? simulated the first come first serve
model and showed that the blocking of Wi-Fi traffic is under 2% (which is the accepted
rate for voice and calls in cellular networks). However, in a congested network scenario, due
to heavy traffic from Wi-Fi stations, the resource sharing policy becomes heavily biased
towards Wi-Fi users which significantly increases the number of WiMAX requests being
blocked. The unfair blocking of WiMAX under conditions of heavy Wi-Fi traffic is shown
in the graph in Figure 3.4. So, the motivation for the work in this thesis is to suggest
changes to the already existing first come first serve model to provide an efficient solution
39
for unbiased resource sharing between Wi-Fi and WiMAX users under conditions of heavy
traffic.
In Chapter 3, two different resource sharing models were suggested as an improvement
over the first come first serve model. The two new models are Forced termination and
Prioritized sharing models. This chapter provides an analytical model for studying the
efficiency of the two new resource sharing models. Call blocking probability and channel
utilization are the dependencies used to analyze resource sharing efficiency of the proposed
models. Measure of call blocking probability indicates if either Wi-Fi or WiMAX user is
blocked unfairly and the total channel utilization indicates if the network resource is utilized
to its maximum. Typically, a first step in the analysis via computational probability is to
model the behavior of the system as a Markov chain.
4.1 Markov Models
A Markov chain is a random process where all information about the future is contained
in the present state (i.e. one does not need to examine the past to determine the future).
To be more exact, if a process has the Markov property it means that the future states of
the process depend only on the present state, and are independent of past states. Since it
is a stochastic process all state transitions are probabilistic. Modeling as a Markov chain
is possible when system parameters such as service demand (time needed to complete a
job) and interarrival time (time between two consecutive arrivals of jobs) are exponential
random variables or some combinations of exponential random variables. By analyzing the
Markov chain, we can understand the performance of the system with respect to channel
utilization and call blocking probability.
To analytically study forced termination and prioritized sharing models, a traffic model
for wireless networks has to be developed first. The two main characteristics of a traffic
model are:
1. Call arrival process,
40
2. Call holding time
Prior research indicates that call arrival in wireless networks follows a Poisson process
[9], [26], [23]. However, for call holding time, two different approaches can be found. One
approach is to model the call holding time (service rate) as general independent identically
distributed (iid) process. This approach has been intensively studied by Fang et al. [15],
Lin et al. [16], Fang et al. [17], and Bolotin in [18]. The second approach is to model the
call holding time as an exponential distribution. This method has been extensively used by
Hong and Rappaport [4], Zeng et al. [8], and Nanda [19]. Both approaches have their own
advantages. In this thesis call holding time is modeled as an exponential distribution for
simplicity of analysis. Also, in the previous works related to resource sharing in integrated
networks service rate is modeled as exponential distribution [9], [26].
With the base of the analytical modeling established, the next sections will describe in
detail the forced termination model, the prioritized model, the Markov models representing
them and results of analysis of each model.
4.2 The Forced Termination Model
In this section, a Markov model representing the forced termination model is designed.
The Markov model is simulated using MATLAB. The first come first serve model has
been analyzed in [9] and [26] from the perspective of Wi-Fi users. In section 3.3.1, we
analyze the first come first serve model and observe that the first come first serve model is
biased towards Wi-Fi users traffic when the network is congested. Therefore in this section
a forced termination model which improves over the drawbacks of first come first serve
model is designed. The proposed model provides a sharing mechanism within the common
channels such that the users of Wi-Fi network which are given more reserved channels,
gets lesser priority in the common channels. Common channels are another term used for
shared channels which can be accessed by both Wi-Fi and WiMAX users. In first come
first serve model, Wi-Fi users occupy all the shared channels as well as channels reserved
for the them, when the network is congested by high arrival rate of Wi-Fi users. However,
41
in forced termination model since WiMAX users are given higher priority in the shared
channels they can forcefully drop or terminate a Wi-Fi user service in the shared channels.
But, this privilege is provided to WiMAX users only when all the channels are full and
Wi-Fi users have occupied all the shared channels.
Figure 4.1: Forced sharing mechanism where WiMax user has more priority in the shared
channels and is replacing a WiFi user
For example in Figure 4.1, if WiFi network is given three channels out of seven, accord-
ing to first come first serve model the Wi-Fi users get a minimum of five channels including
the common channels. The channels are distributed on a first come first serve basis. Ac-
cording to the first come first serve mechanism if the arrival rate of WiFi users is more than
that of WiMax users, then WiMax users get only two channels in the worst case scenario.
In our model even in the worst case scenario, , WiMAX with higher priority would obtain
four channels rather than two as shown in Figure 4.1. So finally WiMax will have four
channels and WiFi will have three channels, which indirectly implies an unbiased sharing
of resources between Wi-Fi and WiMAX users in terms of channel bandwidth. The forced
42
termination model is analytically modeled using Markov chains to observe the unbiased
sharing of channels between Wi-Fi and WiMAX users.
In order to design the Markov model, it is necessary to understand all the parameters
in the forced termination model. Figure 4.1 represents the forced termination model. The
lighter shaded channels are reserved for WiMAX users and the darker shaded channels are
reserved for Wi-Fi users. The number of channels reserved for WiMAX users is denoted by
t1 and that for Wi-Fi users by t2. It can be observed from the Figure 4.1 that more channels
are reserved for Wi-Fi users i.e. t2 ? t1. From the discussion in Section 3.3.1, the reason
for providing more channels for Wi-Fi users is made clear. However, it can also be seen
that using forced termination model, when more WiMAX users arrive they can drop Wi-Fi
users being served in the shared channels and thereby not be blocked due to unavailability
of channels as compared to first come first serve model.
Figure 4.2: Markov Model representing Forced Termination Sharing
In Figure 4.2, each state in the system is represented by a three-tuple of non-negative
integers (i,j,k), where
? i is the number of channels used by WiMAX users or the number of WiMAX users
served by the system at any instant.
43
? j is the number of channels used by Wi-Fi users or the number of Wi-Fi users being
served by the system at any instant.
? k represents the states in which users are dropped. k takes two values 0 and 1. If k
is 1, then that state represents a dropped Wi-Fi user.
Value ranges of the discrete parameters are i ? [0,N?t2], j ? [0,N?t1] and k = [0,1].
As shown in Figure 4.2, the first come first serve model is represented by a three dimensional
Markov chain. The total number of possible states in the Markov chain shown in Figure
4.2 is given by,
TE = [[((N?t1)+1)?t1]+12[[(((N?t1)+2)?((N?t2)+1))]?12[((t2?(t2+1)))]]+(N?t1?t2)+1]
(4.1)
Based on the Markov chain diagram in Figure 4.2 the number of equations and number
of states can be found by substituting the values
? ?1 or ?w1 is the arrival rate of WiMAX users,
? ?2 or ?w2 is the arrival rate of Wi-Fi users,
? ?1 or ?w1 is the service rate of WiMAX users, and
? ?2 or ?w2 is the service rate of Wi-Fi users.
? P(i,j) represents state transition probability which is the probability of a user staying
in a state(i,j).
In Figure 4.2, consider a state (2,2). This state indicates that there are two WiMAX
users and two Wi-Fi users in this system. When a new WiMAX user arrives with arrival
rate ?w1 the system moves from state (2,2) to state (3,2) since i represents the number of
WiMAX users in the system. Similarly, when a WiFi user arrives with arrival rate ?w2
the system moves from state (2,2) to state (2,3). When the WiMAX user is serviced, the
system moves back to state (2,2) from state (3,2) with probability ?w1. States in which
44
Figure 4.3: Balance equations representing first come first serve model
the sum of i and j is equal to N i.e. state (3,4), state (4,3), state (2,5) and state (5,2),
both Wi-Fi and WiMAX users are blocked. However,in the last state at each row, for
example at state (5,0), only WiMAX users get blocked since they have used all available
channels. The available channels include three shared channels and two channels reserved
for WiMAX(which provide a maximum of five channels for WiMAX). Any WiMAX user
arriving after this stage is blocked. Similarly at state (0,5) all WiFi users are blocked
according to Figure 4.2, since Wi-Fi users would have used all their available channels.
From this discussion we derive the formula for blocking probability of Wi-Fi and WiMAX
given by the equations in Figure 4.4.
45
Figure 4.4: Equations to calculate Blocking probability of both WiFi and WiMAX users
4.2.1 Simulation And Results Of Forced Termination Sharing Model
Blocking probability of WiMAX and Wi-Fi users using Forced Termination
Model
In the first come first serve model discussed in section 3.3.1, due to biased channel
bandwidth distribution between Wi-Fi and WiMAX users, WiMAX users experience unfair
blocking. This is because all shared channels get occupied by Wi-Fi users and WiMAX users
are starved in terms of channel availability. In the first come first serve model suggested by
Yahiya et al. [9], the authors? make two assumptions, which are:
? The traffic on WWB in Figure 3.1, through which Wi-Fi users connect to WiMAX
BS, have higher traffic (arrival rate) as compared to the WiMAX BS.
? More reserved channels are allocated for Wi-Fi users than for WiMAX.
The reasons for these assumptions are discussed in section 3.3.1. The forced termination
model is designed to overcome the drawbacks of the first come first serve model. Therefore,
46
Figure 4.5: 802.11 blocking probability variation with increase in number of reserved chan-
nels
the Markov chain representing the forced termination model is simulated using the same
assumptions listed above to give a fair comparison with first come first serve mode. The
arrival rate of Wi-Fi users, ?w2 is kept as 0.8 and the arrival rate of WiMAX users,?w1 is
increased from 0.1 to 1. The service rate ?w1 and ?w2 are kept as 0.05 as per the simulations
in first come first serve model suggested by Yahiya et al. [9].
The graph in Figure 4.5 shows variation of Wi-Fi and WiMax user blocking probability
with increase in WiMAX user traffic in forced termination model. Comparing the graphs
in Figure 4.5, representing forced termination model and Figure 4.6, representing first come
first serve model, it is observed that the blocking probability of WiMAX reduces significantly
using forced termination model. Using forced termination model it is observed that both
the Wi-Fi and WiMAX users experience almost similar blocking probability. This implies
that either both Wi-Fi and WiMAX users are being served in an unbiased manner or they
are not being served (blocked) in an unbiased manner. This shows that forced termination
model is successful in achieving the objective for which it was designed, which is to achieve
47
Figure 4.6: Direct comparison of blocking probability of 802.11 and 802.16 users with
variation in arrival rate of 802.11 users.
unbiased resource distribution between Wi-Fi and WiMAX users even when Wi-Fi user
traffic is high compared to WiMAX.
Effect of increased available channels on WiMAX and Wi-Fi users blocking
probability
The effect of variation in the number of channels on the blocking probability of Wi-Fi
and WiMAX has not been studied by Yahiya et. al in [9] for the first come first serve model.
The knowledge of this effect can help service providers to estimate the minimum QoS they
can provide to the subscribers in terms of call blocking probability. This section studies
the effect of the variation in the number of available channels on the blocking probability
for WiMAX users. From the graph in Figure 4.7 it is clear that with increase in number of
available channels, there is a decrease in the blocking probability of WiMAX users.
48
Figure 4.7: 802.16 blocking probability variation with increase in number of channels
Variation of reserved channels on blocking probability of WiMax and WiFi
Users
The graphs in Figure 4.8 shows the variation of blocking probability experienced by
both WiMAX and Wi-Fi users with the change in number of channels revered for each of
the network. The blocking probability is plotted against the arrival rate of the WiFi and
WiMAX users. The first graph in Figure 4.8 studies how the blocking probability of WiMAX
user changes with respect to two factors. The first factor is number of reserved channel and
the second factor is change in arrival rate of WiMAX users. This study can be done by
choosing an arrival rate on the first graph in Figure 4.8. For a 0.8 arrival rate of WiMAX
users, we observe the change in WiMAX blocking probability with variation in number of
channels reserved for WiMAX. It is seen that when the number of channels reserved for
WiMAX increases the blocking probability of WiMAX users decreases because for a fixed
number of users in the system more channels are provided. More channel provided to a
49
Figure 4.8: (a)Blocking probability of WiMax with change in WiMax load as well as change
in channels reserved for WiFi users.(b)Blocking probability of WiFi with change in WiFi
load as well as change in channels reserved for WiFi users.
fixed number of users increases the probability that they get served and therefore decreases
blocking.
The blocking probability of WiMAX users can also be studied against the change in
WiMax user traffic keeping the number of reserved channels a constant. The blocking
probability of WiMAX users increase with increase in WiMAX arrival rate because more
users are competing for a fixed number of channels. The trend indicated by the analytical
study shows similarity in the pattern of variation of blocking probability of 802.16 and
802.11, which is blocking probability increases with the increase in load. In the case of
802.11 blocking probability variation with channel reservation, the difference is that for
more channels reserved for 802.16 the blocking probability of 802.11 users increase. This
is because 802.11 users end up having lesser channels when more channels are reserved for
802.16 users and moreover 802.16 has more priority in the shared channels. The same trend
is observed in Wi-Fi users also.
Throughput improvement of WiMAX users
In the forced termination model, when the WiMAX uses get starved as their traffic
increases the WiMAX users forcefully drop WiFi users who already have higher priority in
50
Figure 4.9: Comparison of WiMAX throughput between first come first serve model and
forced termination model
terms of number of reserved channels. This results in resource starved WiMAX users getting
more channels than in the first come first serve model, and their blocking probability reduces
significantly. The throughput of WiMAX users is calculated from the equations in Figure
4.4. Throughput is given obtained by taking the product of service rate and the probability
that a user is served. The graph in Figure 4.9, compares the throughput of WiMAX users
using forced termination model to the first come first serve model. The throughput of
WiMAX users is observed with variation of WiMAX user traffic. The arrival rate of Wi-Fi
users is kept high (0.8) as the assumptions made in first come first serve model by Yahiya
et al. The WiMAX throughput increased by an average of 30% using forced temination
model. This can be attested to the fact that WiMAX users get more channels in forced
termination model than they get in first come first serve model. Therefore more WiMAX
users get served.
However, the WiFi users that used to get free access of the shared channels apart from
channels reserved for them, will experience an increase in blocking probability as shown in
51
Figure 4.10: Comparison of Wi-Fi blocking probability between first come first serve model
and forced termination model
Figure 4.10. The graph in Figure 4.10 shows the comparison of WiFi blocking probability
in forced termination with first come first serve model. Even though the Wi-Fi users
experience an increase in blocking probability they will receive the minimum guaranteed
channels which are reserved for them. The number of reserved channels for Wi-Fi users
is more than the channels reserved for WiMAX users in forced termination as well as first
come first serve model.
Throughput of WiMAX and Wi-Fi with the change in number of channels
reserved
An alternative analysis of the forced termination model would be to keep the arrival
rates of both WiMAX and Wi-Fi users constant and observe the variation of throughput
of both the networks. The variation in throughput is analyzed with the change in number
of channels reserved for both the network. When more channels are reserved for WiMAX
the throughput of WiMAX increases but throughput of WiFi decreases since they are given
lesser reserved channels. Similarly, when more channels are reserved for Wi-Fi the through-
put of Wi-Fi increases but the throughput of WiMAX decreases. The graphs in Figure 4.11
shows that when the reserved channels are equally distributed between WiFi and WiMAX
52
Figure 4.11: (a)Variation of WiMax throughput with change in channel reservation.
(b)Variation of WiFi throughput with change in channel reservation.
users, the throughput of both Wi-Fi and WiMAX network is the same. All the other ways
of distribution of reserved channels between both the users result in either the throughput
reducing or staying almost constant after the maximum value of throughput.
Figure 4.12: 802.11 dropping probability variation with increase in number of reserved
channels
However, considering the dropping probability shown in Figure 4.12, it is observed
that dropping probability is maximum when the reserved channels are distributed equally
53
between Wi-Fi and WiMAX users. This can be explained by the fact that when the reserved
channels are equally divided between Wi-Fi and WiMAX, the number of shared channels
become highest when compared to any other division. When there are more shared channels,
the probability that the Wi-Fi user gets dropped increases proportionally. So equal division
of reserved channels can be beneficial for throughput of WiMAX users but it increases the
dropping probability of WiFi users too. Therefore, the distribution of reserved channels
should be made adaptive to the network characteristics such as traffic. It can also be varied
based on the network condition and requirement of the subscribers too.
The forced termination model proves beneficial for the WiMAX users in terms of un-
biased resource sharing and improved throughput. The forced termination model provides
an efficient improvement over the drawbacks of first come first serve model. The first come
first serve model is designed based on hybrid sharing policy discussed in chapter 3. From
a total of N channels, P channels are reserved for different users in the network and the
remaining (N-P) channels are maintained as shared channels which can be accessed by any
user when they have exhausted the number of channels reserved for them. In first come first
serve model, the users of one network are not allowed to use the channels reserved for users
from other network even when they are not being used. This leads to under utilization of
the spectrum. The next section discusses another model proposed to improve the channel
utilization in integrated wireless networks.
4.3 Prioritized Sharing
In prioritized sharing instead of dividing the N total channels into shared and reserved
channels as in first come first serve model shown in Figure 4.13, the N total channel are
divided among the users of the networks that are integrated. The channels are divided
among the users of both Wi-Fi and WiMAX network based on pre-assigned priorities. But,
the prioritized model differs from first come first serve model or hybrid model by avoiding
allocation of a specific set of shared channels, instead in prioritized sharing any of the
unused channels or free channels act as shared channels. For example, in Figure 4.14 and
54
Figure 4.15, we see that from total of twelve channels, any of five channels are prioritized
for Wi-Fi and the remaining seven channels prioritized for WiMAX users. The prioritized
sharing model is explained in detail using three possible scenarios. The scenarios are:
Figure 4.13: Threshold based channel distribution between 802.11 and 802.16 users. N=
Total number of channels.
? Scenario 1: Wi-Fi users have occupied the number of channels allotted for them which
are seven and WiMAX users have also occupied all the five channels allotted for them.
Here all twelve channels are occupied and the both the networks have occupied all the
allotted channels for them. In this situation if a new Wi-Fi or WiMAX users arrives
then they are blocked due to no available channels for transmission.
? Scenario 2: Wi-Fi users have occupied only four out of the seven channels prioritized
for them and WiMAX has occupied all five channels assigned to them. If a WiMAX
user arrives then it is allowed to occupy any of the 3 free channels which are not
occupied by Wi-Fi.
? Scenario 3: Wi-Fi users have occupied only four of the seven channels prioritized for
them and WiMAX users have occupied eight channels which are three more than the
number of channels WiMAX is allowed to use. This happens when Wi-Fi traffic is
low and WiMAX traffic is high. But, when Wi-Fi traffic is increasing and one more
Wi-Fi user arrives then Wi-Fi user can drop any three WiMAX user since WiMAX
has exceeded the number of channels allowed for them by three channels.
55
Figure 4.14: Prioritized channel allocation Model.
The criteria for call admission and forced termination for WiMax users in Prioritized
resource sharing algorithm is stated as follows:
1. A WiMax call is admitted if the at least one channel is free.
2. If all the channels are full but the number of channels occupied by WiMax users is
less than the number of prioritized channels, then the WiFi user who has been most
recently admitted is forced to terminate and is queued back until the next channel is
free.
3. If all the channels are full and if the number of channels occupied by WiMax users is
already greater than or equal to the number of prioritized channels the call is blocked.
Prioritized sharing improves channel utilization compared to previous models like hy-
brid sharing, complete sharing and complete partitioning. In the case of hybrid sharing and
forced termination, we observe from the diagram shown above and the table below that the
56
Figure 4.15: Prioritized channel allocation Model.
channel utilization is not good. This is because at any given time the number of channels
that the previous models (hybrid sharing or forced termination model) could obtain is only
the reserved channels assigned to them even if the channels were unoccupied in the reserved
section for the other networks. Another advantage is that this model fits perfectly for tight
coupled networks, where the channels are to be assumed to be homogeneous in [9]. The
traffic of both the networks can access any of the channels available and will be dropped only
in case the channels occupied by one network exceeds the allocated number of channels for
that network and the traffic of the other network is blocked due to no available free channel.
We represent this model using a 2-D Markov chain and simulate it using MATLAB.
To design the Markov model, it is necessary to understand all the parameters in the
prioritized sharing model. WiMax and WiFi network integration architecture is considered
as shown in Fig. 3.1. The traffic from WiFi-WiMAX bridges just referred to as WiFi traffic
and the traffic from WiMax subscriber stations is referred to as WiMax traffic as mentioned
57
earlier. Both the traffics are assumed to have Poisson arrivals with average arrival rates
of ?WiFi and ?WiMax for WiFi and WiMax traffic respectively. ?WiFi and ?WiMax are
varied from 0 to 1. The average service rates are ?WiFi = ?WiMax = 0.005 all through
the simulations. The total number of channels used in this simulation, N is equal to 10.
The division of the channels to both Wi-Fi and WiMAX users and how the channels can
be accessed by WiMax and WiFi users in different models is summarized in Table. 4.1.
?Shared channels? mean that the channels are accessible by any type of users. ?Reserved
channels? mean that the channels are allocated for a particular class of users and others
can?t access them. ?Prioritized channels? mean that the channels are accessible by all users
but priority is given to the particular type of users to whom the channels are prioritized in
Chapter 3.
Table 4.1: Channel Allocation
Allocation Algorithm WiMax Accessible WiFi Accessible
Channels Channels
Complete sharing 10 shared 10 shared
Complete Partitioning 7 reserved 3 reserved
Hybrid Sharing 4 reserved, 3 reserved,
3 shared 3 shared
Prioritized Sharing 10 shared, 10 shared,
7 prioritized 3 prioritized
Figure 4.16, represents the 2-D model for the prioritized sharing model. In this two
dimensional model the rows, which is represented by i indicate the number of WiMAX users
in the network and the columns represented by ?j? indicate the number of Wi-Fi users. Each
circle in the figure represents a state of the Markov model. State (i, j) shows that there are
i WiMAX users and j Wi-Fi users in the system at that instant. The probability of the
system to stay in a state (i, j) is described as steady state probability and is represented as
P (i, j). ?a represents the arrival rate of a WiMAX user and with the arrival of a WiMAX
user the system moves from state (i, j) to state (i+1, j). Similarly with an arrival of Wi-Fi
user represented as ?b the system moves from state (i, j) to (i, j+1). ?a and ?b are the
58
Figure 4.16: Markov Model representing Prioritized Sharing
service time (holding time) or the time taken by the channel to serve the user. In this figure
we focus on the end states, which are represented by states where the sum of i and j equals
the total number of channels N. The number of prioritized channels prioritized from a total
of N channels for Wi-Fi and WiMAX are represented as t and s respectively.
Figure 4.16, shows that when the system reaches state (0, N) which means the all the N
channels are occupied by WiMAX users, then when one new Wi-Fi user arrives represented
by ?b, then one WiMAX user gets dropped to free a channel for the Wi-Fi user since it has
not received its prioritized number of channels which is t channels. The system moves from
state (0, N) to (1, N-1). This process continues till all the prioritized channels of WiFi are
occupied which will be at state (s, t). At state (s, t) both the users would have occupied
all the channels prioritized for them and any new user will be blocked. Apart from state
(s, t) where users of both the networks have occupied their respective prioritized channels,
WiMAX users are blocked when all the channels are full and WiMAX users have occupied
more channels that what is prioritized for them i.e. more than s channels. Similarly, Wi-
Fi users are blocked when all the channels are full and Wi-Fi users have occupied more
channels that what is prioritized for them i.e. more than s channels. We derive the formula
for blocking probability of Wi-Fi and WiMAX users from this model, by adding the state
probability of all the states where both Wi-Fi and WiMAX will be blocked.
59
Simulation Results
In this section, the Prioritized Sharing scheme is compared with complete sharing, com-
plete partitioning and hybrid sharing. The blocking probabilities and the channel utilization
of these allocation schemes are studied. The definitions of these performance parameters is
as follows:
Suppose the set S contains all the possible states, S = {(i,j)|(i,j) ? N} of the two
dimensional Markov chain representing prioritized sharing model. s and t are the number
of prioritized channels for WiMAX and Wi-Fi traffic respectively. Since s channels are
reserved for WiMax traffic, WiMax users will suffer blocking only in those states in which
WiMax users have occupied more than s channels. If all the channels reserved for WiMAX
users are not used, then Wi-Fi users are allowed to access it. Both Wi-Fi and WiMAX
users are blocked at the same time only when users of both the networks have occupied all
the channels prioritized for each of them. Suppose, all the states where WiMax users suffer
blocking are represented as HWiMax = {(i,j)|(i,j) ? S,i ? s} and the states where WiFi
users suffer blocking are, HWiFi = {(i,j)|(i,j) ? S,i ? s}. Blocking probability is defined
as the states in which Wi-Fi and WiMAX users will be blocked due to unavailability of
channels to transmit.
Then the Blocking probability of WiMax users is given by the expression,
BWiMax = ?WiMax summationdisplay
(i,j)?HWiMax
p(i,j) (4.2)
and the expression for Blocking probability of WiFi users is,
BWiFi = ?WiFi summationdisplay
(i,j)?HWiFi
p(i,j) (4.3)
Channel utilization is defined as the average number of available channels that are
occupied for a specified set of arrival and service rates for Wi-Fi and WiMAX users. In
60
other words, it is the percentage of channels from a total of N=10 channel that Wi-Fi and
WiMAX occupy together. Channel utilization can be derived as,
UTotal = N ?100 summationdisplay
(i,j)?S
(i+j)p(i,j) (4.4)
The percentage of channels occupied by WiMax users from the total of N channels and
the percentage of channels occupied by WiFi users from a total of N channels are referred
to as WiMax Channel utilization and WiFi Channel Utilization respectively. Individual
channel utilization helps to observe if user of any network is starved for resources. Channel
utilization is derived as follows:
Percentage of WiMax Channel Utilization,
UWiMax = N ?100U summationdisplay
(i,j)?S
ip(i,j) (4.5)
Percentage of WiFi Channel Utilization,
UWiFi = N ?100U summationdisplay
(i,j)?S
jp(i,j) (4.6)
Blocking probability
The graph in Figure 4.17 shows that for complete sharing the blocking probability is
highest. This is because the traffic from both Wi-Fi and WiMAX networks can access all the
channels that are available. We can see that initially, the blocking for WiMax is negligible
when the WiFi traffic is zero. This is because when WiFi traffic is zero; all the channels
in the system are available to WiMax users. Once the traffic of WiFi increases, the WiFi
users start competing for the channels and start occupying more channels. As a result, the
blocking probability of WiMax users increases with increase in WiFi traffic. In the case of
complete partitioning, specific number of channels is reserved for both Wi-Fi and WiMAX
users. The blocking probability of WiMAX users remain a constant because for any given
61
traffic of WiFi, the number of reserved channels for WiMax remains a constant and no
channels reserved for Wi-Fi users can be accessed by WiMAX even if they are not being
used. Therefore the blocking probability of WiMAX users stays constant with variation in
Wi-Fi traffic.
In hybrid sharing we observe that the blocking probability lies between the blocking
probability of complete sharing and complete partitioning as expected. This is because in
comparison with complete sharing model, hybrid sharing is allocated a minimal number
of reserved channels. So the users are blocked only when they compete for the shared
channels and when the shared channels are full. In complete sharing, there are no reserved
channels, so all the channels are competed for right from the beginning. In hybrid sharing,
WiMAX can only access the shared channels and channels reserved for them. It cannot
access the channels that are reserved for Wi-Fi even if they are free. However in the case
of prioritized sharing, we can see that the blocking probability is lowest. It starts of along
with complete sharing because like in complete sharing when the traffic of WiFi is zero, all
the channels are available for WiMAX, making it blocking negligible. But, when the traffic
of WiFi increases the blocking probability of WiMAX also increases since Wi-Fi users start
occupying more channels. But the blocking probability of WiMAX users remain lower than
complete sharing since there are a minimum number of channels reserved for WiMAX as
well as Wi-Fi users.
The graph in Figure 4.18 shows the variation of WiFi blocking probability with the
increase in WiFi user traffic for the different models simulated. It is observed that complete
partitioning has highest blocking probability. This is because specific number of channels is
reserved for Wi-Fi users and it cannot access the channels reserved for WiMAX users even
if it is not being used. In complete sharing with the increase in WiFi traffic, WiFi users get
more channels since all the channels are open to access. Therefore the blocking probability
of WiMAX users is lesser than complete partitioning and hybrid sharing models.
In prioritized sharing, the blocking probability is same as that of complete sharing
till a Wi-Fi user arrival rate of 0.2. After that, the blocking probability of WiFi reduces
62
 0.1
 0.15
 0.2
 0.25
 0.3
 0.35
 0  0.2  0.4  0.6  0.8  1
WiMax Blocking Probability, 
B WiMax
WiFi user traffic, ?WiFi
Prioritized Sharing
Complete Sharing
Complete Partitioning
Hybrid Sharing
Figure 4.17: Blocking probability of WiMax Users with varying WiFi user traffic.
slightly. This is because even though all the channels are open to access like in complete
sharing, prioritized sharing provides a guarantee of minimal number of channels to Wi-
Fi users due to reservation. This explains why the blocking probability of WiFi users is
lower in prioritized sharing than in complete sharing. Hybrid sharing however has blocking
probability between complete sharing and complete partitioning as expected.
Fig. 4.19 shows the blocking probability of WiMAX users as their traffic is varied
keeping the WiFi traffic constant. Again, CS should have provided the least blocking
because all the ten channels are accessible to the WiMax users from Table. 4.1. But, due
to prioritization of channels in PS allocation, PS scheme offers the least blocking than any
other allocation method.
Channel Utilization
In this section, the total percentage channel utilization and the WiMAX, WiFi channel
utilization as are studied. Fig. 4.20 shows the Total channel utilization of the channels
as WiFi user traffic is varied and WiMax traffic constant. Both total utilization as well
63
 0
 0.1
 0.2
 0.3
 0.4
 0.5
 0.6
 0.7
 0.8
 0.9
 0  0.2  0.4  0.6  0.8  1
WiFi Blocking Probability, 
B WiFi
WiFi user traffic, ?WiFi
Prioritized Sharing
Complete Sharing
Complete Partitioning
Hybrid Sharing
Figure 4.18: Blocking probability of WiFi Users with varying WiFi user traffic.
 0.1
 0.15
 0.2
 0.25
 0.3
 0.35
 0.4
 0  0.2  0.4  0.6  0.8  1
WiFi Blocking Probability, 
B WiFi
WiMax user traffic, ?WiMax
Prioritized Sharing
Complete Sharing
Complete Partitioning
Hybrid Sharing
Figure 4.19: Blocking probability of WiFi Users with varying WiMax user traffic.
as individual utilization is high in prioritized sharing as compared to other models. In
prioritized sharing since all the channels are available for access when the traffic is low, at
64
any given time the total channel utilization is highest. This is similar to both complete
sharing and hybrid sharing. When the Wi-Fi user traffic increases from 0.2 to 1 in each
model the total utilization of the channels becomes full. In the case of complete partitioning
what happens is that when the traffic of one network is low and the other is high. The
reserved channels are occupied but, the channels reserved for other network user is not
allowed to be accessed even when it is free causing the total utilization to be low compared
to the other models. In complete partitioning users are blocked even when free channels
are available. The total channel utilization of hybrid sharing model is close to complete
sharing and prioritized because with increase in traffic all the channels are utilized. This
is because apart from reserved channels that are provided to each, there are also shared
channels which can be accessed by users of either network.As was discussed, total channel
utilization of CS scheme is ideal since all classes of users can access all channels. Since, PS
carries the same property that any channel can be accessed by any class of users when they
are free, the channel utilization is exactly equal to CS. Complete partitioning has the least
channel utilization because free channels of one class of users can not be utilized by other
class of users and hence they lay idle.
Fig. 4.21 summarizes the advantages of Prioritized Sharing by showing the total chan-
nel utilization and WiMax channel utilization together. Each bar represents the total chan-
nel utilization and the shaded region of each bar represents the WiMax share of the utilized
channels. It can be seen that the total channel utilization is almost the same for hybrid,
complete and prioritized sharing. What is the improvement achieved by Prioritized sharing
then? The answer lies in the observation of individual channel utilization. In this study we
choose channel utilization of WiMax users. It can be observed that even though the total
channel utilization is the same in complete sharing and prioritized sharing, the WiMAX
channel utilization in prioritized sharing is highest, which shows that WiMAX users are no
longer starved for channels with the increase in Wi-Fi users traffic. In complete sharing
the WiMax channel utilization reduces drastically with increase in WiFi user traffic. This
65
 55
 60
 65
 70
 75
 80
 85
 90
 95
 100
 0  0.2  0.4  0.6  0.8  1
Percentage Channel Utilization
WiFi user traffic, ?WiFi
Prioritized Sharing
Complete Sharing
Complete Partitioning
Hybrid Sharing
Figure 4.20: Total Percentage Channel Utilization with varying WiMax user traffic.
is because of a first come first serve policy that completes sharing uses. The channel uti-
lization of WiMAX users in hybrid sharing is better than complete sharing because of the
reserved channels which guarantee a minimum number of channels for WiMax in hybrid
sharing. Complete partitioning the WiMax channel utilization is same as that in prioritized
but their total channel utilization is lowest. In prioritized both total channel and individual
channel utilization is highest.
66
 30
 40
 50
 60
 70
 80
 90
 100
0.2 1 0.2 1 0.2 1 0.2 1
% Channel Utilization
WiFi user traffic, ?WiFi
WiMax
Utilization
Total
Utilization
Complete
Sharing
Complete
Partition
Hybrid
Sharing
Prioritized
Sharing
Figure 4.21: Total Percentage Channel Utilization with varying WiMax user traffic.
67
Chapter 5
Conclusion
With the advance in wireless technology and the demands of high bandwidth, total
channel utilization, global roaming etc. in 4G networks, integration of existing wireless
standards is an imminent solution. However, it has been observed that along with such
an integration, there would be technical challenges to envision it. Resource allocation
which deals with channels utilization, bandwidth usage and various other network efficiency
features, is the most important technical issue in layer 2. Two type of resource allocation
methods were analyzed and performance analysis and the importance of the allocation
schemes were studied.
The first scenario where the two networks, WiFi and WiMax, were tightly coupled and
WiFi was given priority in terms of more reserved channels, it is observed that there is a
starvation problem. The first algorithm alleviates the starvation problem by terminating
the users with undue advantage, and making space for the other network users. Since this
algorithm uses a two threshold mechanism the channel utilization is compromised even
though it provides a solution to the network starvation problem.
The second scenario however, takes into account the starvation as well as the channel
utilization issue. Prioritized Sharing (PS) algorithm is proposed for resource sharing in
WiMax-WiFi integrated networks and later extended to application in multiple integrated
networks. The Complete Sharing (CS) approach which is well studied in literature offers
best channel utilization and Complete Partitioning (CP) approach assures minimum quality
of service to each class in terms of allocated channels. However, both complete partitioning
and complete sharing do not offer good channel utilization. Prioritized Sharing algorithm is
designed such that the best of the two algorithms is achieved. The algorithm was modeled
and analyzed using two dimensional continuous time Markov chains. It is observed through
68
simulations that PS resource sharing algorithm offers least blocking probabilities and highest
channel utilization for WiMax and WiFi users compared to other models.
These solutions clearly indicate the necessity of a resource allocation method specifically
for WiFi WiMax integrated network. The existing solutions were designed for homogeneous
network and networks with multiple user class. Two important Issues in integrated network
related to resource sharing was studied and two algorithms were proposed and analyzed to
measure the improvement it provides in the integrated WiFi-WiMax networks.
69
Bibliography
[1] B. B. Chen, ?Heterogeneous network resource management,? in
http://www.comp.nus.edu.sg/ chenbinb/heterogeneous/index.htm.
[2] J. Ibrahim, ?4g features,? in Bechtel Telecommunications Technical Journal, vol. 1,
no. 1, Dec. 2002, pp. 11?14.
[3] K. Young Kyun and R. Prasad, ?4g roadmap and emerging communication technolo-
gies.? Norwood, MA, USA: Artech House, 2006, pp. 12?13.
[4] S. Hussain, Z. Hamid, and N. S. Khattak, ?Mobility management challenges and issues
in 4g heterogeneous networks,? in InterSense ?06: Proceedings of the first international
conference on Integrated internet ad hoc and sensor networks. New York, NY, USA:
ACM, 2006, p. 14.
[5] I. T. G. d, ?Ieee std 802.16-2004,? October 2004.
[6] H. Mahmoud, H. Arslan, and M. Ozdemir, ?Initial ranging for wimax (802.16e) ofdma,?
in Military Communications Conference, 2006. MILCOM 2006. IEEE, Oct. 2006, pp.
1?7.
[7] R. Iyengar, P. Iyer, and B. Sikdar, ?Analysis of 802.16 based last mile wireless net-
works,? in Global Telecommunications Conference, 2005. GLOBECOM ?05. IEEE,
vol. 5, Dec. 2005, pp. 5 pp.?3127.
[8] D.-H. Cho, J.-H. Song, M.-S. Kim, and K.-J. Han, ?Performance analysis of the ieee
802.16 wireless metropolitan area network,? in Distributed Frameworks for Multimedia
Applications, 2005. DFMA ?05. First International Conference on, Feb. 2005, pp. 130?
136.
[9] T. Ali-Yahiya, H. Chaouchi, A.-L. Beylot, and G. Pujolle, ?Threshold based wimax
resource reservation,? in Mobility ?06: Proceedings of the 3rd international conference
on Mobile technology, applications & systems. New York, NY, USA: ACM, 2006,
p. 40.
[10] Motorola and Intel, ?Wimax and wifi together: Deployment models and user scenario,?
vol. 2, Sept. 2007.
[11] K. Gakhar, A. Gravey, and A. Leroy, ?Iroise: a new qos architecture for ieee 802.16
and ieee 802.11e interworking,? in Broadband Networks, 2005. BroadNets 2005. 2nd
International Conference on, Oct. 2005, pp. 607?612.
70
[12] N. Thomas, M. Willis, and K. Craig, ?Analysis of co-existence between ieee 802.11
and ieee 802.16 systems,? in Sensor and Ad Hoc Communications and Networks, 2006.
SECON ?06. 2006 3rd Annual IEEE Communications Society on, vol. 2, Sept. 2006,
pp. 615?620.
[13] V. Gunasekaran and F. Harmantzis, ?Financial assessment of citywide wi-fi / wimax
deployment,? Communications and Strategy, vol. 3rd Quarter, no. 63, pp. 131?153,
2006.
[14] X. Jing, S.-C. Mau, D. Raychaudhuri, and R. Matyas, ?Reactive cognitive radio al-
gorithms for co-existence between ieee 802.11b and 802.16a networks,? Proceedings of
IEEE Globecom, vol. 5, 2005.
[15] X. Jing and D. Raychaudhuri, ?Spectrum co-existence of ieee 802.11b and 802.16a
networks using reactive and proactive etiquette policies,? Mob. Netw. Appl., vol. 11,
no. 4, pp. 539?554, 2006.
[16] L. Berlemann, C. Hoymann, G. Hiertz, and B. Walke, ?Unlicensed operation of ieee
802.16: Coexistence with 802.11(a) in shared frequency bands,? in Personal, Indoor
and Mobile Radio Communications, 2006 IEEE 17th International Symposium on,
Sept. 2006, pp. 1?5.
[17] J. Nie, X. He, Z. Zhou, and C. Zhao, ?Communication with bandwidth optimization
in ieee 802.16 and ieee 802.11 hybrid networks,? in Communications and Information
Technology, 2005. ISCIT 2005. IEEE International Symposium on, vol. 1, Oct. 2005,
pp. 27?30.
[18] L. Berlemann, C. Hoymann, G. Hiertz, and S. Mangold, ?Coexistence and interworking
of ieee 802.16 and ieee 802.11(e),? in Vehicular Technology Conference, 2006. VTC
2006-Spring. IEEE 63rd, vol. 1, May 2006, pp. 27?31.
[19] H.-T. Lin, Y.-Y. Lin, W.-R. Chang, and R.-S. Cheng, ?An integrated wimax/wifi
architecture with qos consistency over broadband wireless networks,? in Consumer
Communications and Networking Conference, 2009. CCNC 2009. 6th IEEE, Jan. 2009,
pp. 1?7.
[20] R. Fantacci and D. Tarchi, ?Bridging solutions for a heterogeneouswimax-wifi sce-
nario,? Journal Of Communications and Networks, vol. 8, no. 4, pp. 369?377, 2006.
[21] T. Ali-Yahiya, K. Sethom, and G. Pujolle, ?Seamless continuity of service across wlan
and wman networks: Challenges and performance evaluation,? in Broadband Conver-
gence Networks, 2007. BcN ?07. 2nd IEEE/IFIP International Workshop on, May
2007, pp. 1?12.
[22] S. R. Ib?anez, R. A. Santos, V. R. Licea, A. E. Block, and M. A. G. Ruiz, ?Hybrid
wifi-wimax network routing protocol,? in CERMA ?08: Proceedings of the 2008 Elec-
tronics, Robotics and Automotive Mechanics Conference. Washington, DC, USA:
IEEE Computer Society, 2008, pp. 87?92.
71
[23] D. Niyato and E. Hossain, ?A hierarchical model for bandwidth management and ad-
mission control in integrated ieee 802.16/802.11 wireless networks,? in Wireless Com-
munications and Networking Conference, 2007.WCNC 2007. IEEE, March 2007, pp.
3763?3767.
[24] ??, ?Wireless broadband access: Wimax and beyond - integration of wimax and wifi:
Optimal pricing for bandwidth sharing,? Communications Magazine, IEEE, vol. 45,
no. 5, pp. 140?146, May 2007.
[25] ??, ?Integration of ieee 802.11 wlans with ieee 802.16-based multihop infrastructure
mesh/relay networks: A game-theoretic approach to radio resource management,? Net-
work, IEEE, vol. 21, no. 3, pp. 6?14, May-June 2007.
[26] T. Yahiya, A.-L. Beylot, and G. Pujolle, ?Policy-based threshold for bandwidth reser-
vation in wimax and wifi wireless networks,? in Wireless and Mobile Communications,
2007. ICWMC ?07. Third International Conference on, March 2007, pp. 76?76.
72