Content uploaded by Babatunde Samuel Ogunleye
Author content
All content in this area was uploaded by Babatunde Samuel Ogunleye on Sep 25, 2015
Content may be subject to copyright.
1
A Novel Model for WiMAX Frequency Spectrum
Virtualization and Network Federation
Babatunde S. Ogunleye, Alexandru Murgu
Department Electrical Engineering, University of Cape Town, Private Bag X3, Rondenbosch
7701, South Africa
OGNBAB001@myuct.ac.za, alexandru.murgu@uct.ac.za
Abstract. Network virtualization is an emerging and trending subject and it is
currently defining research ideas in areas such as the future internet and wireless
communication systems. The notion of combining the concept of network
virtualization together with network federation which basically involves the
interconnection of independent network domains, possess the potential to provide a
far more enriching environment where network resources like spectrum will be
better harnessed and utilized. This paper focuses on spectrum virtualization
implemented on the Worldwide Interoperability for Microwave Access (WiMAX)
network where virtualized WiMAX networks exist in a federated arrangement for
the purpose of sharing spectrum resources. A novel entity known as the VS-
Hypervisor, which is responsible for virtualizing the WiMAX spectrum is fully
described theoretically and metrics such as time delay, request rejection probability
was used in expressing its basic behaviour.
Keywords: Virtualization, Virtual Spectrum, Virtual Network Federation, VS-
Hypervisor, WiMAX.
1 Introduction
The emergence of new generations of wireless and mobile technologies has increased
the demand for advanced telecommunication infrastructure coupled with the need for
radio frequency spectrum that has the capacity to support high-speed transmissions with
extremely huge data content. Radio spectrum is a finite resource and its demand is
constantly increasing, most especially, by mobile network operators and this constitutes a
global challenge [1]. To illustrate the increasing demand for radio spectrum, at the ITU
World Radiocommunication conference 2007 [2], members considered the expansion of
radio spectrum for 4G networks also known as IMT-Advanced. The expansion was made
© Institute for Computer Sciences, Social Informatics and Telecommunications Engineering 2015
R. Agüero et al. (Eds.): MONAMI 2014, LNICST 141, pp. 1–13, 2015.
DOI: 10.1007/978-3-319-16292-8_8
2
to cater for the spectra needs of the new emerging 4G technologies. Further expansions
was also made at the 2012 edition of the aforementioned conference for IMT-Advanced
and other telecommunication technologies.
The limited nature of radio spectrum accounts for it being very expensive especially the
licensed portions. Some of the factors that affect the high costs of radio spectrum
includes: a) Propagation range b) In-building penetration; and c) Capacity i.e. the
bandwidth available in the band. In the implementation of mobile technologies, low
frequencies (below 1GHz) have a longer propagation range and are more suited to
deployments having few base stations which obviously means lesser cost and this is more
viable for low density areas. As for the higher frequencies (between 2-3GHz and above)
they have shorter propagation ranges, higher capacities (bandwidth) and very effective for
high density regions with closely packed cells and high density of traffic [3]. Because of
the varied uses of these portions of the radio spectrum there is very high competition
amongst mobile network operators to secure them which accounts for its very expensive
nature. The number of base stations in a cellular or mobile network is a major factor that
affects its cost which ultimately affects the capital expenditure (CAPEX) and operating
cost (OPEX) of the network provider. These underlying issues of radio spectrum have
stimulated the need to develop methods and systems for optimizing radio spectrum and a
considerable amount of research in this area is being conducted.
One of the new and exciting approaches to future network design is network
virtualization. Network virtualization is a networking concept that enables the deployment
of customized services and resource management solutions in isolated slices (groups or
portions) on a shared physical network [4]. The idea of virtualization is not so new. It was
first introduced by International Business Machines (IBM) in the 1970s and the
technology provided a way of separating computer physical hardware and software
(operating systems & applications) by emulating hardware using a software program.
Essentially, it involves installing a software program (known as a hypervisor) on a
physical computer. This software or hypervisor, in-turn then installs files that define a
new virtual computer otherwise known as a Virtual Machine (VM) [5].
There are several approaches to virtualization currently used on computer systems and
they include: Bare Metal Virtualization, Hosted Virtualization etc. [4] [6]. These
approaches are being closely studied by researchers and attempts are being made to mirror
their application into network virtualization. One major stride in developing an
architecture for network virtualization was done by a European initiative known as the
4WARD Project. The 4WARD Project’s main objective was to develop an architectural
framework for network virtualization in a commercial setting. This project has since
ended in 2010 but ongoing feasibility test is being done on its architecture [7]. To fully
harness the benefits of network virtualization, the inclusion of the concept known as
Network Federation into its overall design and implementation is beginning to capture the
interest of many researchers.
WiMAX which is a leading 4G mobile telecommunication technology has been chosen
as case study in this paper, whereby a novel model for generating virtual WiMAX radio
spectrum is developed for WiMAX network virtualization and federation.
3
2 Related Work
Currently, quite a number of research works have gone into developing architectures
for wireless network virtualization with emphasis on WiMAX. Others have also looked at
the possibility of developing more flexible ways of spectrum access through spectrum
virtualization, while a few have considered spectrum virtualization and management from
a generalized perspective. Federation of virtualized networks has received minimal
attention so far.
2.1 Wireless Network Virtualization
Ravi Kokku et al [8] designed and implemented a network virtualization substrate (NVS)
for the effective virtualization of wireless resources in WiMAX networks. The NVS
system they designed aimed to strengthen the role of cellular networks in providing
enriched experience for users. The NVS system was also designed to facilitate the
provision of greater differentiation of services among content providers and mobile virtual
network operators (MVNO). MVNOs are wireless network providers that lease
the physical network from the mobile network operators (MNO) to provide
wireless services meeting specific and unique needs of customers.
Their virtualization design centered more on isolating service flows which they called
slices. A service flow in the context of WiMAX is a unidirectional flow of either uplink or
downlink where packets are sent with a particular set of quality of service (QoS)
parameters. In their research, no particular attention was made at looking specifically at
the virtualization of the air-interface (spectrum) of WiMAX; where within a physical
based station multiple virtual networks will exist with each having an allocation of a
specific slice of the available spectrum.
Guatam Bhanage et al [9] proposed a virtual basestation design that will house multiple
MVNOs to on a single physical basestation to share its spectrum resources. Their work
was basically a discussion of how to design the infrastructure for supporting a virtualized
WiMAX framework. They also elaborately discussed the various options available for
building such a substrate and the tradeoffs of each option. They did not go into details
about how this can be done especially with regards to virtualizing the spectrum as a
singular resource.
The Global Environment for Network Innovations (GENI) [10] is a suite of research
infrastructure sponsored by the National Science Foundation in the United States. GENI is
made of federated virtual laboratory of multiple testbeds used mainly for research
purposes. GENI started a WiMAX project and part of the scope was to develop an
open/virtualized WiMAX basestation with external control and data APIs. The project has
4
made significant milestones so far, but nothing concrete with regards to WiMAX
spectrum virtualization and network federation [11].
2.2 Spectrum Virtualization
Kun Tan et al [12] argued that spectrum programmability – which means the ability to
change spectrum properties of a signal to match an arbitrary frequency allocation, is an
independent property that can be separated from the general PHY modulation. They
proposed to add a new spectrum virtualization layer (SVL) which they called layer 0.5.
The propose that layer will support flexible spectrum programmability and enable flexible
spectrum access for general wireless networks. Their research is amongst the few that
practically described how radio spectrum could be virtualized and dynamically accessed.
Yasir Zaki et al [13] implemented their ideas of network virtualization using the
Long Term Evolution (LTE) as their case study. They narrowed their perspective of
network virtualization into two sections: a) virtualization of the physical nodes of the
network and 2) virtualization of the air-interface (spectrum) with the latter section being
the main focus of their research. The approach they used in the LTE air-interface
virtualization, involved using a hypervisor which they termed and called the LTE
Hypervisor. This LTE Hypervisor is responsible for virtualizing the enhanced Node B
(eNodeB) i.e. the LTE basestation into a number of virtual eNodeBs. The LTE hypervisor
is also responsible for scheduling the air-interface resources between the virtual eNodeBs.
They stated that there are already solutions for building virtualized base stations
identifying the VANI MultiRAN solution which supports multiple virtual base stations all
running a single physical infrastructure.
Similar to [13], this research paper centers more on the air-interface virtualization of
WiMAX, especially for WiMAX networks existing in a federated arrangement.
3 Motivation for WiMAX Spectrum Virtualization
Wireless broadband technologies are growing currently at a very high rate, having a major
influence on how people communicate. This has resulted in an inexhaustible need for
radio spectrum and ultimately bandwidth by network operators and WiMAX is not left out
in this struggle. According to recent WiMAX market report analysis, the global WiMAX
equipment market is expected to grow from $1.92 billion in 2011 to $9.21 billion in 2016,
while the service market is expected to grow form $4.65 billion in 2011 to $33.65 billion
in 2016 [14]. Reports like this further stress the need and importance to efficiently
optimize and utilize scare wireless resources such as spectrum which can be adequately
achieved through spectrum virtualization.
5
Spectrum virtualization for WiMAX will greatly reduce the number of base stations
needed for deployment and overall energy usage. With WiMAX virtual networks
operating in a federated arrangement, it will further enhance the sharing of network
resources and exponentially improve the efficient use of spectrum. This will in the long-
run attract smaller network providers to come into the market to provide better services,
hence creating a richer experience for end-users.
4 Network Federation Concept
The concept of network federation is a new and exciting idea that it is being currently
proposed by many researchers as part of the building blocks for the future internet.
Federation in the network domain is a model for establishing very large scale and diverse
infrastructure for the purpose of interconnecting independent network domains in order to
create a rich environment with increased benefits to users of the independent domains
[15]. This concept can easily be expressed as shown in Fig. 1 where independent
Network providers (NP) existing in an interconnected framework with each of them
having their own Network Operators (NO). The NOs are dependent on the NPs for the use
of their network infrastructure and the federated network setting allows for resource
sharing amongst all the various NPs.
Fig. 1. Illustration of Network Federation
6
5 Overview of WiMAX PHY Layer
WiMAX is defined under the IEEE 802 family as IEEE 802.16. For the IEEE 802.16, the
physical layer was defined for a wide range of frequencies from 2-66 GHz. Sub
frequency range of 10-66 GHz is essentially for line-of-sight (LoS) propagation, whereas
for 2-11GHz bands, communication can be achieved for licensed and un-licensed bands
and they are also used for non-line-of-sight (NLoS) communication.
WiMAX uses a number of legacy technologies amongst which are: Orthogonal
Frequency Division Multiplexing (OFDM), Time Division Duplexing (TDD) and
Frequency Division Duplexing (FDD). OFDM is a multiplexing technique that subdivides
the bandwidth of a signal into multiple frequency sub-carriers. These multiple frequency
subcarriers are modulated with data sub-streams and then transmitted. OFDM modulation
is realized with efficient Inverse Fast Fourier Transform (IFFT) that generates large
number of subcarriers (up to 2048) with minimal complexity.
In an OFDM system, resources (e.g. spectrum) are available in the time domain by
means of OFDM symbols and in the frequency domain by means of subcarriers. These
resources in the time and frequency domain can be rearranged into sub-channels for
allocation to individual users. An offshoot of OFDM is the Orthogonal Frequency
Division Multiple Access (OFDMA) which is a multiple access/multiplexing scheme that
provides multiplexing operation of data streams from multiple users both on the downlink
and uplink sub-channels. The OFDMA subcarriers are shown in Fig. 2. WiMAX IEEE
802.16e-2005 otherwise known as mobile WiMAX is based on scalable OFDMA (S-
OFDMA) which supports a wide range of bandwidths enabling the need for flexible
various spectrum allocations. The scalability is achieved by adjusting the Fast Fourier
Transform (FFT) size and the same time fixing the sub-carrier frequency at 10.94 kHz. S-
OFDMA supports bandwidth ranging from 1.25MHz to 20MHz [18].
Fig. 2 WiMAX S-OFDMA Subcarriers [17]
7
6 WiMAX Spectrum Virtualization
The process of virtualizing the WiMAX radio spectrum or air-interface will require that
the basestation hardware components must also be virtualized. As similarly proposed by
[13] for LTE eNodeB virtualization which follows the principle of node virtualization
already done in the field of computer systems we propose a near similar architecture for
the virtualization of WiMAX basestation with emphasis on spectrum virtualization. As
previously discussed, the general approach for computer system virtualization involves
the use of a hypervisor. In similar fashion, our proposed model for WiMAX basestation
visualization is shown in Fig. 3. Considering that our emphasis is solely on spectrum, the
generic hypervisor has been renamed Virtual Spectrum Hypervisor (VS-Hypervisor).
6.1 VS-Hypervisor
The VS-Hypervisor is the entity responsible for virtualizing the air interface and ensuring
proper management of spectrum allocation. Its primary job is the scheduling of spectrum
resources amongst the virtual networks to meet their individual bandwidth requirements.
WiMAX resources in the frequency domain are represented as S-OFDMA subcarriers and
the numbers of subcarriers are directly proportional to the bandwidth size. The VS-
Hypervisor works by receiving bandwidth estimates from the individual virtual WiMAX
networks done by a bandwidth estimation unit. The estimated bandwidth values are then
mapped unto the appropriate number of S-OFDMA subcarriers or sub-channels (grouped
sub-carriers) and a scheduling algorithm schedules these subcarriers/spectrum resources to
Fig. 3. Approach for WiMAX Basestation
Virtualization
8
the appropriate virtual networks. The amount of spectrum allocated will be based on
contracts or strictly on need and request. Fig. 4 depicts the VS-Hypervisor having access
to the entire spectrum channel bandwidth of a physical WiMAX basestation while it
schedules the spectrum to the virtual networks.
In this paper we focus more on the basic description of the VS-Hypervisor and its
functionality. We also look at this functionality in terms of handling requests from VNOs
using metrics such as time delay and request rejection rate.
6.2 VS-Hypervisor Operational Characterization
Considering that the VS-Hypervisor will be constantly receiving spectrum allocation
requests from VNOs, it is very necessary to analyze its operational characteristics one of
which is; how it long it takes (time delay) for a request to be processed. Theoretically, lets
denote the time delay as Ti, assuming that Di(t) is the demand or request for bandwidth
(BW) allocation or spectrum allocation issued by VNOi at time t arriving instantaneously
at the VS-Hypervisor. Let us apportion Ri for rates (bits/sec) at which request are issued
by VNOi.
Fig. 4. WiMAX Spectrum Virtualization Using VS-hypervisor
9
The response of the VS-Hypervisor after a demand can be denoted as Yi(t) which can
also be assumed to be instantaneously received by the VNOi; meaning that no delays
occurs between the input and output of the VS-Hypervisor. These are reasonable
assumptions for an ideal situation but in real terms, the travel times of the
request/response signals could be represented in the definition of Yi(t - Ti). Due to the
various system modeling assumptions, the appropriate block diagram describing the VS-
Hypervisor functionality in terms of time delay is shown in Fig. 5 below.
The signal definition Di(t) can be represented as seen in Fig. 6 below:
Fig. 6. VS-Hypervisor Input Information
Where is the time stamp at which the request Di is issued,
is the amount of bandwidth that is requested or demanded by the VNOi at time ti,
is the current priority/rank of VNOi in terms of bandwidth allocation and
is the
antecedent priority/rank of VNOi with regards to BW allocation in the previous allocation
cycle of the VS-Hypervisor.
Signal Yi(t) has the following structure is expressed in Fig. 7 below:
Fig. 7. VS-Hypervisor Output Information
Where is the time stamp at which the allocation Yi is issued, is amount of
bandwidth that is allocated by the VS-Hypervisor to VNOi, is the priority/rank of
VNOi in the next cycle of spectrum allocation by the VS-Hypervisor, is the frequency
Fig. 5. VS-Hypervisor Time Delay Block Diagram
10
plan (that is, indexes of frequency channels to be used in the next allocation cycle to
implement ), is the time plan (that is, indexes of time slots to be used in the next
allocation cycle to implement , is the start time of the next allocation cycle of
BW (spectrum) addressing the VNOi demand currently processed and is the end time
of the next allocation cycle of BW addressing the current VNOi demand. Formally the
VS-Hypervisor time flow can be expressed as shown in Fig. 8.
Alternatively it can be expressed as described in equation (1) below:
(1)
Where is the VS-Hypervisor mapping function corresponding to VNOi which can
be represented as shown in Fig. 9 below:
The specific definition for, , , …, are part of the VS-Hypervisor design
process to be implemented on the spectrum virtualization. Their definitions will form part
of the future work of this paper.
The available bandwidth for the VS-Hypervisor to allocate can be expressed as:
Fig.9. VS-Hypervisor Input/Output Map
Yi (t - Ti)
Di ()
- Ti
(Ti)
VS-Hypervisor
Fig. 8. VS-Hypervisor Time Flow Structure
11
(2)
can be defined as the anticipated BW needed at the moment of request i.e.
. Fundamentally, the map function will be controlled by a state machine
containing external events D(t), having an internal state evolution subject to system
capability functionality constraints which can be expressed as:
(3)
(That is the sum of BW demanded by the VNOs cannot exceed system capacity C)
(4)
Where T is the overall duty cycle of the WiMAX system.
6.3 VS-Hypervisor Workload Control
The ability the VS-Hypervisor to control its workload is a very important issue that has to
be considered knowing that not all requests made to the VS-hypervisor will be processed.
There is the possibility that some may be rejected because there is not enough bandwidth
(spectrum) to be allocated. Using classic telephony congestion probability formula like the
Engset loss formula, we can be able to measure the probability at which a request made to
the VS-Hypervisor will be rejected [16]. Consider the Engset equation (7) below:
(5)
Where M, N are the input requests and granted requests by the VS-Hypervisor
respectively, is the ratio between rate of requests ( are made by a VNO to the rate at
which a request granted), i.e. . The probability is expressed as the ratio of the
lost stream of requests into the VS-Hypervisor and to the requests that where granted.
Where the state i is the state of the VS-Hypervisor when it has the capacity to allocate
spectrum and state N is the VS-Hypervisor reaches it maximum capacity. Where PN is the
probability that a request will be granted and Pi is the probability of the state at which the
VS-Hypervisor can grant requests.
12
8 Federation of Virtualized WiMAX Network
The establishment of a virtualized and federated mobile network will essentially consist of
the following key participants:
Virtual Network Provider (VNP). The VNP is the owner of the network physical
infrastructure that has the ability to provide the existence of virtualized mobile networks.
Virtual Network Operator (VNO). The VNO is a mobile operator that operates a leased
virtual network infrastructure and in turn hosts their network services on these virtual
mobile infrastructure provided by a VNP.
Diagrammatically, the federation of a virtualized WiMAX network can be modelled
using the Fig.10. The figure shows three virtual network operators (VNO1, VNO2 &
VN03) that are separately hosted by four virtual network providers (VNP1, VNP2, VNP3 &
VNP4). The allocation and management of resources (spectrum) in this federated setup is
coordinated by the VS-Hypervisor.
The aggregation of all demands of the individual VNOs can be expressed as:
(6)
Where the demand D is the transpose of the vectors D1, D2 and D3. The capacity C is the
aggregate capacity in terms of spectrum for all the network providers. It can also be
expressed as: (7)
Fig. 10. Virtualized Federated Mobile Network Implemented
Using the VS-Hypervisor
13
The total capacity C of the entire federated network is represented as the transpose of the
vectors C1, C2, C3 and C4. The VS-Hypervisor’s cumulative magnitude in terms of the
amount of spectrum requests/demand it can handle can be represented with the vector Ƶ
which is defined as:
(8)
Where Ƶ1, Ƶ2, Ƶ3 and Ƶ4 are the basis vectors which are linearly independent vectors
representing the VS-Hypervisor for the individual VNOs as shown in Fig. 10. The real
numbers are the coefficients of Ƶ.
(9)
Irrespective of how the federated networks are designed, the total number of demand
made by the VNOs must never be more that the available bandwidth B. This is
represented as a dot product between Ƶ and demand D as show in equation 10 below:
(10)
To assess the efficiency of the VS-Hypervisor within the federated network, a
mathematical expression showing the ratio between the total bandwidth and capacity of
the federated network can be described as show in the equation 11 below.
(11)
In summary, these equations in general try to express the basic behavior of the VS-
Hypervisor in a federated network setting describing how the VS-Hypervisor should be
able reliably allocate virtual spectrum for VNOs within a VNP and enable spectrum
sharing between VNPs.
8 Discussions and Future Work
In our research so far, we have been able to provide a basic view about the concept of
network virtualization existing within a federated arrangement using WiMAX as our case
study. Our emphasis on WiMAX spectrum virtualization was to develop a system and
framework where WiMAX spectrum can be fully utilized and better harnessed for current
network operators and to encourage the entry of new players into the WiMAX market. We
described the workings of the VS-Hypervisor which is an innovative system tailored for
the virtualization of the WiMAX air-interface.
14
This work is still at its infancy stage and a lot of issues are yet to be addressed. The
issues waiting to be resolved includes: The bandwidth estimation algorithm needed for
proper fair allocation of the virtual spectrum, developing a scheduling algorithm for the
VS-hypervisor on how spectrum resources will be allocated to the virtual networks either
based on contracts or service level agreements (SLAs) and running tests simulations to
evaluate the overall performance of the entire federated virtualized WiMAX network.
References
1. R. MacKenzie, K. Briggs, P. Gronsund and P. Lehne:Spectrum micro-trading for mobile
operators:Wireless Communications, IEEE , vol. 20, no. 6, pp. 6-13, (2013)
2. International Telecommunication Union: Final Acts WRC-07: World Radiocommunication
Conference, Geneva, (2007)
3. R. Martyn, "Competition, Spectrum for Mobile Broadband in the Americas: Policy Issues for
Growth and," GSMA, Atlanta, 2011.
4. C. Jorge and J. Javier:Network Virtualization – a View from the Bottom: MobiCom, Chicago,
(2010)
5. B. David: Virtualization - Is it Right for You?: LAD Enterprizes, Inc. (2008)
6. N. Susanta and C. Tzi-cker:A Survey on Virtualization Technologies:SUNY, Stony Brook, NY,
(2005)
7. The FP7 4WARD Project, The FP7 4WARD Project: The Project, http://www.4ward-
project.eu/index.php?s=overview&c=project.
8. K. Ravi, M. Rajesh, Z. Honghai and R. Sampath, :NVS: A Virtualization Substrate for WiMAX
Networks: MobiCom, Chicago (2010)
9. B. Gautam, S. Ivan and R. Dipankar: A Virtualization Architecture For Mobile WiMAX
Networks:ACM SIGMOBILE Mobile Computing and Communications Review (2011)
10. GENI, http://www.geni.net
11. GENI, WiMAX-GENI:geni, http://groups.geni.net/geni/wiki/WiMAX.
12. T. Kun, S. Haichen, Z. Jiansong and Z. Yongguang:Enable Flexible Spectrum Access with
Spectrum Virtualization: IEEE DySpan, Washington (2012)
13. Z. Yasir, Z. Liang, G. Carmelita and T.-G. Andreas: LTE Wireless Virtualization and Spectrum
Management: Mobile Networks and Applications, vol. 16, no. 4, pp. 424-432 (2011)
14. Markets and Markets: WiMAX Market - TDD/FDD Spectrum Analysis and Global Forecast
(2011-2016), http://www.marketsandmarkets.com/Market-Reports/wimax-market-335.html.
15. Technopedia: What is Federation?, http://www.techopedia.com/definition/2500/federation.
16. Ran Pang: Engset Distribution: Department of Electrical and Computer Engineering North
Carolina State University: (2012)
17. Orthogonal Frequency Division Multiple Access,
http://en.wikipedia.org/wiki/Orthogonal_frequency-division_multiple_access
18. WiMAX Forum: Mobile WiMAX – Part 1: A Technical Overview and Performance Evaluation:
(2006)
