Financial Assessment of Citywide Wi-Fi / WiMAX Deployment

advertisement
Financial Assessment of Citywide
Wi-Fi / WiMAX Deployment
Vinoth GUNASEKARAN & Fotios HARMANTZIS (*)
Stevens Institute of Technology, USA
Abstract: There are several ways by which a Wireless Internet Service Provider (WISP)
can deliver city wide wireless broadband services. However, it is necessary to determine a
profitable business case and at the same time a cost-effective service model, which is
affordable to all types of users and different classes of society. This paper proposes a
service model (both data and voice) that uses two emerging wireless technologies (Wi-Fi
and WiMAX) to deliver cost-effective broadband services. Wi-Fi / WiMAX have not only
the potential to compete on a cost-per-megabyte basis with cable and Digital Subscriber
Line (DSL), but also make ubiquitous broadband a reality. If engineering and economics
are correctly applied, a Wi-Fi network can be built around an entire city with a WiMAX
backhaul, instead of providing limited coverage for hotspots. On the other hand, internet
telephony over a Wi-Fi network is the public access version of Voice over Internet Protocol
(VoIP): users can enjoy the handiness of a cell phone-like service, while avoiding the cost
of traditional cellular carriers. This paper focuses on the techno-economic modelling of WiFi hot zones into a WiMAX infrastructure mesh while addressing Voice over Wi-Fi (VoWiFi) issues. Our study demonstrates that low-cost broadband services can be offered, while
remaining economically advantageous for service providers
Key words: Wi-Fi, WiMAX, VoWi-Fi, 2.5G/3G
T
he Wi-Fi service industry is undergoing a fundamental shift towards
ubiquitous Wi-Fi, with the onset of citywide Wi-Fi deployment. The
deployment of municipal broadband networks has also been
increasing in the United States 1, and the market for embedded Wi-Fi chips,
including laptops, PDAs and mobile phones is growing. A market study 2
(*) Acknowledgements: Many thanks to Dr. Audrey Curtis, Director of Telecommunications &
Project Management, and Dr. Kevin Ryan, Associate Professor of Telecommunications
Management, both at Stevens Institute of Technology, for their countless discussions and
helpful suggestions regarding this research. We also thank Dr. N. K. Shankaranarayan and Dr.
Byoung Jo J. Kim, both with AT&T Labs, NJ, USA, for their valuable comments the helped us to
better understanding the economics and technical aspects of infrastructure mesh topologies.
1 http://www.muniwireless.com/reports/docs/June2004Report.pdf (URL accessed on February
2005).
COMMUNICATIONS & STRATEGIES, no. 63, 3rd quarter 2006, p. 131.
132
No. 63, 3rd Q. 2006
shows that in 2005, the annual shipment of Wi-Fi chipsets exceeded 100
million units; while over 90% of the laptop computers sold were Wi-Fi
enabled. The number of Wi-Fi users in the United States is expected to
almost equal or surpass 2.5G/3G data users in the future 3. Millions of
people in developed countries already own Wi-Fi client devices to access
their existing home Wi-Fi networks or office wireless networks. Taking all
these factors into account, a WISP can strategically plan to deploy a city
wide Wi-Fi network without spending on the client device or incurring
customer acquisition costs.
WiMAX, another emerging wireless technology, is defined as Worldwide
Interoperability for Microwave Access, by the WiMAX Forum 4. IEEE
standard 802.16 is the foundation of WMAN (wireless metropolitan area
network) of the next few decades (EKLUND et al., 2002). The forum was
formed to promote WMAN broadband technology and to support vendor
interoperability. It also aims to promote the conformance and interoperability
of the IEEE standard 802.16. WiMAX has received broad industry support
both from equipment makers and service providers as a means of
broadband wireless access. Mobile WiMAX penetration is expected to be
widespread between 2009 and 2012, when most wireless carriers accept
WiMAX as a mobile wireless broadband service alternative to other mobile
technologies. It may take a few more years for the technology to mature and
compete with other wireless technologies, and also for WiMAX PC cards to
be embedded in end user devices as currently seen with Wi-Fi cards.
WISPs can also take advantage of delivering voice to end users, as
VoWi-Fi is now becoming a reality. WISPs can maximize their revenue by
optimizing their network resources, providing both voice and data services.
Wi-Fi phones are the next generation intelligent IP communications devices.
Two protocols are currently being used: the H.323 and SIP (Session
Initiation Protocol). The phones add SIP/H.323 based VoIP communications
together with Wi-Fi installations. These phones can be used in any Wi-Fi
network. There are "soft" phones that can be downloaded into a PDA or
laptop with additional software, and turn into wireless speakerphones when
2 http://nwc.networkingpipeline.com/ Wi-Fi Chip sale to hit 120 million, Market study by InStat/MDR & Wi-Fi Alliance (URL accessed on July 2006).
3 http://www.unstrung.com/ Wi-Fi to surpass 3G,By Pyramid Report (URL accessed on July
2006).
4 http://www.WiMAXforum.org (URL accessed on May 2005).
V. GUNASEKARAN & F. HARMANTZIS
133
connected to Wi-Fi networks 5. A soft phone is software that simulates a real
phone and runs on a general purpose computer, rather than a dedicated
device. Soft phones are typically part of VoIP environments and can be
standard-based SIP/H.323 or proprietary. There are also Wi-Fi hard phones
based on both H.323 and SIP protocols. A Wi-Fi phone is a hard phone with
a built-in Wi-Fi transceiver unit to connect to a Wi-Fi access point (AP),
instead of an Ethernet port. It does not require a personal computer or any
software to be run on a personal computer to make and receive VoIP phone
calls.
Internet telephony technology allows phone calls to be made over
broadband internet access (both wired and wireless). In the near future,
more people will be using internet telephony as the market migrates from the
traditional public switch telephone network (PSTN) towards VoIP. As a
market report indicates, the customer base for broadband VoIP services is
gradually growing both in North America and in major European countries,
such as France, Germany, Italy, Spain, Sweden and the UK. In Europe,
VoIP has already penetrated nearly one-fifth of all businesses with Spain (24
percent) and the United Kingdom (22 percent) at the forefront 6.
On the enterprise level, many companies have already established Wi-Fi
networks and integration of VoWi-Fi is anticipated. A market study 7 also
indicates that VoWi-Fi handsets will represent about 7% of all handsets by
2009. Therefore, it can be inferred that a home or office Wi-Fi network forms
a convenient platform for internet telephony using Wi-Fi enabled phones.
The customers' broadband wireless phones (VoWi-Fi hard or soft) become a
substitute for mobile phones when they are on the move. By taking all these
parameters into consideration, Internet Telephony over public Wi-Fi hotspots
threatens to remove a considerable amount of traffic from cellular networks.
There is already fierce competition between Wi-Fi and cellular networks for
data services, and if internet telephony is made available in all public Wi-Fi
hotspots this should constitute a serious threat to cellular operators.
Currently, hotspot utilization is low and a lot of capacity is left unused. There
is consequently capacity available for increasing hotspot usage and
providing internet telephony in the hotspot. This increase in the utilization
5 http://www.telesym.com/ (URL accessed on Jan. 2005).
6 The Global Information Technology Report 2003-2004, Oxford University Press 2004.
7 http://www.tmcnet.com/ "Enterprise To drive Dual Mode Cellular / VoWi-Fi Handsets", by ABI
Research (URL accessed on May 2006).
134
No. 63, 3rd Q. 2006
rate due to voice services along with data should have a significant impact
on hotspots' regular business model.
Broadly speaking, there are two types of Wi-Fi deployment as shown by
(CAMPONOVO et al., 2003): selected locations or hotspots and extensive or
outdoor coverage. Recently, some cities around the world have started
actively engaged in deploying extensive outdoor city wide Wi-Fi networks in
many different ways. As explained by BAR et al. (2006), there are nine
possible options in deploying a city wide Wi-Fi network as no model fits all
cities, and each option leads to a unique set of business models and policy
issues.
This paper explores a single private owner model for operating the Wi-Fi
service and selling it directly to customers. In this model, local government
makes an agreement with a single private company to build and own the
networks. Local government can also make a deal with private owners by
granting a license to operate; making the WISP offer its inhabitants the
service for a reduced fee (the monthly service fee should be far less than the
traditional broadband service, i.e., cable and DSL). The service provider in
turn may gain access to the city’s "urban furniture" (i.e., lamp posts or street
poles) for reduced leasing costs to mount its antennas and equipment
boxes. Extensive research has already been carried out in the area of Wi-Fi
integration with cellular networks (SALKINTZIS et al., 2002). This paper
focuses on how Wi-Fi will integrate with WiMAX to offer ubiquitous access,
playing a key role in the emerging fourth generation wireless networks.
Infrastructure cost models have previously been applied in investment
analysis for various types of wireless access provisioning (FURUSKAR et
al., 2005). However, this is the first time a cost model is applied in
investment analysis for the use of both Wi-Fi and WiMAX. In the cost model
presented here, Wi-Fi is used to reach the end user and WiMAX provides a
backhaul to offer a cost effective broadband service. Pursuing the model,
this paper is organized as follows. In the next section background
information is given about Wi-Fi and WiMAX technologies and how they
complement each other. The third section presents a service architecture
overview of Wi-Fi / WiMAX and two deployment scenarios for offering both
voice and data. The engineering and economics model, followed by the cost
model makes up the forth section, while the fifth section examines a citywide Wi-Fi case. The results of the analysis are presented in this section
along with some sensitivity analysis, which is followed by our conclusions.
V. GUNASEKARAN & F. HARMANTZIS
135
Background of emerging wireless technologies
Wi-Fi versus WiMAX
Wi-Fi currently has a clear advantage over WiMAX since it is already
available in most end user devices. The customer acquisition cost for Wi-Fi
service is consequently much lower compared to other broadband
technologies. Although WiMAX as the last mile alternative is a viable option,
the customer acquisition cost increases as it is highly dominated by the
customer premise equipment (CPE) cost. Clearly, WiMAX technology today
is following in the same footsteps as Wi-Fi a few years ago. The
standardization and interoperability between different vendor products may
lead to higher output levels, which will result in very low equipment prices in
the future. When Wi-Fi systems based on 802.11 protocols were first
developed, interoperability was of paramount importance. As a result, any
Wi-Fi product can easily communicate with other Wi-Fi products. Another
advantage of Wi-Fi is that large scale service-level roaming between
different WISPs is possible, as Wi-Fi certification has become a de facto
standard for IEEE 802.11b based products (HENRY et al., 2002). It is also
expected that, at some stage, WiMAX will also accomplish price and
performance levels similar to Wi-Fi. Until the mobile version of WiMAX (i.e.,
IEEE 802.16e) becomes a reality, both Wi-Fi and WiMAX technologies can
coexist, addressing different tasks. At present, a WISP can leverage the
most mature technology, namely Wi-Fi, to reach the end user; at the same
time, it can take advantage of WiMAX to minimize backhaul cost and
efficiently reduce the time for service provisioning. If properly planned and
deployed, Wi-Fi with WiMAX can turn the whole region within the geographic
boundaries into a "hot zone".
Why infrastructure WiMAX mesh
In the proposed model, WiMAX is used as backhaul to feed the Wi-Fi
APs. This is because the rental of wired backhaul networks constitutes a
major cash outflow (BJORKDAHL et al., 2004). Thus, to reduce backhaul
cost and achieve efficient use of the wired backhaul, infrastructure mesh can
be used. Aggregating backhaul lines into higher capacity lines is not only
cheaper, but also reduces the physical space compared to smaller speed
circuits. Unlike other mesh networks, the infrastructure WiMAX mesh
136
No. 63, 3rd Q. 2006
network is slightly different. It is a type of mesh which is unlike "ad-hoc" or
"client" mesh. In client meshing, every client device can pass along traffic for
other devices; therefore, any client can hop through other neighbouring
devices or routing nodes to reach other clients in the network 8. The IEEE
802.16 options are the PMP MAC (Point to Multi Point Medium Access
Control) and the mesh MAC. The PMP MAC option is the default
architecture, which is supported and enhanced by the WiMAX Forum. The
current mesh mode in standard IEEE802.16 -2004 is not compatible with the
PMP mode and at the same time it has no relay function. A new task group,
namely the Mobile Multihop Relay (MMR) study group (IEEE802.16j), is
actively working on the mesh/relay mode 9.
Advantages of Wi-Fi in the last mile
Wi-Fi technology options: Wi-Fi chip makers have already announced a
tri-mode chip, with IEEE802.11b/g and 802.11a as their flagship product 10.
WISPs can strategically plan to deploy their APs to support as many
technologies and standards as possible. This would allow the client's
software to "sniff" and select the best technology available at any given spot.
Another advantage of having different technology options in a Wi-Fi mesh
might be the use of one technology for mesh node communications and
other technologies for client to node communications. 802.11b, the most
widely known, supports a smaller number of audio streams when compared
to the high performance standard 802.11a or 802.11g. Nevertheless,
802.11a with eight channels can be a technology of choice for voice
applications, making it an attractive alternative to 802.11g, which has only
three non-overlapping channels. Wi-Fi service providers may consider
installing APs that include both 802.11a for voice users and 802.11b for data
users. This is a business decision; it may or may not be economical. A
crucial parameter for the justification of such a decision is the amount of
traffic at specific hotspot locations.
8 http://www.techworld.com/ Client Mesh vs. Infrastructure Mesh, (URL accessed on April
2006).
9 http://www.ieee802.org/16/sg/mmr/index.html: Chair of IEEE802.16j Task group- Mitsuo
Nohara , KDDI corp , (URL accessed on May 2006). This task group considers multi hop relay
capabilities and functionalities of interoperable relay stations (RS) and base stations. It also
considers 2-hop MMR networks as mandatory.
10 http://www.pcworld.com/ Intel Eyes Tri-mode Wi-Fi (URL accessed on July 2005).
V. GUNASEKARAN & F. HARMANTZIS
137
Wi-Fi service options: customers have many different options in terms of
bandwidth upgrade. There are also many options for data and voice services
at a lower cost. With any portable devices that have Wi-Fi clients, people
can connect to the network while roaming in different metro zones
throughout the city. Anyone who has a Wi-Fi client device in the city can
access the network from any location by logging on to the network.
Traditionally, cable and DSL networks offer a simple service model with flat
pricing: a monthly subscription fee. However, Wi-Fi service providers can
supply more innovative service options offering on demand service plans for
both data and voice users. There are several payment options, such as a
subscription fee on a monthly basis, a one time charge, for example, per
connection charge, or usage-based pricing 11.
Architectural overview of Wi-Fi with WiMAX
A model is proposed incorporating WiMAX mesh with Wi-Fi systems for
two scenarios where there is a mix of multi-dwelling and individual houses.
The first type of architecture is the multi dwelling unit in a dense urban area
where there are many subscribers per square mile; the second type is the
low-density area, with individual buildings and houses where Wi-Fi / WiMAX
serves as the last mile.
Type 1: Wi-Fi / WiMAX serving a multi dwelling unit
In this type of model, the majority of small offices, home offices (SOHO),
and households are in multi-dwelling units or apartment complexes. As seen
in figure 1a, WiMAX can co-exist with Wi-Fi to deliver megabits of data to
apartments or office buildings. From there, Wi-Fi can be used to distribute
services to individual houses, office rooms, lobbies, conference room, etc.
Though the WiMAX standard does not describe how much capacity an
operator can feed each Wi-Fi AP, a single WiMAX base station (BS) could
handle hundreds of megabits per second of data and can feed one or more
Wi-Fi APs mounted on tall buildings.
11 www.boingo.com (URL accessed on Nov. 2004).
No. 63, 3rd Q. 2006
138
Figure 1a - Wi-Fi / WiMAX serving multi-dwelling unit
Figure 1b - Wi-Fi / WiMAX serving independent houses/SOHO's
Type 2: Wi-Fi / WiMAX serving independent houses
In the second type of model, individual buildings and houses may be
packed close to each other or scattered. In this scenario, a large number of
Wi-Fi APs are needed to cover the entire region. The leasing cost can also
be significantly reduced, as the APs can use the lamppost or the rooftop of a
residential building, reducing operating expenditures (OpEx) significantly.12
There is a provision of using APs with higher gain antennas to extend the
coverage, while still limiting the maximum effective isotropic radiated power
(EIRP) within the legal limit, as described by the Federal Communications
Commission (FCC). The FCC, like the European Telecommunications
12 www.tropos.com (URL accessed on May 2005).
V. GUNASEKARAN & F. HARMANTZIS
139
Standard Institute (ETSI) in Europe, has defined the emission characteristics
for all unlicensed spectrum, including the transmission power and the
allowable antenna gain in Wi-Fi devices. The Wi-Fi in an outdoor
environment needs adequate power level in both the client device and the
base station antennas. Since there is a power limit to the Wi-Fi client device,
the uplink coverage can be obtained using high gain antennas at the APs.
For a given antenna gain there is also a limit in the maximum transmit power
that dictates the downlink coverage. The normal Wi-Fi AP (802.11 b or
802.11g) only covers 300 feet, which is roughly 0.0102 square miles. For
outdoor Wi-Fi access coverage can be increased by using higher gain
antennas. It is also feasible to extend coverage further by bearing additional
cost on smart antennas or phased array antennas. Such antennas based
Wi-Fi systems offer cost effective coverage and access in a ubiquitous
environment. Phased array antennas give immunity to interference, have a
greater coverage range and lower the overall deployment cost, compared to
traditional omni antennas.
A techno-economic model for Wi-Fi / WiMAX
Infrastructure WiMAX cell layout
The WiMAX main BS with wired backhaul should be at the center of the
WiMAX mesh BSs. Wi-Fi cells are then embedded in both WiMAX main or
mesh cells. For example, if one considers a cluster size of nine square cells
(the basic mesh architecture with one hop), there will be one WiMAX main
BS surrounded by eight mesh BS as shown in figure 2. Each WiMAX mesh
and main BS of one square mile area can include as many Wi-Fi cells as
possible, as long as the BS has enough capacity to aggregate all the APs
traffic. In this architecture, each WiMAX mesh BS aggregates all the traffic
from the Wi-Fi APs and then wirelessly backhauls to WiMAX main BS. From
there, it is taken to the wire backhaul and finally to the point-of-presence
(POP). For example, considering this architecture for a coverage area of 135
square miles, with each WiMAX cell having a one mile radius, there will be
seven WiMAX clusters with 56 WiMAX mesh BS and seven main BSs. Since
there are only seven clusters, there is a need for only seven WiMAX wired
backhaul facilities in the main BS for the entire 135 square miles area, to
serve all the Wi-Fi APs.
No. 63, 3rd Q. 2006
140
Figure 2 - One WiMAX cluster with mesh and main BS
WiMAX mesh and main BS radio capacity planning
For radio capacity expansion, sectorization are adopted in WiMAX main
and mesh BS, instead of cell splitting. The techno-economic analysis of pure
WiMAX deployment is performed by GUNASEKARAN & HARMANTZIS
(2005). However, this model extends that study to incorporate Wi-Fi in the
last mile, while keeping a WiMAX mesh backbone architecture. The model
assumes that each WiMAX BS is a micro cell with a radius of 1 mile. In this
type of cell layout, the BS towers are typically around 30ft-50ft below rooftop.
Operators can use both licensed and unlicensed spectrum. The WISP has
the advantage of using unlicensed spectrum, and is able to deploy services
immediately.
In cases where multiple operators are operating, licensed spectrum is an
appropriate choice for avoiding interference. Though many wireless vendors
offer flexible channel size, a standard size of six MHz is used for the
calculations. With a channel size of six MHz, and spectral efficiency of three
bits/Hz, the capacity per sector for each WiMAX main or mesh BS will be 18
Mbps. These are dimensioned so as to produce a maximum of only six
sectors, i.e., only 60-degrees configuration.
The model implements two, four and six sectors with only two frequency
channels. For more capacity, one can add more channels and still keep six
sectors with 60-degree configurations. An effective sector is defined as the
V. GUNASEKARAN & F. HARMANTZIS
141
total number of sectors available due to addition of more channels in the
actual given sector. The effective sectors and channel numbers selected are
even, so that the same channels can be used in alternate sectors to avoid
co-channel interference.
Wi-Fi cell layout in the last mile
The Wi-Fi cells are optimized to fit in the WiMAX main or mesh cells. The
coverage provided depends upon the surroundings and can be affected by
buildings, hills, foliage and weather conditions. A typical installation supports
Wi-Fi cell spacing of one-fifth to one-fourth of a mile, leading to typical
densities of 15 – 25 Wi-Fi cells embedded in each square mile area. But with
initial capital expenditure (CapEx) on APs, the coverage is given to the entire
area in the first year. A maximum of 25 Wi-Fi APs with high gain antennas
are enough to give sufficient coverage for each square mile. By deploying
more APs per square mile, the demand per unit area never exceeds the
capacity of a single AP at any given time of the project life. The network
modelled is coverage-limited over the entire study period (ZHANG et al.,
2004). Another advantage of having a large number of Wi-Fi APs is that
operators can avoid outdoor CPE. This is also because operators have good
coverage and adequate capacity to run all applications on top of their
network, even during the first year. The subscribers within the city coverage
area can freely roam without re-association or re-authentication. The whole
city is covered with Wi-Fi radios and the APs are not only deployed in
residential and SOHO buildings, but also in locations where groups of
people meet: coffee shops, restaurants, malls, bus stops, subways, railway
stations, universities, airports, convention centres. Smart antenna
technologies can also be used to reduce inter-access point interference.
Wi-Fi Mesh: IEEE is setting up another new standard called 802.11s 13 to
extend mobility to Wi-Fi APs in very large Wi-Fi networks. IEEE is working
on the wireless LAN medium access control (MAC) and physical layer (PHY)
for Extended Service Set (ESS) in mesh networking (RAMAN et al., 2005).
Its mission is to develop new protocols for auto-configuring paths between
APs over self-configuring multi-hop topologies in a wireless distribution
system (WDS) to support both broadcast/multicast and unicast traffic in an
13 http://grouper.ieee.org/groups/802/11/ (URL accessed on Nov. 2005).
No. 63, 3rd Q. 2006
142
ESS Mesh 14. Although this standard is targeted to be approved by 2008,
many vendors are already developing Wi-Fi mesh systems using their own
proprietary technologies.
Figure 3 - Wi-Fi cells embedded in one WiMAX cluster
(One WiMAX main and eight mesh BS)
The proposed cost model
In any type of environment, there will be a mix of both the abovementioned scenarios. Figure 4 analyzes the main principle of the
methodology used for the cost analysis. The analysis is based on the
assumption that nearly 75% of the area consists of individual buildings,
houses and SOHO's; while the remaining 25% is made up of multi dwelling
houses and tall buildings with both houses and enterprises. The goal is to
cover the entire region with wireless broadband access so that all
individuals, namely residents, businesses, guests and tourists, will have the
opportunity to access broadband wireless anywhere and anytime in the city.
The economic feasibility of this proposed model in terms of net present value
(NPV) over a five-year period is calculated. Reasonable assumptions on
CapEx, OpEx, backhaul cost, leasing costs, etc., are made. With CapEx,
OpEx and different kinds of service revenue streams, the life-cycle
14 www.wikipedia.org (URL accessed on May 2006).
V. GUNASEKARAN & F. HARMANTZIS
143
economics are calculated in terms of investment and profitability. Unlike
cable or DSL operators, a Wi-Fi / WiMAX business model has a wide range
of pricing, roaming, payment and collection options that give operators the
positive cash flow to offer the most innovative broadband services at
affordable prices by lowering ongoing OpEx. With the proposed wireless
broadband business model, the end user enjoys many different service
options. Users with Wi-Fi client devices inside the city coverage area can get
on demand broadband services, as well as monthly subscription access. As
a long-term strategy, WISPs can build broadband wireless networks that can
support both voice and data for all residential and small business customers
in the serving area (WANICHKORN et al., 2002).
Figure 4 - Cost model flow diagram
No. 63, 3rd Q. 2006
144
A city-wide Wi-Fi business case
The total service area is assumed to be 135 square miles, which is
roughly equal to the size of the city of Philadelphia. This model uses Wi-Fi
as the last mile to reach the end users and uses infrastructure WiMAX mesh
for backhauling. Many cities have planned to deploy a new wireless mesh
network based on the pure IEEE802.11b Wi-Fi standard. Deploying pure WiFi mesh for larger areas presents a lot of engineering challenges, since this
requires multi-hop nodes with self-organizing and self-healing wireless mesh
networks. Though Wi-Fi mesh is also a viable option, it should be restricted
to a maximum of three to four hops 15. Therefore, for larger coverage areas,
there should be mix of both Wi-Fi mesh with limited hops and long distance
Wi-Fi with phased array antennas.
The analysis is based on the assumption that residential light users use
an average bandwidth of about 250 kbps (web browsing), while heavy users
use about 500 kbps. It is also assumed that 25% of all residents are heavy
users. Small businesses are assumed to use 1 Mbps on average. As shown
in table 1 in appendix 1, the number of subscriptions from residential users
(light and heavy users) in type 1, i.e., individual building and houses per
square mile in year 1, is assumed to be ten and the number of SOHOs
subscribed to the service in the first year is assumed to be five 16. In type 2,
i.e., multi dwelling units, the number of subscriptions from residential users
and SOHOs per square mile area is assumed to be three times that of
type 1. Since there is a huge variation in the growth rate of internet users, a
sensitivity analysis is done for this figure ranging from 20% to 80% per
year 17. The study consequently covers the impact of engineering and
economics, as a result of varied growth rates for different countries. If the
maximum utilization is only 25%, the service can be oversubscribed by 4
to 1.
Both the annual subscriber growth and the oversubscription factor impact
the engineering and economics of networks. The subscriber growth rate and
the oversubscription factor are varied to show the impact on the number of
15 www.chaska.net (URL accessed on June 2005).
16 In United Kingdom, there are 13 internet users per 1,000 inhabitants; in France there are 27.
Source: The Global Information Technology Report 2003-2004, Oxford University Press 2004.
17 The average internet user growth for 1999-2002 in United States was 52%; in France this
figure was 249%. Source: The Global Information Technology Report 2003-2004, Oxford
University Press 2004. For this base case analysis an average growth rate of 60% is assumed.
V. GUNASEKARAN & F. HARMANTZIS
145
radio channels needed for the WiMAX main BS and its impact on wired
backhaul cost. For both scenarios, the bandwidth needed for residential
users (light users and heavy users) in one WiMAX infrastructure mesh
cluster for one year can be calculated as follows:
Subscribers/square mile * Area (individual or multi-dwelling) * average
bandwidth width per subscriber (light and heavy users).
The bandwidth needed for SOHO's in multi dwelling and individual units
can also calculated in the same way.
For Voice over Wi-Fi, even if the basic codec techniques with the worst
compression schemes are used, each AP can support over 20 simultaneous
voice calls. The 802.11a has the capacity to handle roughly four times as
much voice traffic as 802.11b. VoWi-Fi calls use no more than one percent
(i.e., 100Kbps) of the total bandwidth available in either 802.11b/a or g.
Therefore, the data rate required for a two way voice channel is no more
than 200kbps/channel. The model assumes that only 50% of all subscribers
have both monthly data and voice plans. The voice codec 18 on the handset
is 64 Kbps with 20 millisecond voice frames. There are different codec types
based on the number of bits per second that need to be transmitted to
deliver a voice call. But even with the worst codec technique of 200 Kbps,
the net present value (NPV) is much better. This does not have a major
impact on the cost model. Since there are enough APs per square mile, the
capacity needed never exceeds the available capacity of a given AP.
However, in this model one is interested in capacity allocation for the WiMAX
mesh and main BS, rather than in the last mile Wi-Fi access. If one assumes
an average usage of ten minutes per hour, then the traffic generated per
handset is 0.17 erlangs 19. For on demand voice service, the model does not
take any busy hour voice traffic for any hand set. This is because this model
assumes that on demand VoWi-Fi calls will be placed only in cases where
mobile users need to make longer duration calls; as this gives them the
incentive to use the cheaper Wi-Fi network instead of their regular cellular
network. Therefore, on demand calls will be of longer duration than the
18
http://www.cisco.com/en/US/tech/tk652/tk698/technologies_tech_note09186a0080094ae2.shtml
(URL accessed on Oct. 2005).
19 http://www.proxim.com/learn/library/whitepapers/
voice_over_wifi_capacity_planning.pdf
(URL accessed on April 2005). As an example, with a total of 10,000 VoWi-Fi subscribers
(approximately) at the end of year five, each having a 17% chance of being active at any given
time, there are (on average) 1,700 active handsets. Therefore the active handsets in each
cluster would be 1700/7 and the total voice traffic generated in each cluster would be
approximately 50 Mbps at the end of year five.
146
No. 63, 3rd Q. 2006
regular fixed service residential VoWi-Fi calls, and so the model assumes
that there is always fixed capacity allocated to long duration callers.
Assumptions for capital and operating expenditures
WiMAX Micro BS Sector Controllers should be available for around USD
10,000 during the second half of 2005 – 2006. A WiMAX Pico BS controller
with wireless self-backhaul will be available for around USD 5,000 by the
end of 2005 20. Therefore, a two sector BS should cost approximately USD
10,000. The APs should have dual-band capabilities, providing Wi-Fi
connectivity for 802.11a/b/g devices and at the same time provide wireless
self-backhaul to WiMAX BS controllers. These outdoor APs will cost more
than the normal APs; the operators can purchase them on a wholesale basis
for approximately USD1,000-2,000 per outdoor AP 21. Appendix 2 shows the
number of WiMAX and Wi-Fi systems needed for the entire geographic area.
OpEx include tower leasing, backhaul costs and maintenance costs.
Maintenance costs are assumed to be 15% of equipment costs. The
assumptions made for wired backhaul costs are as follows: T1 (1.54 Mbps)
costs USD 250 per month; T3 (45 Mbps quantum) costs USD 2,000 per
month; and OC3 (155 Mbps quantum) costs USD 5,000 per month. The
wired backhaul is needed only in the main BS; in this case there are seven
for the total coverage area of 135 square miles. The WiMAX cells are made
of pico-cells; therefore, the BS antennas can be less than 50 feet. That is an
advantage, as operators can use the rooftops or the lamp-posts on the
streets. Since WiMAX mesh BS antennas can be less than 50 feet,
operators can negotiate for lower prices for the unused tower portion. Based
on this assumption, the cost of leasing space for placing a mesh BS
controller is estimated at USD 100 per month; while the per-sector leasing
for mounting antennas is assumed USD 150 per month. For the WiMAX
main base antennas, which should be at least 100ft up on the tower, a persector leasing cost of USD 300 per month and another USD 300 per month
for placing the BS controller is assumed. This model is based on the implicit
assumption that the city is providing its urban furniture at low cost for
mounting Wi-Fi APs, antennas or other equipment boxes, which is
approximately USD 120 per annum for a single light post or stop light.
20 Some of the assumptions in this model are based on private conversations with members of
the AT&T broadband wireless group on August 2004, and WiMAX vendors in the CTIA wireless
trade show on March 2005.
21 http://www.windowsmarketplace.com (URL accessed March 2006).
V. GUNASEKARAN & F. HARMANTZIS
147
The revenue model
Operators can have three sources of revenue: a) residential users (heavy
data/light users and monthly VoIP), b) the SOHO's, and c) on-demand
service (based on per connection fee for data and per call fee for voice).
Since the wireless service offered should be affordable to all classes of the
society, this model assumes a monthly subscription fee of USD 15 for light
users and USD 30 for heavy users. This is about 50% less than the regular
broadband fee of cable or DSL service. For business users the service can
be offered at a rate of USD 60 per month. For on-demand service this model
assumes an average Wi-Fi connection fee of USD 5 (good for a day). At this
rate, users can transmit unlimited volume of data traffic on top of a Wi-Fi call
fee of USD 2.50 per call. Only business travellers are considered using on
demand service for this analysis. This revenue model assumes that 2% of
business visitors in the city 22 use the pay-as–you-go service 23. Even with
this small percentage of visitors using the service, operators can generate
significant revenues compared to the other service models. However, not
only business travellers use the on demand service. People who go to coffee
shops, bookstores or libraries, etc., also use the Wi-Fi service (pay-as-yougo service). In the same way, this model assumes that 2% of them use
Voice over Wi-Fi service for longer duration voice calls instead of cellular
networks. By paying a Wi-Fi connection fee for data service or a VoWi-Fi call
fee for voice service, users can connect to any Wi-Fi APs in the city
coverage area on a given day. With a monthly plan for VoWi-Fi services
(USD 20 per month), residential users can make unlimited calls, both local
and long distance. Similarly, the SOHO VoWi-Fi service can be offered for
USD 30 per month.
Based on these numbers, the gross revenue and NPV with terminal
value 24 of the project is calculated as shown in appendix 3. For NPV
calculations, the model considers a weighted average cost of capital
(WACC) of 12%, which is an average for the telecom industry. The ongoing
22 The Philadelphia region attracted 6.3 million business visitors; leisure travellers rose to 17.9
million in 2003. http://www.centercityphila.org/docs/SOCC05_TOURISM.pdf (URL accessed on
Oct. 2005).
23 The British Library in central London, one of the most active and the largest public Wi-Fi
hotspot, has an average of 1,200 Wi-Fi connections or sessions per week. This main indoor WiFi zone, allows 3,000 visitors to connect to the internet and access e-mail using either their
existing service provider or the Library's pay-as-you-go service.
http://www.4ni.co.uk/nationalnews.asp?id=35396 (URL accessed on May 2005)
24 Investment Valuation: Tools and Techniques for determining the Value of any Asset by
Aswath Damodaran.
148
No. 63, 3rd Q. 2006
CapEx cost in the following years is minimal, since only initial investment is
needed to provide the full coverage. Hence, the cash outflow in the following
years is only due to the OpEx rather than capital costs. The operators can
break even in the second year, as long as they are able to attract 2% of total
business users 25 visiting the city, for on-demand service. To calculate the
terminal value, the model assumes a growth rate of 4%, a rate below the
nominal growth of the global economy. With 4% growth after the end of the
life of the project (i.e., five years), the value of the project in today's worth
with terminal value is calculated as shown in appendix 3. This cost model
does not take into account either the marketing or the customer acquisition
cost. These costs vary significantly among different WISPs and each service
provider has its own marketing strategy to acquire customers.
Sensitivity analysis
The analysis shows that the ongoing CapEx cost in the following years is
minimal. This is due to the fact that only an initial investment is needed to
provide full coverage: the cash outflow in the following years is due to OpEx
rather than capital costs. Several assumptions were made in the study and
the results were tested via sensitivity analyses. Since spectrum is the most
valuable asset, the total number of radio channels needed in a WiMAX main
BS was tested, as the subscriber growth rate is varied from 20% to 80%. As
seen in Figures 5 and 7, for higher subscriber growth, and less oversubscription factor, more spectrum is needed, i.e., additional 6MHz channels
in the proceeding year. Figure 6 shows that as the subscriber growth rate
increases, the wired backhaul leasing cost in WiMAX increases, although
not significantly. This is because of the traffic aggregation and the quantity
advantage. The study assumes 30 APs per square mile; this shows that
even with more Wi-Fi APs per WiMAX cell the NPV looks better.
25 To take a conservative view, the study assumes an average of only 2,500 session
connections per week in the whole coverage area. Even with 2% of the city visitors using the on
demand service, operators can generate significant revenues, compared to the monthly
subscription plan.
V. GUNASEKARAN & F. HARMANTZIS
149
Figure 5 - Number of WiMAX radio channels (6MHz) in main BS per year,
assuming different (annual) subscriber growths
14
20% Subs
Growth
12
10
BS
WiMAX Channels in Main
16
40% Subs
Growth
8
6
60% Subs
Growth
4
80% Subs
Growth
2
0
1
2
Ye ar
3
4
5
Figure 6 - Wired Backhaul cost from WiMAX main to POP per year,
assuming different (annual) subscriber growths
Backhaul cost
Millions
$2
Subs
Growth(20%)
Subs
Growth(40%)
$1
Subs
Growth(60%)
Subs
Growth(80%)
$0
1
2
Year
3
4
5
Number of WiMAX channels in
Main BS
Figure 7 - Number of WiMAX channels (6MHz) in main BS per year,
for different over-subscription factors
30
Oversubscr
iption (1:4)
25
20
Oversubscr
iption(1:3)
15
Oversubscr
iption(1:2)
10
5
Oversubscr
iption(1:1)
0
1
2 Year
3
4
5
No. 63, 3rd Q. 2006
150
Summary and conclusion
This paper explores a viable business model for the deployment of a city
wide Wi-Fi access network that uses WiMAX backhaul systems to lower
infrastructure costs. In the short run, Wi-Fi access to the last mile is
beneficial both for service providers and users, until mobile versions of
WiMAX systems (IEEE 802.16e) become a reality. Since the backbone
infrastructure is WiMAX, it should be easier to migrate to mobile WiMAX
access in the future, by using the same infrastructure. The main advantage
of this hybrid model (using both WiMAX and Wi-Fi) is that it turns the entire
geographic area into a wireless hot zone in a very short period of time.
Wi-Fi APs cost very little and can be easily mounted on lamp posts or
stop lights at a significant cost than leasing tall towers to mount the BS and
antennas. The advantage of using Wi-Fi as a last mile solution is that,
despite the higher density of Wi-Fi APs, it is still more economical compared
to alternative infrastructure costs. This is due to the fact that the capital costs
of APs are much lower and that additional investment in APs should not
impact the net profit. The main economic advantage of WiMAX infrastructure
mesh architecture is its low backhaul cost, due to traffic aggregation.
However, additional research and standardization work is needed to bring
the full benefits of mesh architecture or infrastructure mesh to
802.16/WiMAX.
A profitable business strategy for a WISP would be to serve a wide
variety of customers with the same infrastructure. Another main advantage
of this model is that residential and business users can access an on
demand (on a daily or hourly basis) service along with their option of
traditional monthly subscription. The result is one-time CapEx that can be
leveraged across different customer bases, making this an optimal solution
for broadband deployment. It was found that a higher operating profit can be
achieved even with a smaller number of subscribers in the initial stages of
the deployment. For a larger intake, the business case looks significantly
better. Therefore, with lower backhaul costs and zero dollars on CPE
subsidies and truck rolls, the combination of Wi-Fi with WiMAX to provide
both voice and data services, represents an attractive solution for deploying
city wide wireless broadband access.
V. GUNASEKARAN & F. HARMANTZIS
151
References
BAR F. & PARK N. (2006): "Municipal Wi-Fi Networks: The Goals, Practices, and
Policy Implication of the U.S. Case", COMMUNICATIONS & STRATEGIES, no. 61,
pp. 107-125.
BJORKDAHL J., BOHLIN E. & LINDMARK S. (2004): "Financial Assessment of
Fourth Generation Mobile Technologies", COMMUNICATIONS & STRATEGIES,
no. 54, pp. 71-94.
CAMPONOVO G., HEITMANN M., SLABEVA K.S. & PIGNEUR Y. (2003): "Exploring
th
the WISP industry Swiss case study", Proc. of 16 Bled Electronic Commerce
Conference.
EKLUND C., MARKS R.B., STANWOOD K.L. & WAND S. (2002): "IEEE Standard
802.16: A Technical Overview of the Wireless MAN Air Interface For Broadband
Wireless Access", IEEE Communication Magazine, vol. 40, no. 6, pp. 98-107.
FURUSKAR A., ALMGREN M. & JOHANSSON K. (2005): "An Infrastructure Cost
Evaluation of Single- and Multi-Access Networks with Heterogeneous User
Behavior", Proc. of IEEE Vehicular Technology Conference, vol. 5, pp. 3166-3170.
GUNASEKARAN V. & HARMANTZIS F.C. (2005): "Affordable Infrastructure for
Deploying WiMAX Systems: Mesh vs. Non Mesh", Proc. of Vehicular Technology
Conference, vol. 5, pp. 2979-2983.
HENRY P.S. & LUO H. (2002): "Wi-Fi: What’s Next?", IEEE Communication
Magazine, vol 40, no. 12, pp. 66-72.
LEHR W. & MCKNIGHT L. (2003): "Wireless Internet access: 3G vs. Wi-Fi?",
Telecommunications Policy, vol. 27, pp. 351-370.
RAMAN B. & CHEBROLU K. (2005): "Design and Evaluation of a new MAC Protocol
for Long-Distance an 802.11 Mesh Networks", Proc. of 11th Annual International
Conference on Mobile Computing and Networking, pp. 156-169.
SALKINTZIS A.K., FORS C. & PAZHYANNUR R. (2002): "WLAN-GPRS Integration
for Next-Generation Mobile Data Networks", IEEE Wireless Communications, vol 9,
no. 5, pp.112-124.
WANICHKORN K. & SIRBU M. (2002): "The Role of Fixed Wireless Access
Networks in the Deployment of Broadband Services and Competition in Local
Telecommunication Markets", Proc. of Telecommunications Policy Research
Conference.
ZHANG M. & WOLFF R. (2004): "Using Wi-Fi for Cost-Effective Broadband Wireless
access in Rural and Remote Areas", Proc. of IEEE WCNC / ICC, vol. 3, pp. 13471352.
No. 63, 3rd Q. 2006
152
Appendix 1: data and voice subcribers
The tables below show the data and voice subscribers for both scenarios: individual houses and
multi-Dwelling Units. Table 1 and 2 show total figures for a 1 square mile area on an annual
basis. Basic assumptions for subscribers apply for the first year; for the forthcoming years, a
growth rate of 60% is considered. Table 3 shows the total number of data and voice subscribers
for the entire coverage area of 135 square miles.
Table 1 - Total number of data subscribers per square mile area (individual houses)
Data Subscribers (Individual houses)
Year 1
Year 2
Year 3
Year 4
Year 5
10
8
2
5
16
12
4
8
26
20
6
13
41
31
10
21
66
50
16
34
Average subscribers / square mile
Average light users / square mile
Average heavy users / square mile
Average SOHOs / square mile
Table 2 - Total number of data subscribers per square mile area (multi dwelling units)
Data Subscribers (Multi Dwelling Units)
Year 1
Year 2
Year 3
Year 4
Year 5
30
23
7
15
48
36
12
24
77
58
19
39
124
93
31
63
199
150
49
101
Average subscribers / square mile
Average light users / square mile
Average heavy users / square mile
Average SOHOs / square mile
Table 3 - Total data and voice subscribers
for the whole coverage area of 135 square miles
Data and Voice Subscribers
Year 1
Year 2
Year 3
Year 4
Year 5
810
203
507
507
254
912
304
810
608
304
1,114
304
1,317
709
355
1,215
405
2,127
810
456
1,418
405
3,443
912
557
777
237
507
507
254
1,215
405
810
810
405
1,958
642
1,317
1,300
659
3,139
1,047
2,127
2,093
1,064
5,063
1,654
3,409
3,359
1,705
120,000
120,000
138,000
138,000
158,700
158,700
182,505
182,505
209,881
209,881
Individual houses
Residential light users
Residential heavy users
Business SOHOs
Residential VoIP(VoWi-Fi)
SOHO VoIP(VoWi-Fi)
Multi dwelling units
Residential light users
Residential heavy users
Business SOHOs
Residential VoIP(VoWi-Fi)
SOHO VoIP(VoWi-Fi)
Number of data connections
Number of VoWi-Fi connections
Appendix 2: number of WiMAX BS and Wi-Fi APs
Table 4 - Number of Wi-Fi and WiMAX systems (BS, controllers, APs)
Wi-Fi and WiMAX systems
Year 1
Year 2
Year 3
Year 4
Year 5
WiMAX main BS (135 sq. miles)
WiMAX mesh BS (135 sq. miles)
Total WiMAX main BS controllers
Total WiMAX mesh BS controllers
Wi-Fi APs (30 APs /square mile)
7
56
42
112
4,095
7
56
42
112
4,095
7
56
42
224
4,095
7
56
42
336
4,095
7
56
42
336
4,095
V. GUNASEKARAN & F. HARMANTZIS
153
Appendix 3: cash flow analysis
From table 4, the total CapEx and OpEx needed for Wi-Fi/WiMAX systems is calculated. From
table 3 the total revenue generated from all different services is calculated. The cash flow
breakdown is shown below.
Table 5 - Cash flow analysis
Total CapEx
Total OptEx
Revenue
Cash Flow
Year 1
Year 2
Year 3
Year 4
Year 5
$ 9,170,000
$1,339,800
$2,835,180
($7,674,620)
$0
$2,209,800
$4,145,400
$1,935,600
$560,000
$2,579,400
$6,225,630
$3,086,230
$560,000
$3,033,000
$9,449,468
$5,856,468
$0
$3,453,000
$14,550,426
$11,097,426
Assuming WACC (r%) = 12% and
TV =
g stable
= 4%, and the terminal value is given by
Cashflowt +1
r − g stable
Terminal value in 5 years = $11,514,323 = $144,266,533
12% − 4%
The present worth of the project including terminal value is
$7,674,620 + $1,1935,600 + $3,086,230 + $5,856,468 + $11,097,426 + $144,266,533
=
1.12
1.12 2
1.12 3
1.12 4
1.12 5
1.12 5
Therefore, the value of the project in today’s worth is $88,767,000.
Appendix 4: sensitivity on stable growth rate
The terminal value represents most of the value of the NPV because after the fourth year the
cash flow increases due to a decrease in the CapEx and OpEx (traffic aggregation). By
considering the cash flow (Cashflow t+1) for the t+1th year with 0% growth, the terminal value
will represent most of the present worth as shown in the table below. In fact, even changing the
stable growth rate after 5 years to 0% has little impact on current value of the project.
Table 6 - Sensitivity to stable growth rate
Growth
Cash flow
(t+1)th year
Terminal value in
year “t” (here t=5)
Terminal value in
today’s worth
NPV with
terminal value
0%
$11,097,426
$92,478,547
$52,474,1811
$59,381,106
1%
$11,208,400
$101,894,544
$57,817,700
$64,723,995
2%
$11,319,374
$113,193,741
$64,229,168.
$71,135,463.
3%
$11,430,348
$127,003,871
$72,065,407
$78,971,702
4%
$11,541,323
$144,266,533
$81,860,705
$88,767,000
Download