Affordable Infrastructure for Deploying WiMAX Systems: Mesh v. Non Mesh Vinoth Gunasekaran, Dr. Fotios C. Harmantzis Stevens Institute of Technology Telecommunications Management Castle Point on Hudson Hoboken, NJ 07030, USA {vgunasek, fharmant} @stevens.edu

Abstract— This study compares the economics of deploying conventional broadband wireless systems with different broadband Mesh topologies (i.e. 1 hop or 2 hop) that deliver high speed data services to residential, SOHO, and Large enterprises by a competitive local exchange carrier based on future WiMAX certified products. Regardless of fixed or mobile broadband wireless, the service offered should be affordable to all classes of the society. There are many new solutions, new income opportunities and sometimes-complicated business models for each type of mesh architecture. The different architectures of broadband wireless have been examined and we propose a cost effective infrastructure for deploying WiMAX systems.

Keyword- WiMAX, Mesh, Non Mesh, Backhaul, Leasing I.

INTRODUCTION

Internet which includes both narrowband and broadband has reached nearly 65 % in the US and Canada. However, broadband penetration is less, compared to the overall Internet penetration. While the backbone networks are heavily matured and more reliable with more bandwidth, the last mile to the end user is weak [6,8]. The economics of FBWA (Fixed Broadband wireless access) technology never made it suited for last mile; it was also thought that it can be deployed only in areas where there is no-preexisting infrastructure. But recent developments in WiMAX have been changing the whole BWA industry dynamics, making it an attractive alternative to DSL and Cable as a last mile to reach the customer. The WiMAX deployment as the last mile not only serves the residential and enterprise users but it can also be deployed as a backhaul for Wi-Fi hotspots and backhaul between the conventional cell towers[3]. There are different opinions of whether BWA will be successful as a last mile. Our study lays the groundwork for deploying future BWA systems based on WiMAX certified products. There are challenges in deploying WiMAX, but it has the huge potential to compete on a cost-per-megabyte level with cable and DSL, if both engineering and economics are carefully applied. We mainly focus on backhauling and tower leasing, by exploring several opportunities for significant cost savings like aggregating the backhaul traffic and optimal use of tower space.

This paper is organized as follows. Next Sections gives the background about WiMAX, Wi-Fi v. WiMAX, Mesh V. Non Mesh and also the different topologies that can be formed of infrastructure mesh. Section 3 dealt with how we have modeled our study. Section 4 presents the analysis of comparing different mesh WiMAX systems with non- mesh systems. In this section we have also shown in detail our analysis of how we have dimensioned the network with some reasonable assumptions for a particular case. We draw our conclusion in Section 5. II. BACKGROUND A. WiMAX to rescue BWA A group of vendors and service providers, among those who founded the WiMAX forum, believe that it would be widely deployed in the same way as that of Wi-Fi [2]. Standardization will not only reduce equipment and components costs, allowing mass production, but it also allows interoperability between equipments of different vendors. Though the standard does not describe how much of that capacity an operator should provide each user, a single base station could handle hundreds of megabits per second of data. The operators can offer more than 1 Mbps to a small or medium-sized business and 250 Kbps to 500Kbps to the residential users. The most suitable frequency band for WiMAX would be 3.5 GHz band followed by 5.2-5.8 GHz band. It is also expected that 2.5-2.7 GHz band would also be a potential band for WiMAX in many countries. B. Wi-Fi v. WiMAX A single Wi-Fi access point can cover approximately 0.0102 square miles. A typical urban cell site covers a radius of approx 1.212 square miles. The WLAN access points can only be deployed in strategic locations and it is very hard for the WISP to give a full coverage using Wi-Fi. WLAN regardless of the residential or business, need a backhaul connection to transmit data traffic to the Internet. It is very hard to justify, city wide Wi-Fi deployment in an economics perspective. Though Wi-Fi has a clear advantage over WiMAX as to reach the end user in the view of the fact that Wi-Fi is already been available in most of the end user devices and brings customer

acquisition cost much lower than would any other broadband technologies. So the other alternative broadband service model would be with both Wi-Fi and WiMAX where Wi-Fi can be used to reach the end user and at the same time can take advantage of WiMAX to minimize backhaul cost, reduce the time for service provisioning and reduce customer acquisition cost [7].

C. Mesh Networks vs. Non Mesh Unlike other mesh networks, the type of mesh network we are dealing is slightly different. Infrastructure Mesh is a type of mesh where subscriber nodes do not forward packets. It is contrasted with “ad-hoc” or “client” Mesh. Options in IEEE 802.16 are PMP MAC option and Mesh MAC option[1]. PMP MAC Option (Point-to-multipoint mode) is the default architecture, which is supported and enhanced by the WiMAX Forum. Mesh MAC Option is a type of “client mesh”. This option is not actively discussed or supported (still on research). Additional research and standardization work is needed to bring full benefits of mesh architecture or infrastructure mesh to 802.16/WiMAX. Infrastructure mesh has many advantages over the client mesh as it is more secure, more predictable, easier to manage, and does not suffer from initial seeding issue.

Figure 1. Mesh cluster 1:6(six one hop)

Figure 2. Mesh cluster 1:12(6 one hop & 6 two hop)

D. Different Topologies of Infrastructure Mesh Infrastructure mesh is a new way of delivering broadband access for residential and SOHO’s. There are different types of architectures by which mesh systems can be formed. Weighing the advantages and disadvantages of each different systems and more careful analysis is required before WiMAX deployment. We have considered different topologies from one hop to three hops for our analysis. We have chosen hexagonal cell and the size of the clusters in such way it tessellates the plane. The number of cells that can form a regular cluster pattern is given by the formula m^2 + n^2 + m*n where m, n are integers. This gives 3, 4, 7, 9, 12, 13, 16, 19, 21.etc. So it has a N-sized cluster with one main base and (N-1) mesh BS. Since we need one main BS surrounded by mesh BS, the size of the clusters of mesh BS (Base Station) with one Main BS chosen would be: 7, 13, 19, 27, and 37. As seen in the figures below we have taken very limited hops: 1 (no forwarding) to max 3 with symmetrical pattern to form a regular pattern with main BS in the center.(1:6, 1:12, 1:18, 1:26, 1:36). Main BS with wired backhaul at the center of a cluster of Mesh Base Stations is connected wirelessly to one (or more, for redundancy) Main BS. For example as shown in Figure 1, if we consider a cluster size of 7 cells, there will be one Main base station surrounded by six Mesh base stations. In this architecture the Main BS aggregates all the traffic from the Mesh base stations and then takes them via wired backhaul to the POP. In the same way we have considered different topologies of maximum up to three hops for our analysis.

Figure 3. Mesh Cluster 1:18(6 one hop & 12 two hop)

Figure 4. Mesh 1:36 (6 one hop,12 two hop ,18 three hop)

Considering the 1:6 case with the coverage area of 900 square miles of 1 mile cell radius there are 297 mesh BS and 50 main BS. So there will be a total of only 50 wired backhaul facility needed to serve all the users in 297 mesh BS coverage apart from its own users in the main BS coverage area. As seen in Fig 5 the number of main BS decreases as the mesh becomes

deeper. For a 1:6 case it needs 50 main BS in compare to only 10 main BS in 1:36 case.

5

337

10

4

334

13

328

3

Mesh BS

19

Main BS 320

2 1

297

260

280

27 50 300

320

340

360

Figure 5. Number of Mesh cells v. Main cells in Different Topologies (900 square miles of 1 mile cell radius)

III. ENGINEERING AND ECONOMIC MODELING Engineering should be more carefully applied to have a economical network deployment when a broadband service is established and also at each phase the operators can save money, without compromising the quality of service. Engineering include both radio capacity planning as well as backhaul capacity planning. Each and every parameter is linked with the business case, in order to come up with a cost effective system. The product cost of WiMAX solutions should also be very less compared to the current fixed wireless system apart from the customer premise installation cost elimination. A. Backhaul Infrastructure planning Regardless of any wireless systems, operators need wired backhauling. The more number of T1 or E1 ports on the Base station controller, the more bulky the controller, the harder to manage them. Aggregating backhaul lines into higher capacity lines will not only be cheaper but also reduces the physical space compared to smaller speed circuits. The mesh topologies have less backhaul cost compared to the Non Mesh case. A deeper mesh has more advantages in regards to backhaul cost, when compared to a lesser hop mesh. In conventional Non mesh as access networks are built to support wireless services, each newly deployed services requires additional dedicated leased lines. Optical standard pipes available are T1 , T3, OC3 ,OC-12,…….where T1(1.54 Mbps), T3 (45 Mbps quantum), OC3 (155 Mbps quantum), OC-12 ( 623 Mbps quantum) . B. Radio Capacity Planning Spectral efficiency is one of the most difficult factors to quantify, and it has a significant impact on the network economics as well as the quality of service. Spectral efficiency is defined as the bits/sec transmitted per Hz of the bandwidth and it is indirectly linked to the modulation schemes and the frequency reuse ratios. Radio equipment costs can be held

down by enhanced use of spectrum utilization, and sectorization. We made reasonable assumptions on data rate and the sectorization plan. We also incorporated frequency reuse plan in our model. The operating frequencies are between 2 to 6 GHz, which includes both licensed and unlicensed spectrum. In the case of licensed spectrum, we assume that the operator owns the spectrum already and it is taken to be a sunk cost. IV. ANALYSIS AND RESULTS We analyzed by comparing the economics of the above mentioned network topologies with some reasonable assumptions on data rate, sectorization and frequency reuse plan and also the type of cell (microcell-1 mile radius). The use of smaller cell and sectorization is to gain additional capacity as the system grows. As an example we have shown here how we dimensioned the network for the infrastructure mesh topology of 1:6. Here, we considered an area of 900 square miles (big metro city) which includes both urban and sun-urban area. The horizon of the project is five years. The base stations need to be of shorter reach (1 mile cell for our analysis) because only true plug and play CPE devices with non-line of sight operation can reduce the cost of installation, including truck rolls and service personnel. So if the radius of a Main BS or Mesh BS is 1 mile then the area covered by one main or mesh base station would be 2.597 square mile (hexagonal cell). Area covered by one mesh cluster (1:6) is 18.18 square miles with the total number of clusters to be 50. A. Bandwidth Calculation The total bandwidth needed is the capacity needed for residential, SOHO’s and the large enterprise & hotspots (WiFi). The operators can also plan to use the same infrastructure for residential and SOHO’s in the alternate times of the day. Large businesses are usually not oversubscribed and they have dedicated bandwidth. Bandwidth needed for residential users: For light users, average bandwidth needed (Web Browsing) might be around 250 kbps and for a heavy user it may be around 500 kbps. We assume that the number of light users per square mile in year 1 is 3 subscribers and the heavy user per square mile is 1 subscriber. We include a subscriber growth of 30% every year. If we assume that if the maximum utilization is 50%, it can be oversubscribed 2 to 1. For example if the user is doing web browsing, there may be an inactive period during which other users can use the available bandwidth with the guaranteed QoS. If it takes 5 to 10 sec to download a 30 kbps web page and if the user spends a minute to read then it means that the user is active only for a maximum of 10 seconds, the time he/she spends in downloading the data. So the rest of the inactive time can be shared by the other users in the system. The bandwidth needed for light users in one infrastructure mesh cluster for year 1 (Mesh 1:6 case): Subs/square mile * Area * average BW/Subs (Heavy users) =13.65 Mbps. In the same way the bandwidth needed for heavy users in one

infrastructure mesh cluster for year 1 (Mesh 1:6 case): Subs/square mile * Area * average BW/Subs (Heavy users) =9.09 Mbps Bandwidth needed for SOHO’s and large Enterprise users: If we assume that the average bandwidth per SOHO is 1000kbps and the number of small businesses per square mile at year 1 is 2 and if the number of SOHO’s increases every year by a factor of 2 then at the end of year 5 there will an average of 10 SOHO’s per square mile which is a reasonable figure. The bandwidth needed for a SOHO in an infrastructure mesh cluster (1:6) for year 1 is given by: SOHO/square mile * Area * average BW/SOHO =36.36 Mbps. In the same way for large enterprise, we assume in such a way that at the end of year 5, there should be at least one enterprise or Wi-Fi hotspot in a given area (with a constant growth rate from year 1). Therefore bandwidth needed for large businesses in an infrastructure mesh cluster (1:6) for year 1 is given by: Large enterprise/square mile * Area * average BW/Large enterprise or Wi-Fi hotspot=5.45 Mbps. B. Backhaul Capacity calculation If the total bandwidth needed for residential users in one infrastructure mesh cluster (1:6) is 22.726 Mbps. then the maximum backhaul needed with 50% over-subscription is only 11.36 Mbps. In the same way for small businesses the backhaul needed with oversubscription would be 18.18Mbps. As seen in the Fig.10 the backhaul needed in main base station for all mesh topologies is large compared to that of non mesh case. Since there is no aggregation of traffic in non mesh case the backhaul must be provisioned by one or more T1 lines. If the leasing cost of a T1 connection is $250 per month, T3 is $2000 per month and an OC3 is $5000 a month, we see from Fig.11 that the cost of backhaul provisioning for all the mesh topologies regardless of one hop, two hops or three hops has a clear advantage over the non mesh case. Although one T3 has significantly larger capacity compared to 10 T1’s, the cost of leasing one T3 is lot cheaper than 10 T1’s.The cost of backhaul provisioning increases as the number of T1 line increases but mesh takes advantage of the volume discount due to aggregation.

then the capacity per sector for each main or mesh BS would be 18 Mbps. The radio capacity needed for each mesh BS is very less compared to the main BS because main BS needs radio capacity for the users in its own cell in addition to wireless backhaul from all the other 6 mesh BS (1:6). On the other hand mesh needs only capacity for its own users in its cell coverage area and for its own backhaul. We implement sectorization in the order of omni, 2 sectors, 4 sectors, or 6 sectors. We have dimensioned in the way that there will be a maximum of only six sectors i.e. only 60 degree configuration. We implement 2 sector or 4 sector or 6 sectors with only two frequency channels as seen in Fig.6, 7, & 8. An effective sector is defined as the total number of sectors available due to addition of more channels in the actual given sector. The channel numbers and the sectors selected are even i.e. 2, 4, 6, 8, 10,…etc, so that the same channels can be used in alternate sectors to avoid co-channel inference. As seen in Fig.12 & 13, mesh topology (1:6) needs two channels with two sectors during the initial phase and as the system grows sectorization and addition of channels go hand in hand. For example with an actual sector size of six with four 6MHz channels the effective sectors would be 12 in year 4 and 5. Since we assume that the operators have both the licensed and unlicensed spectrum, we haven’t set any upper bound on the number of channels. As shown in Fig 10 the effective sectors in main BS may vary from 2 to 10 in the mesh 1:6 case due to addition of more channels in the following years. For deeper mesh topologies i.e. with more than two hops the main BS will need more radio channels (> 4) when compared to the lesser hop mesh architecture. The deeper the mesh the operating cost increases in terms of tower leasing for the main BS and on the other hand the spectrum requirements increases as the need for effective sectors increases.

C. Radio Capacity calculation We have adopted sectorization for capacity expansion instead of cell splitting. The cell size deployment we have opted to be micro cell (1 mile) so that the base station antenna can be of less than 70 feet. This has an advantage of using the lower part of the tower or the operators can use the rooftops or the lamp posts mounted on the streets. The operators can use both licensed and unlicensed spectrum and since licensed spectrum is most precious resource there should be more resource left over for future capacity expansion. Though many wireless vendors are offering flexible channel size we have taken a standard size of 6 MHz for our calculation. If we have a channel size of 6 MHz with the spectral efficiency of 3 bits/Hz

Figure 6. Sectors (2), channels(2)

Figure.7 Sectors(4), channels(2)

Figure 8. Sectors (6), channels(2)

Figure 9. sectors (6), channels(4)

CONCLUSION

Backhaul needed in Mbps

Size of Backhaul pipe per wired backhaul location

Mesh 1:6

790 740 690 640 590 540 490 440 390 340 290 240 190 140 90 40 -10

Mesh 1:12 Mesh 1:18 Mesh 1:26 Mesh 1:36 Non Mesh 1

Year 3

2

4

5

Figure 10. Size of Backhaul per wired backhaul location

Millions

Total Backhauling (annual cost) $9 $8 $7 $6 $5 $4 $3 $2 $1 $0

Mesh 1:6 Mesh 1:12 Mesh 1:18 Mesh 1:26 Mesh 1:36 Non Mesh 1

2

3 Year

4

5

Figure 11. Total backhauling (Annual Cost) Number of radio channels needed for Main Base station in different Mesh topologies 18

Mesh 1:6

No of 6 MHz Channels

16 14

Mesh 1:12

12 10

Our results show that with reasonable assumptions for data rates, sectorization and frequency reuse plans mesh based WiMAX systems is more affordable and viable compared to non mesh case. Our results are based on high level study which is technology based using economic modeling. Since our study is generalized we like to conduct a site specific design to include parameters like terrain factors and actual traffic in a particular area and also to include an upper bound on the channel number. This analysis focused mainly on providing broad band wireless Internet access services without considering the VoIP services to residential and business users. We also like to extend our future study by considering voice over WiMAX. Tower leasing and backhaul costs play a significant role in deploying BWA systems. Economic advantages of mesh architecture are: (a) lower backhaul costs due to aggregation of traffic at wired backhaul sites; and (b) possibility of using lower antenna heights. Economic disadvantages of mesh architecture: (a) the main BS has high tower leasing costs and spectrum requirements, because of radio backhaul; and (b) operating costs increase as mesh becomes deeper (more hops). So from economics perspective Mesh option is promising for medium-sized cells compared to conventional non mesh case. ACKNOWLEDGEMENTS Many thanks to N. K. Shankaranarayan and Byoung Jo J.Kim, both with AT&T Labs, NJ, USA, for their valuable comments in understanding the economics and technical aspects of Infrastructure Mesh topologies.

Mesh 1:18

8 6

Mesh 1:26

4 2 0 1

2

3 Year

4

5

Mesh 1:36

Figure 12. Number of radio channels needed for Main BS in different Mesh Topologies Number of Effective sectors needed for Main BS due to sectorization & addition of channels

Mesh 1:6

REFERENCES [1] [2] [3] [4] [5] [6]

60 Mesh 1:12

Effective sectors

50 40

[7]

Mesh 1:18

30

[8]

20

Mesh 1:26

10 0 1

2 Year 3

4

5

Mesh 1:36

Figure 13. Number of Effective sectors needed for main BS due to Sectorization & addition of channels

802.16 Working Group. http://www.ieee802.org/16 WiMAX Forum, http://www.wimaxforum.org http://www.alvarion.com B.J.Kim,N.K.Shankaranarayanan,P.S.Henry,K.Schlosser and T.K Fong, “The AT&T Labs Broadband Fixed Wireless Field Experiment,”IEEE Communication Magazine”,Vol 37 Issue 10, pp.56-62, Oct 1999. C.Eklund, R. Marks, K. Stanwood, & S.Wand “IEEE Standard 802.16: A technical overview of the wirelessMAN air interface for broadband wireless access” IEEE Communication Magazine, June 2002. C. Cordeiro, H. Gossain, R. Ashok and D. Agarwal ,”The last mile: wireless technologies for broadband and home networks” Center for distributed and Mobile Computing ,University of Cincinnati.,OH V.Gunasekaran, F.Harmantzis, “Migration to 4G-ubiquitous broadband: Economic modeling of Wi-Fi with WiMAX “To appear in WWC 2005”, in press. K. Wanichkorn, M.Sirbu “The role of Fixed wireless Access Networks in the deployment of Broadband Services and Competition in Local Telecommunication Markets” Dept of Engineering and Public policy, Carnegie Mellon University. Pittsburgh, PA