Fibre access network for India as India poises for growth

Fibre access network for India as India poises for growth A. Jhunjhunwala Abstract: Fibre plays a major role in backbone networks, not only making it capable of carrying large amounts of traffic but also making it cost-effective. However, the decrease in backbone cost has so far been nullified by an increase in the cost of local loops. Fibre access networks in developing countries like India are set to play a major role in expanding the teledensity. In dense urban areas of India, fibre access is not only the most cost-effective access technology, but is also contributing to building a multiservice network integrating telephony, TV and video-on-demand services.

1

Introduction

India has a population of more than one billion [1]. It has barely 40 million fixed telephone lines, 12 million mobiles and 3–4 million Internet connections [2]. Even compared to other developing countries, its teledensity is rather poor [2]. It has set itself a target of 100 million telephone lines by 2005 and about 200 million by 2010 [3]. A year ago these numbers looked unrealisable, but today there is little doubt that India will achieve these numbers and may even exceed them. This paper examines the role that optical fibre systems, particularly fibre access systems, are playing in India to achieve these numbers. The author examines what has led to the confidence in the targets being achieved in India. An understanding of affordability and reduction in capital expenditure (CAPEX) is considered key to this growth. The telecom network and contributions to CAPEX from its different elements are examined. Access is recognised as the key contributor to CAPEX, and the role that fibre access systems are playing in India towards reduction of CAPEX is examined. Fibre-to-the-building, and copper in the ‘last meters’ is becoming a techno-economic reality, and this is illustrated by a business case. The applicability of the approach to other developing countries is considered. 2 Towards 200 million telephone and Internet connections by 2010 The confidence in achieving 200 million lines by 2010 comes from an environment today in which multiple operators are enthusiastically expanding. Such competition is bringing down prices rapidly and customers are beginning to enjoy significant benefits. Long distance call charges were Rs. 30 [Note 1] per minute two years age, but have now fallen to under Rs.5 per minute. International call charges have gone down by 40% and are likely to fall further in the coming months. Voice-over-IP communications has been legalised and mobile call charges have been falling very rapidly [4]. r IEE, 2003 IEE Proceedings online no. 20031143 doi:10.1049/ip-cds:20031143 Paper first received 2nd April and in revised form 3rd September 2003 The author is with the Department of Electrical Engineering, Indian Institute of Technology, Chennai 600036, Tamil Nadu, India IEE Proc.-Circuits Devices Syst., Vol. 150, No. 6, December 2003

The process started in 1993 when the government of India decided to open up telecom operations in India and invite private players as operators. Unfortunately, the liberalisation process was initially unsuccessful largely due to the inexperience of the government as well as potential operators. In 1999 a new telecom policy, referred to as NTP’99, put the telecom liberalisation process in India back on track [5]. Subsequently, in 1999 Internet services were opened up and by 2002 national long distance and international long distance services were opened up. The government-based, hitherto monopoly, operators were converted to corporates and a National Regulatory Authority was set up [4]. Within two years, six–eight operators were operating in most circles and true competition began. Customers were able to obtain not only better prices, but also better services. Everyone started working in earnest to expand the subscriber base.

2.1

Affordability in India

Liberalisation in telecom alone was not sufficient to start telecom expansion in India. The upbeat mood has emerged primarily because Indian policy-makers, regulators and operators have, for the first time, a reasonable understanding of what would make the market in India grow. Until about two years ago, the fully loaded CAPEX involved in setting up a telephone line in India was about Rs. 30 000 [5]. With such a CAPEX, what kind of average revenue per user (ARPU) would an operator require to break even? Finance costs in India hover around 15%. As the equipment installed may need replacement in about 10 years, 10% depreciation is required. Operation and maintenance costs are another 10% of CAPEX, and licence fees, spectrum charges and taxes may amount to another 5% of CAPEX. Thus, a total of 40% of Rs.30 000 or around Rs.12 000 per year is required as ARPU. This works out at Rs.1000 per month and is unaffordable for most Indian households. India has 62 million urban and 135 million rural households. Assuming that an Indian household would be willing to spend 3% of its income on telecoms, Figs. 1 and 2 show the amount that Indian urban and rural households would be willing to spend today on telecom [6]. Note 1: 50 rupees equal one Euro today

467

well as the lowering of equipment costs due to rapid expansion of the Chinese market, Indian operators have reduced their CAPEX to around Rs.18 000 per line. Efforts continue, and most operators are confident of going below Rs.10 000 per line cost in the next few years. How has this happened?

millions of rural households

102.1

3 17.0 10.0 3.9 130

Fig. 1

220

1.9

1.0

330 440 630 940 telecom expenditure, Rs/month

0.3

0.3

1560

3760

Indian rural telecom affordability

millions of urban households

21.2

15.8 12.8

5.8 3.2

130

Fig. 2

220

2.0

330 440 630 940 telecom expenditure, Rs/month

0.8

0.9

1560

4420

Indian urban telecom affordability

It is clear that less than 3.7 million or 6% of urban and 1.6 million or 1.2% of rural households can afford to spend anywhere close to Rs.1000 per month. At Rs.30 000 CAPEX, India has a very narrow market. But the same figures show that India has a huge potential market, if CAPEX can be reduced to Rs.10 000 per line. At 40% of CAPEX required for ARPU, this works out to a little over Rs.300 per month. 55% of urban households, and 25% of rural households can spend this amount on telecom. But as some households can spend more, it is possible to get an average revenue of Rs.300 per month from over 50% of the 197 million households in India. Add to this those households that will have two or three telephones, and the business connections, and 200 million connections look very possible. The key to the growth of telecom in India is therefore, to bring down the CAPEX to under Rs.10 000 per line. This was not recognised until a few years back. Upto the mid1990 s, the government monopoly operator provided telecom services on a cost plus basis. The teledensity was just above 2% and the ARPU was close to Rs.1000 per month. When private operators began their services, they concentrated on cherry-picking high-paying subscribers from the incumbent operator. CAPEX was not a serious concern. Only after NTP’99 came into being, did the operators started seriously looking at expansion of the telecom market in India. Policy-makers and regulators for the first time started looking at the affordability numbers. They started looking for lower cost technologies and started bargaining to reduce CAPEX. Over recent years, helped by low-cost technologies emerging from India and China, as 468

Elements of telecom network

Let us examine what contributes to CAPEX on the telecom network today. The numbers provided here are for Indian geography and topology and may differ in other countries. The backbone network in India today constitutes 10% of CAPEX. It consists of a fairly dense optical fibre network, and fibre cost and installation (digging) is the primary cost. India already has a fairly dense fibre network. While the incumbent operators had already laid fibre to almost every taluk (county) town, private operators have feverishly added fibre recently. Several other agencies such as Railways (Railtel) and Gas Authority of India Ltd (GAIL) are also laying fibre and selling fibre or bandwidth. Most parts of India will have a 20 km by 20 km grid and 85% of the 650 000 villages in India will soon be within 10 km of a fibre point. The electronics used for lighting the fibre consists primarily of STM-1, STM-4 and STM-16 equipment. Today this equipment is increasingly supplied by either Chinese companies such as Huawei, UTStarCom, and Zhongxing Telecom Equipment Corp (ZTE) or by Indian companies such as Tejas Networks. This equipment has better features and is more cost-effective than that from the West. WDM equipment is used, but to a much lesser extent. The bandwidth requirement is moderate and there is enough fibre. Major investment in WDM equipment has therefore been postponed for several years. Backbone switches and routers contribute between 5 and 10% of CAPEX today. The equipment is mostly brought in from Alcatel, Siemens, Ericsson, Nortel, Lucents, and Cisco. Huawei from China is trying to gain entry but has not succeeded so far. Service platforms contribute another 10–15% of CAPEX on telecoms today. They include operator and maintenance console (OMC), network management systems (NMS), customer care and billing systems, intelligent network (IN) services, platforms for web-hosting and mail services, etc. Systems are mostly from IBM, HP or Sun and use database management systems such as Oracle. Application software is obtained from a host of companies from the East as well as the West. In recent times, some operators have started using Linux based personal computers to provide these services. The access network, however, dominates CAPEX and can contribute as much as 65% of it. This includes the access switch, local loop and subscriber terminal. Mobile wireless, fixed wireless and fibre access system are the dominant systems, which sometimes compete, but mostly supplement each other in providing services. It is in this part of the network that major innovations have occured in recent years. Reduction in the cost of the access network has driven the overall CAPEX down. 4

Fibre access solutions

Fibre plays a major role in the backbone network, making it not only capable of carrying large amounts of traffic, but also making it very cost-effective. However, the decrease in backbone cost has so far been nullified by an increase in the cost of the local loop. Could fibre make a major difference to the access network of the future? [7]. IEE Proc.-Circuits Devices Syst., Vol. 150, No. 6, December 2003

Figure 3 shows the evolution of fibre/copper local loop technology. In the early 1980s, exchanges in urban areas served subscribers within a 6–8 km radius using long copper local loops. The late 1980s saw the emergence of small remote line units (RLUs) with fibre connections to the main exchange fibre, reducing, the copper loop to 3–4 km [8]. The digital loop carrier technologies of the late 1990s took the fibre closer to the subscriber. Fibre rings were terminated on a remote terminal, which was now cost-effective serving as few as 500 subscribers. The copper loop was now barely 500–700 m [9]. However, even with these developments the network remained a voice-centric network.

traffic, but also video/TV services) with fibre as a driver. Could fibre be taken closer to subscribers (fibre to the building or fibre to the curb), help build this multi-service network and still be cost-effective? How deep could this fibre go? The cost drivers for fibre being deeper depend on the cost of opto-electronics for terminating the fibre, the cost of electronics to provide the telephony interface on copper and the cost of power back-up required. The business case, however, equally depends on the revenue and therefore the customer density that an operator can hope to get and the paying capability of these potential customers.

Early 1980s exchanges with long copper…6−8 km

trunk lines

expensive time consuming to deploy difficult to maintain

E X C H A N G E

RLU Late 1980s RLUs last mile copper…3−4 km 0.4 mm RLU

Late 1990s DLCs fibre rings copper in 500−700 m

fibre ring network

short copper

Fig. 3

Evolution of fibre/copper in the local loop

Data networks were emerging in parallel (see Fig. 4), and were generally based on a dial-up modem on an analogue copper loop. However, two other access technologies for the internet emerged over time. One was based on coaxial cable used for video delivery for television [10] and the other used DSL technology [11] on a copper local loop, for telephony. The time had come to merge these networks and make a multiservice network (providing not only voice and Internet

VDSL/ Ethernet

50M 10M ADSL

8M HDSL/ sHDSL

2M ISDN 128k

Fig. 4

Analogue modem

Speeds driven by application Bandwidth requirements touching 10 Mbit/s

Evolution of data access network on copper

IEE Proc.-Circuits Devices Syst., Vol. 150, No. 6, December 2003

With falling costs of optoelectronics and integration of termination electronics for telephones and power-backups, it started making sense to take fibre right to the building in the dense urban areas in India. Somewhere in the middle of 2002, the calculations started favouring this, and new kinds of fibre access systems started emerging. Fibre-to-the-building (FTTB) and fibre-to-the-curb (FTTC) have copper limited to about 100 m within the building. Copper in the last metres enables use of Ethernet as the final delivery vehicle. The high volume production of Ethernet and the fact that it sometimes eliminates the use of customer premises equipment, coupled with the 100 Mbit/s throughput that this interface can provide make it the preferred technology. Services could now converge. Convergence, however, depends on the features that such a network can support. While quality of service (QoS) features [12] are important, it is features like VPN [13] and transparent LAN services [14] which are key to the convergence. However, it may not always make business sense to take the fibre right up to the building; sometimes it may be better to terminate it at the curb. The copper loop, in this case could be typically 300 m and one may have to use some 469

FTTB / FTTC

RT

DSL/ Ethernet

PC

multiservices CO CPE

fibre ring

TV telephone

Fig. 5

Emerging integrated services network

kind of DSL. At yet other locations, fibre PON, with Ethernet over passive optical network may be the optimum choice for India. Figure 5 shows the emerging integrated services network in India. The remote terminal (RT) located in the building provides voice telephony on separate copper, and Internet and TV services on Ethernet or DSL on another CAT-6 copper pair. It is possible in future that voice will be provided as voice-over-IP and a separate pair to carry analogue voice may not be required. It is also conceivable that wireless LAN technologies like 802-11 may be connected to the RT tomorrow and provide tetherless services within a building. Different architectures are emerging for these fibre access systems (FAS). One such architecture involves two fibres being terminated on an RT. One fibre carries STM-1 traffic and supports narrowband services like POTS, ISDN and n  64 kbit/s services, and the other fibre supports a gigabit Ethernet ring carrying broadband traffic such as Internet, TV services, and video-on-demand. The remote terminal is wall-mounted, has an eight-hour battery backup powered mains and is cost-effective serving as few as 24 subscribers. The system is coupled with a sophisticated multilevel network management system, which performs comprehensive management of services to each subscriber and the total network. Another architecture uses a fibre passive optical network as shown in Fig. 6. 5

Does FAS make business sense?

To understand the business case for such fibre access systems in dense urban areas, let us take an example. Chennai, a large metropolis in South India, has a dense urban area roughly 10 km by 10 km. Dwellings in this area are closely packed and many are multistoried. Approximately 5 million people live in this area in about 1 million

homes. The number of homes per sq km therefore works out at 10 000, or 100 homes in a 100 m  100 m area [15]. To serve subscribers in this area using a FAS, two approaches are used. The first approach uses a fibre grid 100 m by 100 m and proposes to serve about 24 subscribers from each grid point. The copper length is less than 100 m and (100 BaseT) Ethernet is used as a ‘last metres’ technology. As there are 100 homes covered by a grid point, an operator has to be able to acquire 25% of the homes as customers. As shown in Table 1, the access cost works out to be Rs.7000 per subscriber and the total CAPEX would be approximately Rs.10 000 per subscriber (assuming that 70% of CAPEX comes from access in dense urban areas) [15]. Assuming Rs.300 as revenue per month from telephony and adding video and Internet revenue of Rs.100 each per month, an operator may hope to earn Rs.500 per month or Rs.6000 per year. There is a clear business case. The second approach uses a fibre grid of 400 m by 400 m and a copper loop of 300 m. Obviously higher cost DSL has to be used in such situations. Each grid point needs to serve about 96 subscribers (in about 1,600 homes in the area covered) to have a total access cost of Rs.10 000 (DSL cost is higher) and per-line CAPEX of about Rs.14 000 [15]. One now needs higher revenue per subscriber. As shown in Table 1, a Rs.400 per month telephony revenue, and video and Internet revenue of Rs.200 and Rs.300 per month respectively would drive the per-user revenue to Rs.900 per month or Rs.10 800 per year, making a business case. But, can one obtain this higher revenue? As an operator needs to cover only about 6.5% of the homes in an area, it should be possible to target this revenue number. The business case is not limited to large conurbations. Table 2 presents a mid-size city, Madurai and a smaller town Erode, in south India. Madurai will have a 3 km  3 km area and Erode will have a 1 km  1 km area

PC ET PC RT

fibre backhaul ring

fibre PON TV

Fig. 6 470

Fibre passive optical network in the last few metres of local loop IEE Proc.-Circuits Devices Syst., Vol. 150, No. 6, December 2003

Table 1: Business case for fibre and Ethernet or DSL on copper Ethernet option: deep fibre, higher sub density

DSL option: not so deep fibre, medium sub density

Fibre grid

100 m  100 m

Fibre grid

Copper hop

100 m

Copper hop

400 m  400 m 300 m

No. of lines

24

No. of lines

96

Customers needed to be acquired

25%

Dense urban market share

6.5%

Cost of fibre per line

Rs.2000

Cost of fibre per line

Rs.2000

Electronics/CPE

Rs.4000

Electronics/CPE

Rs.6000

Copper

Rs.1000

Copper

Rs.2000

Total access cost

Rs.7000

Total Access cost

Rs.10 000

CAPEX per line

Rs.10 000

CAPEX per line

Rs.14 000

Telephony

Rs.300

Telephony

Rs.400

Video

Rs.100

Video

Rs.200

Internet

Rs.100

Internet

Rs.300

Total revenue per month

Rs.500

Total revenue per month

Rs.900

Revenue per year

Rs.6000

Revenue per year

Rs.10 800

Table 2: Business caseFevery city has a dense part where fibre access makes most sense Chennai

Madurai

Erode

Dense urban area

10 km  10 km

3 km  3 km

1 km  1 km

Homes in dense area

1 million

90 000

10 000

City population covered

75%

35%

12%

Ethernet @ Rs.500 pm

25%

25%

25%

DSL @ Rs.800

6.5%

6.5%

6.5%

Market share of urban homes in dense area

with density of homes/businesses similar to that in the dense part of Chennai. In other words most Indian cities would have a high density area, where a fibre access system with last metres in copper starts to be a sound business proposition. As the number of services that can be offered grows, the area served in a city is only likely to grow. Fibre-to-the-building/fibre-to-the-curb has become a reality and operators in India are making plans to provide such systems in the near future. The equipment supply for this market is increasingly being dominated by Indian/ Chinese companies. 6

Conclusions

The likely percentage of the access market in a country like India which would be based on fibre is a difficult question that only time will answer. The two competing access technologies are mobile wireless and fixed wireless. Mobile wireless based on GSM/GPRS and IS-95/3G-IX provides mobility and may account for 40–45% of new customers. Homes and offices in dense urban areas would be connected by fibre access systems, accounting for 40–45% of new connections. Homes and offices not served by FAS in dense urban areas and most homes in sub-urban areas, small towns and rural areas may be connected using fixed wireless systems. Both mobile wireless and fixed wireless systems, however, need high-speed back haul to connect base stations to base IEE Proc.-Circuits Devices Syst., Vol. 150, No. 6, December 2003

station controllers or radio control units. Fibre is the best medium to provide such a back haul. The role of fibre in the access network is therefore all pervasive. To conclude, fibre has already contributed to building high bit-rate and cost-effective backbone networks throughout the world. It is now all set to make its presence felt to a significant extent in access networks in dense urban areas of emerging economies. Fibre-to-the-building is becoming a commercial reality after a long wait. To what extent is the analysis and technologies presented here relevant to other countries? Developed countries are poorly wired up and the costs of laying fibre in these countries are very high. Also, dwellings in these countries are not as dense as in countries like India. Most of the analysis presented here would be inapplicable in such cases. But in most other developing countries, the situation is somewhat similar to that in India. First, the affordability analysis presented here is likely to be relevant in these countries and the growth in telecom in these countries would again depend on the CAPEX and OPEX involved. Second, most developing countries will have cities and towns, parts of which will be as densely populated as Chennai and Madurai in this article. In other words, most towns and cities of the developing countries will have parts where it would make sense to serve about twenty subscribers within 100 m of a fibre loop. While the costs involved in digging and laying fibre may vary, the business case and technologies presented here will be largely valid. 471

Fibre-to-the-building is likely to make a significant impact in most developing countries. It may enable these subscribers to have services comparable to the best available anywhere in the world. 7

8 9

References 10

1 Economic Survey 2002–2003, Department of Economic Affairs, Government of India http://indiabudget.nic.in/es2002-03/chapt2003/ tab97.pdf 2 Annual Report, 2002–2002, Department of Telecommunications, Government of India http://www.dotindia.com/annualreport/annualreport.pdf 3 Report of the Working Group on the Telecom Sector for the Tenth Five Year Plan 2002–2007, Department of Telecommunications, Ministry of Communications, Government of India, October 2001 4 ‘Telecom in India: Winning through flawless business execution, 2002–2003’, McKinsey and Company, February 2003 5 Jhunjhunwala, A.: ‘Looking beyond NTP ‘99’, J. Bitcom India, January 2000, 1, (1) 6 Agarwal, A.P.: ‘Affordability Analysis’. Proc. Workshop on 201 million connections in India by 2010 – Making a business case, NCC 2003, Chennai, India, February 2003, pp. 5–8 7 Jhunjhunwala, A.: ‘Innovative use of emerging technologies for rapid expansion of telecom networks’. Presented at First session on Access

472

11

12 13 14 15

technologies, Symposium on Telecommunications: Technology alternatives and policy, New Delhi, India, March 1996 Jhunjhunwala, A.: ‘Affordable fibre access network for developing countries’. Proc. OFC/IOOC’99, San Diego, USA, Feb. 1999, pp. 47–49 Jhunjhunwala, A.: ‘Role of Fibre Access Network for the Developing world’. Presented at Conf. on Optical fibre communication, Baltimore, USA, March 2000 Dravida, S., Gupta, D., Nanda, S., Rege, K., Strombosky, J., and Tandon, M.: ‘Broadband access over cable for next generation services’, IEEE Commun. Mag., 2002, 40, (8), pp. 116–125 Ooghe, S., De Clercq, J., De Voorde, I.G., T’joens, Y., and de Jaegher, J.: ‘Impact of the evolution of the metropolitan network on the DSL access architecture’, IEEE Commun. Mag., 2003, 41, pp. 140–145 Hao, F., Zegura, E.W., and Ammar, M.H.: ‘QoS routing for anycast communications: Motivation and an architecture for diffserv networks’, IEEE Commun. Mag., 2002, 40, pp. 48–56 Cohen, R.: ‘On the establishment of an access VPN in broadband access networks’, IEEE Commun. Mag., 2003, 41, pp. 156–163 Sargento, S., Valadas, R., Goncalves, J., and Sousa, H.: ‘IP based networks for broadband multimedia services’, IEEE Commun. Mag., 2003, 41, pp. 146–154 Nair, K.V.: ‘Fiber access network’. Proc. Workshop on 201 Million connections in India by 2010-Making a business case, NCC 2003, Chennai, India, Feb. 2003

IEE Proc.-Circuits Devices Syst., Vol. 150, No. 6, December 2003