Research and Advances
Computing Applications

Building India’s National Internet Backbone

The government wants it to link the country's huge rural population, as well as its own district offices and new private ISPs, to the Web and the world.
Posted
  1. Introduction
  2. Accelerate Net Growth and Penetration
  3. Access through the Phone Network
  4. Conclusion
  5. Author
  6. Figures
  7. Tables

India is a geographically vast country, located across the Arabian Sea from Saudi Arabia and Africa to the west and across the Bay of Bengal and the Indian Ocean from Myanmar, Malaysia, and the Indonesian archipelago to the east. Only about 27% of its 980 million people live in its cities; the rest live in the countryside in more than 600,000 villages.

Here, I describe the Indian Department of Telecommunications’ effort (begun July 1997) to address the country’s emerging but potentially enormous demand for Internet infrastructure. The government’s official policy is now to accelerate addition of Internet nodes for the department and for two state-owned companies by way of the Ministry of Communication—Videsh Sanchar Nigam Limited (VSNL) for international communication bandwidth for users in India, and Mahanagar Telephone Nigam Limited (MTNL) for telecom services in two big cities, Mumbai and New Delhi. The government is also allowing private ISPs to enter the market for the first time and to choose the national and international backbone networks they will use to offer commercial Internet services. The government expects this approach to help boost Indian Internet connections (or subscriber accounts) from the current (April 1999) total of 250,000 to more than five million within five years.

To achieve these ambitious goals, the government has set up a network of 549 Internet nodes, one in each of its district offices throughout the country. And it has authorized and funded the Department of Telecommunications to build a new Internet backbone to serve these nodes and take care of the private ISPs’ connectivity requirements.

The history of Indian telecommunications dates back to 1853, when a 33-kilometer experimental telegraph line was built between Calcutta, then the capital of Imperial India, and Diamond Harbour, the anchorage of the British East India Company. Telephone service began in 1882, when small exchanges were opened in Bombay, Calcutta, and Madras. Service was gradually extended to other areas, though progress was slow. At the time of its independence from British colonial rule in 1947, India had only a rudimentary telecommunications system consisting of about 403 telephone exchanges with a capacity of 91,424 lines. Today, however, India operates the tenth largest telecom network in the world in terms of numbers of users, number of lines, geographical coverage, and traffic volume (see Table 1).

As of April 1999, the country had 2 telephones per 100 people. Along with the new Internet backbone, the government now intends to provide telephone on demand, including coverage of all villages, by 2002 and increase teledensity to 7 per 100 by 2005 and 15 per 100 by 2010. In rural areas, it intends to increase teledensity from the current 0.4 per 100 to 4 per 100 by 2010.

The existing long-haul backbone telecommunications network is mostly digital, employing both fiber-optic and microwave links. The telephone exchanges are largely electronic, and the various local loops are copper-based with extensive remote-line units to reduce the need for copper lengths. The Department of Telecommunications and VSNL already maintain 54 Internet nodes. Today, international gateways, provided by VSNL exclusively, are available at six nodes in six major cities—Calcutta, Chennai, Bangalore, Mumbai, New Delhi, and Pune. These nodes provide service to local subscribers and act as international interconnection points for Internet service. Most of the Department of Telecommunications’ nodes have remote access servers connected to the nearest international gateway nodes through 2Mb digital trunks. The various servers available at individual nodes provide Internet services. All 54 Internet nodes provide customers dial-up TCP/IP and shell accounts, or Internet access for nongraphical applications. Leased-line and ISDN access is provided to only a limited number of customers in selected cities. Of these various accounts, about 50% are businesses across the country.

Persistent complaints from business and private users alike about the performance of existing Internet services finally prompted the government in 1997 to begin planning the new backbone. Today, although connectivity for any given station is through a 2Mbps star network connection (through VSNL via a single link), the links lack redundancy. Moreover, because the network has not been centrally managed, traffic overloads and failures have been common. The new backbone architecture seeks to eliminate such problems.

Most of the complaints focused on the difficulty of getting access, slow response times, and unstable connections. Another was the occasional lack of availability of Internet accounts due to capacity constraints. Such complaints are often prompted by congestion in either international or national access segments, inadequate dial-up ports, or deficiencies in point-of-presence (POP) equipment and lack of synchronization in the local switches connected to a POP.

Back to Top

Accelerate Net Growth and Penetration

The government’s official policy is to accelerate the growth, penetration, and geographical availability of Internet services. In addition to the nodes being set up by the Department of Telecommunications, along with VSNL and MTNL, the government has issued licenses to approximately 100 private ISPs, and another large number is being reviewed for licenses, though few are online as yet.

For the private ISPs, there are three license categories covering various geographical areas: Type A, covering the entire country; Type B, covering individual telecom circles (the country has 20); and Type C, covering individual secondary switching areas.

The new backbone’s architecture will follow a router-based design and will eventually involve greater use of asynchronous transfer mode (ATM) equipment. The equipment and the network design are being based on highly reliable carrier-grade solutions with a high degree of redundancy. The network’s architects in the Department of Telecommunications are keeping expansion and scalability in mind to make future upgrades as easy as possible.

The architecture calls for centralized servers at a few places and distributed POPs at the 549 nodes. The architecture also provides sufficient scope for easy expansion and decentralized operation. Most popular Internet applications, such as email and Web browsing, will be supported with capacity enough for later introduction of value-added services, such as online stock trading. Roaming the Internet within the country will be possible without users having to make any change in their access methods.

Backbone-related equipment will consist of routers (or optional frame-relay switches), remote access servers, LAN switches, application servers, a network management system, a help desk, access devices, and a billing system. The new backbone, which could be operational by next year, will be a widely available access network designed to provide easily accessible Internet access points for the Department of Telecommunications’ own ISPs as well as for the licensed private ISPs, allowing them to connect their Internet POPs to the global Internet.

The architecture will permit aggregation and connectivity on three different tiers:

  • Cities and large towns (Type A stations) connected through high-capacity links to the gateways (first tier). The government has identified a total of 14 Type A stations, all slated for connection in the first construction phase, scheduled for completion this year. The stations with international gateway connectivity are called Type A1, the others are Type A2.
  • Midsized towns (Type B stations) connected to the first tier through medium-capacity links (second tier). In the new backbone’s first construction phase, 31 Type B stations are slated for connection.
  • All government district offices and some small towns (Type C stations) aggregated around the Type B stations (third tier). The construction plan calls for Type C stations to be connected to the new backbone in the second phase, scheduled for completion next year. Locations for as many as 504 Type C stations have been specified. Depending on the number of users they support, Type C stations are labeled C1, C2, and C3. Equipment requirements for different Type C stations have been drawn up.

The plan calls for connecting as many as 549 stations through the new backbone. In the first construction phase, 45 stations will be connected; in the second, all C stations will be connected (see Table 2).

The network’s architecture will consist of core stations (the Type A1 locations) connected on a mesh topology (see Figure 1). The core backbone will provide 34 E3 communication links to start with and then be converted to STM-1 (155Mbps) links once the ATM and synchronous digital hierarchy (SDH) data transmission-based networks are in place. Type A2 stations will be connected to two Type A1 stations (each with 8Mbps capacity) and will be converted to E3 links in the second construction phase. Type B stations will be connected to A1 or A2 stations by 8Mbps links. Type C stations will be connected to Type A1, A2, or B stations, depending on geographical proximity through two communication links of 2Mbps. All stations will have redundant connectivity to the backbone in anticipation of potential link failure as well as in light of how congestion degrades network performance.


The government expects to boost Indian Internet subscriber accounts from the current total of 250,000 to more than five million within five years.


The new backbone will further support the Department of Telecommunications’ own ISP operations by including ISP functionality at its Internet access points. The backbone will eventually provide the following services to its users:

  • Connectivity to the Internet
  • Internet customer accounts, including shell, TCP/IP, dial-up, leased-line, and virtual private networks
  • Internet services, including email, FTP, Telnet remote login, online discussion groups, and information search
  • Network management and billing functions
  • Automatic roaming throughout India
  • Homing for email users

The network will also be flexible enough for its administrators to add features without further investment in infrastructure, including:

  • A scalable security architecture as a framework for the backbone’s formal security policy
  • A global roaming facility
  • Distributed and centralized billing
  • A centralized network monitoring system (NMS)
  • Proxy servers for caching (to minimize dependence on international bandwidth) and for implementing security with the help of an access list of authorized subscribers
  • Free home pages for all customers
  • Centralized authentication, authorization, and accounting

Back to Top

Access through the Phone Network

For the Department of Telecommunications’ own future Internet customers, the backbone will support Internet access through the public switched telephone network; dial-up access to the Internet at port speeds up to 56Kbps; digital signal processor (DSP) modems at 56Kbps, by way of a software upgrade; ISDN dial-up at basic rate interfaces; X.25 access through the packet-switched public data network, commonly known as Inet in India; frame-relay and ATM access; and leased-line Internet access through telecom links in any multiple of 64Kbps bandwidth, using appropriate modems.

Private ISPs will connect to the backbone through leased-line access in any multiple of 64Kbps bandwidth, using appropriate modems. They will be able to use the backbone for connectivity or to reach other networks, such as the railways’ and the power grid’s own national networks, or may set up their own licensed links for Internet access. Meanwhile, the Telecom Engineering Centre in New Delhi, which is the Department of Telecommunications’ standard-setting body, finalized (in February 1998) new backbone and interface requirements for remote access servers to be used when a private ISP connects to the new backbone.

In preparation for the first implementation phase, scheduled for completion in December, the servers and other equipment at Type A and Type B stations were specified in the network’s architecture, the network connectivity diagrams were worked out, and a formal request for IP-address allocations was made to the Asia Pacific Network Information Center (headquartered in Australia), a regional organization providing allocation and registration services for participating in the global Internet. Separate servers are being planned for different applications and placed in the backbone in accordance with the network’s overall architecture. The network architecture thus incorporates the concept of putting centralized servers in places where skilled technical assistance is readily available. Login and caching servers are planned only at the 14 Type A stations in the first construction phase; their place in the second phase at Type C1, C2, and C3 stations has been worked out and released (September 1998) as Amendment 2 of the Telecom Engineering Center’s general requirements (as promulgated by the Department of Telecommunications) for the backbone. Plans were also drawn up for upgrading the servers and routers required at Type A and Type B stations to service the load created by the Type C1, C2, and C3 stations.

The NMS will ultimately represent a two-tier structure: The Mumbai and New Delhi stations will function as the national NMS center, managing not only the nodes, ports, and devices connected in the geographical areas of these cities, but also the national backbone’s core connectivity links. A second NMS is planned for each telecom circle (20 throughout the country) to act as middleware for managing the devices, links, and equipment in each circle, collecting data and passing it on to the centralized NMS in Mumbai and New Delhi. The combined NMS will be based on the simple network management protocol (SNMP), a standard Internet protocol used around the world, to monitor the network’s elements and control network traffic.

The server requirements for the backbone’s first construction phase have been worked out to support a customer capacity of about 300,000 users. The dial-up customer base for remote access servers at Type A nodes will involve 480 DSP modems on 16 E1 trunks for a projected customer base of 4,800; the B nodes will involve 120 DSP modems on four E1 lines for a projected customer base of 1,200. This scheme assumes that only 10% of the customers will be online at any given time. Application servers may be expanded later depending on customer need.

To support TCP/IP on Ethernet on the local network side, the backbone will employ the following equipment:

  • Routers or frame-relay switches to route customer traffic
  • Remote access servers to accept user phone calls
  • LANs to connect ISP equipment
  • Application servers for user authentication, domain name service, service management, and various Internet services
  • The NMS and application workstations
  • The billing system
  • The call management system
  • The help desk
  • Access devices, such as line drivers, and modems
  • A testbed to test new features and functions
  • Firewall servers
  • Uninterruptible power systems, inverters, unshielded twisted pair, untwisted pair, cables, and more

The backbone will thus involve a tiered architecture and a core layer; carrier-class equipment will provide highly redundant features and operation. Because scalability and expandability are two main factors in designing the backbone, the design emphasizes network connectivity with link redundancy and an NMS through SNMP. The power supply will be -48 volts DC, so the backbone’s equipment can run on the exchange battery supply, as needed. Vendors unable to support this power requirement are required to supply inverters that convert -48-volt DC to 220-volt AC power. All equipment also has to be hot-swappable, support the architecture’s redundancy specifications, and be installable on racks or chassis assemblies. Moreover, the plan calls for implementation on IP v.4, though support for IP v.6 has been included in all network and server equipment, so a migration path to IP v.6 will be available soon.

The network structure will also support a formal security plan, to be developed by the government by the end of the year. For example, the firewall architecture will define an overall, integrated security policy distributed across multiple firewall gateways and managed remotely from a central place where the NMS is available. The architecture will support central integration, configuration, and management of the firewall, as well as of other third-party security applications.

The firewall architecture will divide the network into three separate zones, or subnetworks:

Secure. Only authorized and authenticated personnel will be permitted into this zone. DNS, NMS, and application workstations and billing servers will be here.

Demilitarized. This “semiprotected” zone (the perimeter network) will grant access only to users who are checked and authenticated. Application servers, such as the World-Wide Web, caching for Internet content, remote authentication of dial-in users, and email, will be allowed here.

Open. This zone will include remote access servers, routers, and login servers.

ISP nodes will be connected to the Department of Telecommunications’ public switched telephone exchange through analog subscriber lines or junction lines (at 64Kbps or any multiple of 64kbps up to 30 lines). In some cases, subscriber connections to ISPs may be through leased lines, but because they won’t pass through telephone exchanges, no dialing will be required for accessing an ISP node. Traditional telephone subscribers will be able to access ISP nodes on a local-call basis, if an ISP node is available in the same local area. In all other cases, subscribers will have to make long-distance calls to access an ISP node. They will dial the access code + ISP code, or 172 + xxx, where xxx = 100 to 999; these numbers will be centrally allocated to the ISP.

The new backbone will be connected to other networks through its Internet exchange points—initially only to the VSNL and MTNL networks. An Internet exchange can be set up only when a large private ISP wants to be interconnected to the new backbone.

In India today, all international connectivity is through VSNL, and any link with the Asia-Pacific region is set up by way of mutual arrangements between two countries or ISPs.

Under its telecommunications policy, the government still does not permit Internet telephony, though it recently stated it will continue to monitor the related technological innovations and their effect on national economic development, returning to that policy as needed.

Back to Top

Conclusion

Along with the rest of the world, India is experiencing exploding interest in Internet communications, in part because of the dramatic liberalization of the government’s policy toward Internet connectivity. The Department of Telecommunications is now playing a major role in delivering sufficient infrastructure to private Internet entrepreneurs, as it helps maintain the government’s basic monopoly on the Indian telecommunications infrastructure. It is also committed to providing Internet access and services to the country’s vast rural areas where most of the population lives.

Back to Top

Back to Top

Figures

F1 Figure 1. Connectivity and locations in the new national Internet backbone.

Back to Top

Tables

T1 Table 1. Elements of the Indian telecommunications infrastructure.

T2 Table 2. Distribution of backbone node types.

Back to top

Join the Discussion (0)

Become a Member or Sign In to Post a Comment

The Latest from CACM

Shape the Future of Computing

ACM encourages its members to take a direct hand in shaping the future of the association. There are more ways than ever to get involved.

Get Involved

Communications of the ACM (CACM) is now a fully Open Access publication.

By opening CACM to the world, we hope to increase engagement among the broader computer science community and encourage non-members to discover the rich resources ACM has to offer.

Learn More