. : : Marc Sez : : .

1. SEA-ME-WE-3 (VSNL) – total capacity available to VSNL is a few Gbps. Much of this is already used up. This cable connects India to Western Europe, Middle East and East Asia (inc Japan). This used to be the old workhorse.

2. FLAG - Intercontinental, but smaller footprint than SMW3. Owned by Reliance, but by a quirk of history, VSNL owns the access points out of India. Needs the two to tango, which they often don’t. Capacity available on this one.

3. i2i – Whopping 8400 Gbps design capacity. Over 1400 Gbps immediately available. It connects to Singapore, which is a hub for many international cables. This one cable alone debunks any complaints of bandwidth ‘scarcity’. The catch, it is wholly owned by Bharti and SingTel, its partner.

4.SAFE-SAT-3 – VSNL’s alternate route to Europe via Africa, and East Asia (via Malaysia). Lots of capacity Available.

5. Tata Indicom – VSNL, connects to Singapore (like i2i), and has 5400 Gbps capacity. Launched, but awaiting security clearances from Govt of India

6. Coming Soon – SEA-ME-WE-4; VSNL & Bharti – next generation in the SEA-ME-WE family.

As you can see, all three of India’s telecom giants (Tata, Reliance and Bharti) own independent access to international capacity. There is a humongous amount of capacity available, waiting to be used, under the sea.

This kind of capacity is not matched by the quality of the “last-mile”, the stretch from the exchange to homes and offices. In many cities, the legacy BSNL/MTNL lines are poor quality and incapable of supporting more than a few Mbps speeds. Older Cable TV networks similarly suffer from quality problems.

New lines laid by Bharti, Tata, BSNL and others do allow higher speed access, but service offerings are tailored for the mass market which suffers from legacy issues.

In my mind, there are two reasons why higher speed broadband services are not offered in India:

a) Operators have found a ‘comfort spot’ on pricing international bandwidth; with no one ready to fight each other in this space.

b) Last-mile bottleneck creates reluctance on the part of operators to offer higher-speed services

c) India’s internet penetration is miserably low. Looking at market growth rates, operators may be taking a very conservative world view. Hence high prices for low bandwidth. For this change, PC penetration must take off sharply.

What is the cost structure of the connectivity business – I think there are several aspects to this. Let us split this up into its components – connectivity from the user to the service provider, operations at the service provider, connectivity from the service provider to the internet and the speed component. Let us examine each component separately.

Connectivity from the consumer to the provider is piggybacked on phone or cable lines (when talking about DSL or cable modem). In the US at least, there is a redundant and robust networking infrastructure to provide telephone and cable lines to every end consumer. In India, while there is a telephone infrastructure, it isn’t comparably robust
(with cables hanging all around, roads being dug so frequently, improper drainage, lines are easily susceptible to picking up noise and disruption). When a network service is piggybacked on the existing
infrastructure, it adds some cost. I suspect that the incremental cost with a robust infrastructure is almost zero, while the incremental cost in India is non-zero.

The second aspect of connectivity is operational costs at the service provider (essentially efficient routing and rate limiting). This is purely a function of technology and efficiency of operation. Its hard for me to compare these two between India and the US.

The third aspect of connectivity cost is pipes and peering. The cost of available pipes is a function of whether there exists physical pipes, and is essentially about supply/demand. The more points of presence you have on the network (ie the more you peer with other service providers), the better your connectivity. I think someone has listed the pipes out of India though has not spoken about peering. My knowledge is limited to VSNL which has one fat pipe to New York (through Europe but with no peering in Europe). Its expensive to peer
with providers in developed countries. Costs are a function of 95% percentile of peak traffic as opposed to average traffic. I’m not exactly sure of why it is hard to have good pipes connecting India to the ROW and to increase peering with US network providers. I suspect there are more aspects to it than simply supply/demand.

The term speed, I think is misrepresented. There are two aspects to it. The download speed (ie the speed between the last hop and your machine) which is what is advertised by providers, and perceived speed which is also dependant on how good the service providers connectivity is to the ROW. This aspect of it is never discussed when advertising about network connectivity. The latter is what determines user experienced speeds while the former is what is discussed everywhere. I think one would require to do a detailed technical analysis to understand requirements to improve connectivity speed from India to
the ROW.

All that said, I don’t think network connectivity is expensive based on the example you gave. Your comparison is of the Bay Area vs Pune. I think the average cost of a 384 kbps connection in the Bay Area is
around $50 a month (as also pointed out by a couple of others). I myself was paying Comcast around $45 pm on a discounted scheme. The $20 and $30 schemes are essentially all promotional schemes which last anywhere from 3 months to a year. The cost of network connectivity in Pune, for a 256 kbps line, BSNL’s current charges are Rs 250 a month (was Rs 500 a month upto last month). I think Iqara has similar
charges (somewhere around Rs 300 a month). While the download speeds can be compared (256 to 384) perceived speed, uptime of the service, value added services as well as restrictions are different in both cases (like service providers in India impose download limits on a monthly basis), making the complete user experience in both cases very different. (and this is probably what matters more)