Tarun Dua

Buying corporate bandwidth in India is a royal pain. I am sure a lot of IT guys would agree with me.

Most common uses of bandwidth in corporate environment in India at development oriented companies is running the VPN with the US counterpart. Video/web conferencing, download of (free) softwares. Employees at new-media/web companies should have similar experience of Internet as compared to their target audience in the west/far-east otherwise being able to innovate in the webspace is out of question. So yes they need to use Youtube, MySpace and its ilk if the next killer application on the web is coming from India.

ISPs in India act as resellers of Bandwidth?
The ISPs in India instead of acting like network operators, and focussing on better peering with other network operators or reaching out to large content providers are content to be mere resellers of 'scarce' bandwidth to US. Infact most brochures of ISP(s) would read something like 'we have X amount of connectivity to the US'. The ISPs mostly say nothing about their connectivity with other Indian ISPs and datacenters. The way bandwidth is priced in India is the cost of International bandwidth plus amortization of heavy infrastructure sunk cost with atrociously high profits. The high profits are then divided between backbone providers, resellers and sub-resellers based on artificial management/sales structures needed to sell a product for price as high as possible.

Exactly how bad can it get ?
Try downloading a random large file to test the sustained transfer rate on a 2-4 Mbps 1:1(standard) leased line, from a host in the US. You might be able to download it at a sustained transfer rate far exceeding 2 Megabits per second from another host/ISP in the US. If you find sustained transfer rates exceeding 400 Kilobits per second in India with any ISP with the fattest possible pipe the ISP's will sell you, consider yourself lucky. This is because the interconnectivity of Indian ISPs to the other ISPs in the world is pathetic, since they only believe in buying transit and hardly peer with as many other ISPs as they can.

Apart from having bad networks, the ISPs in India actually run traffic shaping while mangling packets and disconnecting customer VPN connections as a ugly side effect. The http/ftp based user downloads are actually shaped into a decreasing throughput, even while the corporates are paying through their nose for this 1:1 bandwidth.

ISP SLAs are not enforceable ?
Most Indian ISPs don't have well defined procedures for planned maintenance of their infrastructure. Un-planned failures are normally denied by the level 1 staff or you are provided with standard explanations during downtime support calls. TRAI has guidelines for compensation applicable for the downtime for ISPs who follow it in violation. ISPs in India also depend on the practice of opening up the pipes when trying to sign up customers initially and the applying the traffic shapers and bandwidth managers on customer Internet "pipes".

Why is bandwidth expensive in India ?
Un-officially ISP officials routinely claim that due to transfer pricing for home users connectivity charges are higher for the corporate users. The home users are treated worse off than corporate users anyway, but the consolation is they are not paying 30 times the cost of bandwidth for corporate use.

The real reasons why bandwidth is expensive is:-

1. High cost of 'international' bandwidth due to under-utilization of fiber links( according to some estimates 90% of the fiber coming into India is dark),
2. The local bandwidth 'produced' locally in India is also sold for the same price causing anaemic network growth in terms of content/datacenter and makes even peering not useful.
3. Reseller mentality of cost plus models not taking into account the usefulness of growing the network.

How can it be fixed ?
Contrary to belief of bureaucrats/government, the answer to fixing this does not lie in having a fairly low-priced broadband connection from a government telcos where the waiting queue for a connection is way too long to be of any practical consequence but by
1. By fostering free-market competition by un-bundling the last mile.
2. Improvement of law and order for fostering competition in laying local cable networks by dealing with the muscle networks that run them in accordance with the law. Better policing against cable cutting when the cable networks "fight".
3. Better regulation of monopolistic backbone providers by un-bundling the backbones. So anyone should be able to buy dark-fiber transit to say closest backbone peering sites like Hongkong, Singapore, UK in Europe or US at the same price as backbone providers sell bandwidth to themselves.
4. Removing barriers to entry and making laws technology neutral(e.g. how stupid is a separate license for VPN ISPs or for ISPs carrying voice traffic). Instead of the current rent seeking regime promoted by current voice based telecom operators, the laws should be technology neutral. Government has no business in defining whether digital certificates are used for security or whether GnuGPGis used and what level of encryption is used or which vendors are allowed to issue certificates. Whether SIP should be used for VOIP or any competing platform can be used. Communities and companies who want to carry their own voice traffic for an integrated EPABX across their US/Europe counterparts should not have to worry about ambiguous laws.
5. Correct definition of 3rd party in IT parlance for fixing responsibility according to IT laws of the country and removing the burden of heavy regulation on ISPs/Cybercafe's, and occasional bouts of tyrannical suppression of free speech on Internet.
5a. Community Internet Links and Companies providing Internet to their employees should not be liable for misuse of Internet by their end-users without prejudice to their right to terminate the connectivity of their end-users at will.
6. Defining a minimum throughput of links for peering between different ISPs at any Peering point of presence at a minimum level say 10 or 40 Gbps and removal of constraints caused by complex billing requirements. Complex billing systems required to take care of bandwidth flow differential between different ISPs at Rs. 50 /GB(administered price) is detrimental to free market growth effects of peering in enlightened self-interest of ISPs.
7. Indian newpaper/TV media reports related to technology and Internet which in general are baseless/un-warranted, patently false, malicious and mis-informed should be objectively analyzed for factual incorrectness by a competent non-government body and appropriately addressed for the benefit of general public.
8. Provision for tax breaks to companies providing subsidized Internet based VPN connectivity to their office employees allowing them to work from home for at least 2 days a week. This also has the additional effect of reducing Indian dependence on Oil economy.
9. The value of Internet network will increase at an exponential rate to the number of users who are able to network with one another at a lower cost and a consequent increase in the number of Internet based services to serve the growing userbase. The 40TB plus internal dark fiber(countrywide backbones) can then be made operational and that should reduce the cost of 100 Mbit connectivity to near zilch as the component of 'expensive' International transit becomes a smaller relative to the overall bandwidth use in the country.

The opportunity
Consider the following hypothetical scenario.
Internet in India happens to be a multi-billion dollar opportunity for any fairly large investor who wants to enter this game.
One way this can play out is:-
A large investor to suddenly reduces the price of bandwidth to half for corporate buyers by buying off a huge slice of dark fiber and making it available to anyone who asks for it at a sliding price scale(based on increasing the number of subscribers) . This large investor can also buy Terabits of nation-wide dark fiber from about half a dozen or so backbone networks(apart from incumbent govt. telcos) in India that have laid redundant dark fiber to every nook and corner of the country. The last mile can be bought at currently high but negotiable prices from incumbent telcos providing city-wide networks(E1 or Metro Ethernet) even today to provide every corporate a 2Mbps pipe from home to office for each of their employee's within the town at less than Rs. 2000/- per month. Companies can be provided with 100Mbps aggregated bandwidth for connectivity to Internet available to their employees as well at nearly the same price as their current measly 2-4 Mbps contrained connections running this E1/Metro ethernet infrastructure. The new userbase this creates gets many more broadband users accessible to the new Investor who can easily connect to a datacenter created for hosting content with hithertho un-heard throughput of nearly 2Mbps per end-user. This datacenter itself is now creating the bandwidth for the large investor which is being sold to content providers as well as the end-users at a lower price but the cost of creation is near-zero. This network creation model can be replicated fairly quickly in all the large cities, leaving the incumbent ISPs who would very likely ignore it for a while before they realize what actually the new Investor did to be able to formulate a response for the same. These ISPs can then be offered free peering with the new bandwidth 'producer' for getting access to their subscriber base on the new Investor's terms. TV channels, other large 'portals' and content sites would now be dying to host themselves on this new network+datacenter+peering entity to get access to the combined subscriber base available through it. Thus Internet expansion in India will take place with a domino effect on prices of bandwidth and Indian ISPs will be able to get out their high cost International 'bandwidth reselling' trap.