Summary

  • By the 2006–07 snapshot, TRAI counted 27 participating ISPs among 135 operational providers, 54 links across four NIXI nodes, and 1.9 Gbps exchanged per week in a later-reported October 2006 measure.
  • TRAI said a majority of domestic traffic still did not pass through NIXI. Node construction had created a place to peer; circuit economics, ASN access, port sizing and route policy determined whether it became a working domestic path.

The exchange was funded before it was adopted

In July 2003, an Indian business report described the Noida exchange as almost ready. The first phase was meant to cover four cities, and roughly twelve providers had signed up, according to Amitabh Singhal, then secretary of the Internet Service Providers Association of India (ISPAI). The report's verbs matter: “almost operational” and “signed up” describe a launch in progress, not a nationwide routing result.

The institutional design had taken shape earlier. A task force convened by the Telecom Regulatory Authority of India (TRAI) in 2002 recommended Internet exchange points under an industry-led, neutral, non-profit body. The Department of Information Technology would provide a one-time capital grant and space in four metropolitan areas. TRAI's 2006 consultation paper says the government approved financial support of about Rs 4 crore, after which NIXI was set up in 2003.

The intended function was narrow and practical: let one ISP exchange traffic with another inside India instead of sending that traffic out through international connectivity and bringing it back.

A later recollection by ISPAI secretary Amitabh Singhal dates administrative approval to March 2003, NIXI's incorporation to June and the Noida node's inauguration to 26 July. That account also remembers reluctance among some ISP members. It adds useful color to the launch, but it is a participant's retrospective, not a substitute for the contemporary report or TRAI's official record.

Four sites made a footprint, not a national mesh

TRAI's 2006–07 annual report says nodes were operating in Noida, Mumbai, Chennai and Kolkata. Across those sites, 27 ISPs had connected 54 links. A later TRAI consultation text gives a weekly volume of 1.9 Gbps across all four locations for October 2006. These measures describe actual use more closely than the 2003 signup count, yet each has a boundary: 54 links are not 54 providers or 54 bilateral peering relationships, and a weekly traffic volume is not a bandwidth rate.

In April 2007, TRAI supplied the blunt comparison: just 27 of 135 operational ISPs had joined NIXI nodes. It also said that a majority of domestic traffic still was not routed through the exchange. TRAI did not publish a national percentage for that majority, so the statement should remain qualitative. It is evidence that the authority considered the platform underused, not a traffic matrix from which to calculate the exact share that bypassed it.

The expensive distance between an ISP and a node

The barrier was partly geographic. Four metropolitan sites could not put a peering port within reach of every provider; stakeholders told TRAI that leased lines to a node cost too much and that nodes were absent from many state capitals. A provider's exchange decision therefore included the recurring cost of reaching the site, not just the exchange membership or port fee. The benefit also varied by operator: an ISP with substantial local destinations and usable peers had a different case from a small provider whose customers and upstream routes were concentrated elsewhere.

The barrier was also institutional and technical. TRAI's consultation record lists the cost of obtaining an autonomous system (AS) number among the obstacles. An ASN identifies a network in interdomain routing. It is one input to Border Gateway Protocol (BGP), the system that lets networks announce which destinations they can reach and choose among paths. Connecting a cable to an exchange does not create those announcements automatically.

TRAI noted that some connected ports were too small for expected traffic, and that connected ISPs did not always announce or accept all routes. In practical terms, a physical connection and a working peering relationship are separate states. If an ISP withholds routes, accepts only some destinations, or lacks capacity at its port, domestic traffic may continue to use an international upstream even while both networks have a live NIXI link. The exchange can provide the control surface; operators' routing policies and commercial incentives decide what runs across it.

This is why the 2007 report's combination of measures matters. Four nodes show that the buildout happened. Twenty-seven providers and 54 links show that participation was real but bounded. The weekly traffic figure indicates the network carried traffic, while TRAI's “majority” judgment says that the system had not captured most domestic flows. No single count is a verdict on NIXI's value. Together, they reveal the distance between funded capacity and network-wide adoption.

A coordination problem with local economics

The historical question is not whether India should have kept domestic traffic at home. TRAI's documents treat that as the purpose: lower dependence on international bandwidth, improve latency and reduce cost. The question is why opening a neutral exchange did not make participation self-executing. Circuit costs, geographic reach, ASN access and peering configuration each sat at a different layer. A government grant could start the institution and equip nodes; it could not make the economics of joining equally attractive to every ISP or compel complete route exchange through the switch.

Heng Lu's Note 64 offers a useful editorial test, not historical evidence: define the minimum common coordination that protects interoperability, then leave decisions that depend on local operating conditions close to operators. NIXI's record shows the practical tension. A shared exchange makes domestic interconnection possible, but the value of each connection depends on who else is present, which routes they accept, and what it costs to reach the node. The aggregate public benefit does not erase those private calculations.

The careful conclusion is narrower than either a triumph story or a failure story. India funded and built a four-node platform; operators did join it; the available official evidence says participation and traffic remained incomplete by 2006–07. The record shows an adoption gap between infrastructure and routing behavior. It does not show that every domestic packet should have crossed NIXI, that the nodes were useless, or that a single policy lever could close the gap.

Sources