ISPs: If It Happens in India…

To accelerate the penetration of Internet in India, or any developing country
for that matter, the cost of Internet connectivity has to be reduced drastically
while significantly improving the quality of service at the same time. One of
the most effective mechanisms to accomplish to establish Internet exchange
points (IXPs). An IXP interconnects ISPs within a region or country,
facilitating them to exchange domestic Internet traffic locally without having
to send them across multiple international hops to reach their destination.

Connecting to other ISPs

Any tier-2 ISP has to connect to a tier-1 ISP to get connected to the ‘global
Internet’. Due to the size of these tier-1 networks and their comprehensive
interconnection with other networks, they can send and deliver Internet traffic
to any network connected to the Internet. Other ISPs, in turn, have to pay their
upstream transit providers to deliver and receive this traffic.    

Therefore, if one ISP in a country like India needs to get connected to
another Indian ISP, he will have to take the long, circuitous route of going
through a network of tier-1 ISPs. This unnecessarily increases the cost, as both
the ISP would be paying transit fees. Moreover, the quality and speed of service
gets deteriorated. Obviously, all this could have been avoided if the two ISPs
were directly connected. There are, however many ISP networks in India and the
number is going to grow exponentially in future. So it won’t be
cost-effective, scalable or manageable for an ISP to connect to all other ISPs
in India.

An IXP is a single physical network infrastructure, often an Ethernet local
area network, to which many ISPs can connect. An ISP that’s connected to the
IXP can exchange traffic with all other ISPs connected to the IXP. Also, by
enabling traffic to take a more direct route between many ISP networks, an IXP
can improve the efficiency of the Internet, resulting in a better service for
the end user. Furthermore, since many networks have more than one connection to
the Internet, it is not unusual to find several routes to the same network
available at an IXP, thus providing a certain amount of redundancy.

Benefits accrued

Several benefits get accrued in the process of ISPs joining the IXP, which
can be further leveraged by ISP to provide superior and differentiated value to
the customer.

n Lower cost and reliance on
international transit

n Reduced latency

n Rapid deployment of peering
sessions

n Single point of contact

n Robust redundancy

n Flexibility and scalability

n Performance monitoring and
analysis

Technology and the Underlying Architecture: The network architectures adopted
by the IXPs are essentially of following types:

The complete mesh peering

n Exchange router

n Exchange switch

n Distributed exchange models

The most basic interconnect between ‘n’ number of ISP’s would be a
complete mesh with n(n-1)/2 connections. This though simple, does not provide
for scaling.

This may be simplified within a local exchange, with each provider drawing a
single circuit to a local exchange and executing interconnections at the
exchange. In this case, N point-to-point circuits are sufficient for complete
interconnection. The exchange router can also be used as an IXP with each of the
service providers connected to the exchange router. Scalable, full
interconnection is now possible.

The exchange router sets the preferred paths to each of the destinations and
control on routing is moved away from each ISP.

This may not be acceptable to all ISPs. For example, although the destination
may have multiple domain presence, the routing would be done through only one.
The exchange router is an active component and not player-neutral, which it
ideally should be. This disadvantage can be overcome by using an exchange switch
operating at layer 2 or by using an exchange LAN.

A policy-based route preference can be maintained on the local forwarding
table for routing to network destinations advertised by multiple peers. A
downside in this architecture is that imposed transit can occur when forced
routing occurs in the absence of a bilateral peering agreement in place between
the ISPs in the absence of any preventive mechanism.

IXPs have a wide variety of LAN architectures to choose from–Ethernet, FDDI
ring, switched FDDI hubs, fast Ethernet hubs, and switched fast Ethernet hubs.
Since IXPs are bound to have high traffic concentration, currently gigabit
Ethernet switches are used in such exchanges.

There are other models in existence, which use distributed exchange
architectures. In these cases, the exchange comes to the provider’s location.
These models trade off the costs of operation of a central location with the
enforcement costs of uniform access technology between every distributed
exchange participant. This also impacts the switching speed negatively. The
general cost structure of this model is unfavorable with higher costs being
incurred by both the provider and the exchange operator.

The Ideal Model for India

Ideally, the Indian IXP should opt for the third-party hosted co-location
model, which can handle high traffic volumes and also switch at high speed.

In this architecture, the exchange hub is a key point in the topology. The
central location must have the following aspects:

n Environmental Controls–ambient
temperature, uninterrupted power supply/power generation, fire suppression
controls

n Physical Security of Location–intrusion detection

n Multiple communication network
connectivity

n Support–24x7 onsite/offsite
monitoring, technical support

Further, DRP and BCP plans must be in place and periodically tested.

A suggested architecture for the Indian IXP, which is loosely based on other
common IXP architectures currently used across the world is given in the figure
below.

Possible Peering Structures in the Internet

Direct Point-to-point Peering: Here, each bilateral peer arrangement is
implemented through a physical interconnection. This, being highly unstable in
terms of the underlying routing and transit agreements, is rarely used.

Multilateral exchange points involve a single multilateral peering structure,
which is stable only when all parties involved have approximately equal mutual
traffic exchange.

Mutual Bilateral Exchange: The exchange points in this structure offer the
greatest degree of flexibility. A single physical ISP connection to an exchange
environment can be used to execute individual bilateral peer agreements with
some, or all, of the similarly connected ISPs.

ISP Peering Models

The Sender Keeps All (SKA) Structure: In SKA, peering traffic is exchanged
between ISPs without any mutual charge.

However, the marginal cost of international traffic transfer to and from the
rest of the Internet is significantly higher than the domestic traffic transfer.
Thus, it may be expected that any SKA peering would only relate to domestic
traffic and international traffic would either be provided by a separate
agreement or provided independently by each party.

The SKA model has stability problems where either one ISP acts in a transit
role for the other or provides a set of services used by the other ISP. In such
cases, there is little incentive for the service providing ISP to enter into an
SKA agreement with the second ISP, and a financial settlement may be the only
viable means of reaching a peering agreement.

Peering with Financial Statements: Financial settlement can be determined
through direct negotiation, traffic exchange levels or network routing
exchanges.

Traffic volume is the most obvious basis of financial settlement. However,
traffic volumes on the Internet are not as rigidly structured as those within
the telecom environment, and within this environment it is not possible to
readily determine which party’s client generated a bidirectional traffic flow
across the peering structure.

Establishing an IXP

Budgeting: The member ISPs should pay fees, ideally representing a
significant financial saving done on the cost of upstream transit by connecting
to the IXP. However, it could be difficult to achieve this in the beginning; so
the founding members may need to invest more than they save in the short term.
But if they build an IXP infrastructure that attracts new member ISPs fairly
quickly, all parties will see significant savings in the medium term and very
large savings in the longer term.

Outside investment: Investment could also be made by parties other than the
ISPs themselves. However, for an IXP with a neutral model, any commercial
investment should be very carefully considered to avoid losing that neutrality.
Government grants or assistance from local councils, technology funds, or
chambers of commerce may be used for this purpose.

Peering Model Suits India

The few ISPs in the Indian market are struggling in the face of low returns
from their businesses. However, if they can share the available infrastructure
and network resources, they can expect to increase the viability of their
business.

In the near future, an IXP based on multilateral peering agreement should be
set up. A consortium of member ISPs who pay annual membership fees sufficient to
cover the operating expenses could do this. This short-term policy would require
all participants to be peered with each other. However, it is required that no
explicit mutual fee be charged–to ensure that all participants operate on a
level platform and are guaranteed the same level of service.

In the long term, once the ISP market shows a greater degree of maturity,
transition to the mutual bilateral system is recommended. This would allow a
participating ISP to bypass the default multilateral agreement in favor of an
individual bilateral peer agreement. It would provide it the additional
flexibility of leveraging its pre-existing private peering agreements with the
local ISPs on a need basis.

However, in both the cases it is expected that the peering refers only to
domestic traffic and that international traffic be provided at the IXP level or
independently by each participant

Legal And Regulatory Issues

There are a number of regulatory issues facing the effective deployment of
IXPs, particularly in developing countries. Several parties have a stake in the
use of IXPs and consequently their own vested interests. The main obstacles
facing the development of an IXP would be:

n Resistance from existing
international traffic carriers (for example, VSNL). A centralized IXP that keeps
domestic traffic within the country would cause a significant cut in the
revenues of VSNL from other ISPs

n Resistance from existing ISPs.
With fierce competition on the price front, one ISP may not welcome the prospect
of making the cost of operation cheaper for other ISPs

n Government interference may
also be a potential barrier

Policy guidelines must be based on the feedback from ISPs as also the
existing best practices abroad. A number of Internet exchanges are in operation
internationally (examples are, Euro-IX in Europe and KIXP in Kenya). KIXP is a
good example of an IXP in a developing country. The following discussion draws
from the policy drafts of such IXPs currently in operation.

Guidelines for effective deployment of an IXP must be drafted in such a way
as to make ISPs feel comfortable about opening up their networks for other
players. The main regulatory and legal issues to be handled in the policy
document are:

n Ensure that no anti-competitive
peering takes place–between two strong ISPs–to the detriment of others

n The model should be limited to
domestic, inter-ISP traffic only

n Each ISP must agree to exchange
traffic with all other members of the Internet exchange. For this, developing
trust between ISPs is imperative

n A monitoring authority needs to be
set up to oversee functioning of the IXP and handle complaints and technical
flaws

The best model for an IXP seems to be in the form of a
nonprofit venture, as practiced in most cases globally. This creates a more
credible, neutral view of the organization and helps instill greater trust and
openness among ISPs. Most of the Internet exchanges globally are nonprofit.

India’s first IXP, INIX Public Peering Point, was rolled
out in the latter half of 2001. It was promoted by BandX, an international
bandwidth exchange company as a no-profit no-loss venture. Despite this, the
company had trouble getting FIPB approval. It got government clearance only
after clarifying that BandX itself would neither be an ISP nor build its own
network. It would simply act as a neutral point for the exchange of Internet
traffic in India.

IXPs holds the potential of accelerating the penetration of
Internet in India as they offer cost and quality advantages at the same time.
The underlying architecture, which seems to be the most suitable for India is
the exchange switch. Indian IXPs should opt for the third-party hosted
co-location model, which can handle high traffic volumes and also switch at high
speed.

Regulation is going to play a major role as several parties
with overlapping interests will use IXPs. Policy guidelines should facilitate
the process of implementation and be amenable for continuous improvement as the
number of ISPs grows in future.

Dipa Balakrishnan, Ruchika Gautam, Sachin Beny, Saurabh
Bansal, and Tejaswini Tilak under guidance of Prof V Sridhar, IIM, Lucknow