Executive Summary
The benefits of maximising local traffic via independent Internet exchange points (IXPs) is well-recognised as essential for facilitating a robust domestic Information and Communication Technology (ICT) sector. From a public policy perspective, ensuring the presence of local IXPs has become an increasingly important priority in order to make sure that online services are equally accessible to all local users, as well as to enhance competitive opportunities, and generally
So far IXPs have only emerged in about half the countries in the world, and these vary greatly in scale and effectiveness. To help accelerate the development of new and existing IXPs this Toolkit has been created to describe best practices for setting up and supporting the growth and enhancement of these crucial Internet facilities.
Ideally, IXPs are needed in every region in which different networks need to exchange local traffic. Deployment of IXPs are, however, sensitive to a variety of local constraints, and initiating and ensuring their efficient operation is not as simple as it would appear, especially in emerging markets (where IXPs are rare). Nevertheless IXPs are not a universal solution to a country’s Internet and Internet access challenges. IXPs can and often do complement and improve the functioning of a local Internet ecosystem, but they cannot gloss over problems such as lack of competitively priced international or local capacity, nontransparent regulation, or poor energy supplies.
Research into the experiences of IXPs around the world was compiled for this report and the variety of case studies and data about IXPs presented in Section 6 amply demonstrate that: IP interconnection is still relatively new and there are a wide variety of fees, institutional models, business models, policy rules, and technical strategies adopted by IXPs across the world. Some of the choices made may have constrained growth in interconnection; in other cases, alternative IXPs have emerged to fulfil needs unmet by the existing IXPs.
As a perusal of the case studies will show, IXPs vary immensely in scale (from a few 100Kbps of traffic to many Tbps), in pricing and in institutional models — from free to use, to nonprofit cost-recovery, to for-profit. There are few geographic trends that can be deduced aside from the commercial/noncommercial divide between the US and the rest of the world; but even this is now blurring with three European IXPs recently launching neutral membership-based services in the United States.
Nevertheless, in broad terms, three main models for operating an IXP have emerged:
1. the for-profit carrier neutral data centre as typified in the United States;
2. the neutral nonprofit member-owned organisation operating on a cost recovery basis, with infrastructure often hosted at a commercial data centre; and
3. the sponsored IXP, supported either by a ccTLD manager, a regulator, an NREN, or a large network operator.
In the last two years, there has been a notable surge in the number of IXPs in secondary cities, particularly in Argentina, Brazil and Indonesia, but also in secondary cities around the world, including Arusha, Adelaide, Buffalo, Cork, Durban, Edinburgh, Grenoble, Leeds, Lyon, Manchester, Manitoba, Mombasa, Port Harcourt, Saint Etienne, Toulouse, Turin, Winnipeg, and Zurich.
This trend reflects increasing local content consumption, decentralisation of content redistribution, and overall growth in bandwidth demand built on the steady extension of high bandwidth cable and wireless networks. While most of this growth has so far been in more developed economies, the same trends are becoming evident in emerging economies.In addition, aggregating outbound traffic and avoiding tromboning is likely to be more critical in smaller secondary city markets where local ISPs typically face higher transit costs and longer routes to the desired content.
At the same time, the scale, reliability, and geographic scope of existing IXPs is growing. Many IXPs now offer multiple sites, remote peering, and ‘partnership programmes,’ often called service-provider or reseller plans. Such programmes leverage the benefits of the remote peering model and low cost national or regional backhaul, minimising technical support needs for the IXP and taking advantage of link aggregation.
Regional extension of networks is also being encouraged in countries where the IXP may operate its own links to a neighbouring city or country. In France, members of France-IX may freely use up to 100Mbps of connectivity between Paris and Lyon, Toulouse, Luxembourg, and Italy, after which they need to purchase their own links.
Global expansion of IXP presence is also a noteworthy recent trend. For example, Dutch IXP AMS-IX now operates an IXP in Curacao and Hong Kong, and is collaborating with KIXP to manage the Mombasa exchange in Kenya, while DE-CIX operates the UAE-IX, the Dubai exchange and DE-CIX New York. While increasing economies of scale and attracting new members are some of the motivations for this, demand for the skills and expertise developed at leading exchanges is another
It is also noteworthy that policies that promote multilateral peering are present among a significant number of IXPs on either mandatory terms or incentivised through discounts on the port fee for the invited party. The majority of IXPs, however, also offer bilateral peering and VLAN services and of late, a few are beginning to offer VoIP or GRX type services.
It is also noteworthy that policies that promote multilateral peering are present among a significant number of IXPs on either mandatory terms or incentivised through discounts on the port fee for the invited party. The majority of IXPs, however, also offer bilateral peering and VLAN services and of late, a few are beginning to offer VoIP or GRX type services.
Many IXPs also host regular social, technical, or industry events to help build the local community of people involved in peering and add another membership benefit. Twinning programmes (programmes in which experienced IXPs partner with developing IXPs) to support emerging IXPs have alsobeen adopted by some of the larger exchanges such as those in London, Amsterdam, and Stockholm. Some IXPs also have created positions for policy staff in order to inform and educate local and global policymakers.
A significant number of IXPs are still operated without charge; however, the majority of IXPs have pricing for participation ranging from simple joining free to charges almost equalling the cost of transit. We have seen that there is great variability in fees, especially for smaller emerging country IXPs, many of which may have donated space and equipment, and so are able to minimise fees to attract members before moving on to achieving a cost recovery position.
The most important variable in IXP pricing is port speed. This may need to be balanced against membership fees (if any) or setup fees (if any) as well as the backhaul costs of getting to the IXP and the availability of link aggregation and discounts for second ports to allow smoothing the costs as network needs grow (for example, a network needing 1.2Gbps could cost the same as 2X1Gbps ports or 1x10Gbps port without link aggregation).
In analysing the current costs for use of IXP services, 1Gbps and 10 Gbps ports are the most commonly available. A minority of IXPs have 100Gbps services and below 1Gbps, ports may not be available or may even be free. (See page 27, figure 4.6, for the annualised port cost for 1 and 10Gbps ports at a variety of IXPs in different locations around the world.) There is greater variability and inconsistency in charges for 10Gbps ports.
The case studies and data samples provided in this report draw on information from a variety of sources, including the IXP websites, national ICT market profiles, and personal interviews for this study with IXP managers. In the course of gathering this data, researchers found little consistency in the presentation of basic information on IXP websites. Few IXP websites in emerging markets provide the three main data points: pricing, membership policies, and peers lists. Traffic statistics are also missing from many sites while some may show disaggregated data with the traffic history of each network connected to the exchange. In other cases, information may be buried in a hard to find web link or may not be current. Overall, only a small minority of IXPs operate websites that fulfil the basic requirements of a prospective peer for up-to-date, easily accessible information.
Further case study information and additional materials can be found on the IXP Toolkit portal at http://www.IXPToolkit.org.
* Note about releasing this “Collaboration Draft” of the Toolkit and Portal. We plan to enhance and amplify many areas of the Toolkit in April, when we release v.1.0 of the Toolkit. Thus, the release of the IXP Toolkit & Best Practices Guide, now, as a “Collaboration Draft.” More input from the community is essential. The Toolkit Portal (www.ixptoolkit.org) is meant to be a more detailed resource, and we also are asking for continuous feedback and information from the community. Feedback about the Toolkit and Portal can be sent to feedback@ixptoolkit.org.
