Internet Exchange Points (IXPs) Explained

When you send data across the internet, it rarely travels in a straight line. Packets hop between networks operated by dozens of different organizations — ISPs, cloud providers, content delivery networks, and universities. The physical locations where many of these networks directly interconnect are called Internet Exchange Points, or IXPs. They are among the most important (and least publicly known) components of internet infrastructure.

What Is an IXP?

An Internet Exchange Point is a physical facility — usually a data center or a dedicated neutral building — where multiple networks connect to a shared switching fabric. Instead of each network needing dedicated links to every other network, all participants plug into a common Ethernet switch (or switch fabric). Traffic passes directly between any two connected networks at very low cost and with minimal latency.

Without IXPs, a data packet from a user in Frankfurt to a server in Amsterdam might travel across the Atlantic to a US-based transit provider before being routed back to the Netherlands. With a local IXP, that same packet can travel 60 kilometers in a few milliseconds.

Public Peering vs. Private Peering

There are two ways networks interconnect:

Public peering happens at an IXP. A network pays a membership fee and a port fee to connect to the IXP's shared fabric. Once connected, it can exchange traffic with any other IXP member that agrees to peer.

Private peering (also called a private network interconnect, or PNI) is a direct fiber cross-connect between two specific networks inside the same data center. Private peering is typically used when two networks exchange enormous volumes of traffic and want dedicated, guaranteed capacity without using the shared fabric.

Most large networks do both: they peer publicly at IXPs with hundreds of smaller networks, and they establish private peering links with their largest traffic partners.

How Peering Works at an IXP

The shared fabric at an IXP is usually a route server combined with a large Ethernet switch fabric. Here's the basic flow:

1. A network connects a router to the IXP's switch fabric via a port (1 GbE, 10 GbE, 100 GbE, or 400 GbE).
2. The router establishes a BGP (Border Gateway Protocol) session — either with the IXP's route server or bilaterally with each peer.
3. The network advertises its IP prefixes over BGP, telling peers "I can deliver traffic destined for these address ranges."
4. Peers that accept the advertisements install routes pointing to that network's IXP port.
5. Traffic flows directly between the two networks across the shared switch fabric.

The route server is an optional but popular shortcut: instead of configuring individual BGP sessions with every peer (which could be hundreds of networks), a participant only needs one session to the route server, which then redistributes routes from all participants.

Peering Policies

Not every network will peer with every other network. Each participant has a peering policy that defines who it will exchange traffic with and on what terms:

• Open peering — Will peer with anyone who asks, no minimum traffic requirements. Common among smaller networks and content providers trying to maximize connectivity.
• Selective peering — Requires traffic volume thresholds, geographic presence requirements, or technical standards (e.g., 24/7 NOC, redundant connections).
• Restrictive / closed peering — Only peers with specific organizations by agreement. Rare among IXP participants.

PeeringDB (peeringdb.com) is the canonical public registry where networks publish their peering policies, contact information, and IXP memberships.

Major IXPs Around the World

DE-CIX (Frankfurt, Germany)

DE-CIX Frankfurt is consistently the world's largest IXP by peak traffic, regularly exceeding 12 Tbps. Founded in 1995, it connects over 1,000 networks from 90+ countries. DE-CIX now operates exchange points in over 30 cities worldwide, including New York, Dubai, Mumbai, and Madrid.

AMS-IX (Amsterdam, Netherlands)

AMS-IX (Amsterdam Internet Exchange) is one of the oldest IXPs, founded in 1994. It connects 900+ members and regularly handles 6–8 Tbps of traffic. Amsterdam's position as a major internet hub reflects the Netherlands' historically strong network infrastructure and its role as a European business center.

LINX (London, UK)

LINX (London Internet Exchange) operates multiple switching platforms in London and regional exchanges across the UK. With over 900 members and peak traffic exceeding 6 Tbps, LINX is the dominant IXP in the United Kingdom and one of the most influential in Europe.

Other Notable IXPs

The Economics of Peering

IXPs save money for everyone. Consider the alternative: if two ISPs don't peer, traffic between their customers must travel through a transit provider — a network that charges per-megabit fees to carry traffic between networks. These transit fees can be significant at scale.

By peering at an IXP:

• Both ISPs avoid paying transit fees for traffic they exchange with each other.
• Latency drops because packets take a shorter physical path.
• Redundancy improves because traffic can take multiple paths.
• Local internet traffic stays local — important for performance and, in some jurisdictions, for data sovereignty.

The economic incentive for developing-world countries is especially strong. Without local IXPs, internet traffic between two users in, say, Lagos would often route through European or American transit providers, incurring both latency and cost. The establishment of IXPs in Africa, Southeast Asia, and Latin America has been one of the most effective levers for reducing the cost and improving the quality of local internet service.

IXP Infrastructure: What's Inside

A modern IXP runs a highly resilient switching fabric. Key components include:

• Core switches — Large chassis switches or aggregation-layer hardware (Juniper, Cisco, Arista, Nokia) with multiple terabits of switching capacity.
• Route servers — Typically running BIRD or OpenBGPd, these servers manage BGP sessions with all participants and distribute routing information.
• Looking glass — A publicly accessible web interface that lets anyone query the IXP's routing table, useful for diagnosing routing issues.
• Traffic statistics — IXPs publish real-time and historical traffic graphs, often in 5-minute resolution, giving a fascinating window into internet traffic patterns.
• Out-of-band management — Separate management networks and remote console access to recover from in-band failures.

Why IXPs Matter for Internet Health

A dense ecosystem of IXPs makes the internet more:

• Resilient — Multiple exchange points mean no single facility failure takes down a region's connectivity.
• Competitive — Access to cheap peering lowers the barrier to entry for new ISPs and content providers.
• Fast — Local traffic stays local, reducing round-trip times for users.
• Economical — Less transit cost means cheaper internet service for end users.

Countries and cities that have invested in neutral IXP infrastructure (especially carrier-neutral colocation facilities) have consistently developed stronger internet ecosystems than those that relied on a small number of monopoly carriers.