Internet Exchange Points

Imagine two neighbors who can only communicate by mailing letters to a distant city, where a postal service forwards them back. Absurd for neighbors—yet this is exactly how Internet traffic worked before exchange points existed. Two networks in the same building might route traffic through another continent because that's where their transit providers connected.

Internet exchange points fix this absurdity. They're neutral meeting grounds where networks connect directly, transforming the Internet from a hierarchy into a mesh.

The Problem IXPs Solve

Without IXPs, the Internet would be purely hierarchical. Your local ISP buys transit from a larger provider. That provider buys transit from an even larger one. Traffic climbs up this ladder, crosses someone else's backbone, and climbs back down to the destination.

This hierarchy creates three problems:

Cost: Every rung of the ladder charges fees. Both the sender's network and receiver's network pay their transit providers for every bit exchanged.

Latency: Traffic takes detours. Two networks in Amsterdam might exchange traffic through Frankfurt because that's where their transit providers interconnect.

Fragility: The hierarchy creates chokepoints. If a major transit provider has problems, traffic between thousands of networks is affected.

IXPs solve all three by letting networks bypass the hierarchy entirely.

What an IXP Actually Is

An IXP is a physical location—typically inside a carrier-neutral data center—where networks connect their routers to shared switching infrastructure. Think of it as a meeting hall where anyone can show up, plug in, and start conversations.

The IXP provides:

• High-capacity Ethernet switches (the shared infrastructure)
• Physical ports for each participant (1G, 10G, 100G, or 400G connections)
• A shared network segment where all participants can reach each other
• Often a route server that simplifies establishing connections

Networks pay for their port—a monthly fee based on capacity—and gain access to every other network at that IXP.

How Peering Works

Connecting to an IXP is just the first step. Networks then establish peering relationships—agreements to exchange traffic directly.

Bilateral peering: Two networks negotiate directly, configure BGP sessions between their routers, and begin exchanging routes. This requires mutual agreement and individual configuration for each relationship.

Multilateral peering via route servers: The IXP operates a route server that acts as a BGP intermediary. Networks configure one session to the route server and automatically peer with everyone else using it. Instead of configuring 200 individual BGP sessions, you configure one.

Route servers are why IXPs scale. A new participant can connect, configure a single BGP session, and immediately exchange traffic with hundreds of networks.

The Traffic Flow

Once peering is established, traffic flows directly across the IXP fabric:

1. A user on Network A requests content hosted on Network B
2. Network A's router (connected to the IXP) has learned Network B's routes via peering
3. Traffic crosses the IXP switches directly to Network B's router
4. Network B delivers the content

No transit providers involved. No climbing the hierarchy. The traffic stays local, takes fewer hops, and arrives faster.

Why This Matters

Cost disappears for peered traffic. Transit bandwidth is expensive—networks pay per megabit. Peered traffic at an IXP costs only the fixed port fee, regardless of volume. For networks exchanging terabits of traffic, the savings are enormous.

Latency drops dramatically. Traffic exchanged at a local IXP might traverse a few hundred meters of fiber. The same traffic routed through transit providers might cross countries. The difference is milliseconds—which matters for video calls, gaming, and financial trading.

Resilience improves through redundancy. A network with peering relationships at multiple IXPs has multiple paths to reach other networks. If one IXP has problems, traffic flows through alternatives. If a transit provider fails, peered traffic is unaffected.

Control shifts to the network. Instead of depending entirely on transit providers' routing decisions, networks can establish direct relationships and implement their own policies.

The Major Exchanges

IXPs exist worldwide, from small regional exchanges to massive global hubs:

DE-CIX Frankfurt handles over 10 terabits per second of peak traffic. Hundreds of networks connect here, making it one of the world's largest interconnection points.

AMS-IX Amsterdam has been central to European interconnection for decades, serving hundreds of networks with multi-terabit capacity.

LINX London connects hundreds of networks and serves as a major hub for European and transatlantic traffic.

Equinix exchanges operate in major cities worldwide—New York, Chicago, Silicon Valley, Tokyo, Singapore—providing global interconnection.

But the magic of IXPs isn't just the giants. Regional exchanges in hundreds of cities enable local traffic to stay local. A city with a healthy IXP keeps its internal Internet traffic from unnecessarily crossing borders.

The Ecosystem Effects

IXPs benefit more than just the networks that connect to them.

Local traffic stays local. Without an IXP, traffic between two networks in Lagos might route through London. With a local IXP, it stays in Lagos. This reduces latency for users and decreases demand on expensive international links.

Barriers to entry drop. A startup ISP can connect to an IXP and immediately peer with established networks. Without IXPs, new entrants would need to negotiate individual relationships or purchase expensive transit.

Developing regions gain independence. Countries with strong local IXPs reduce dependence on international connectivity. Local content providers can serve local users efficiently instead of routing traffic through distant countries.

The Complications

Peering politics: Not everyone peers with everyone. Large networks often have restrictive policies, refusing to peer with smaller networks or requiring specific traffic ratios. Being at an IXP doesn't guarantee peering—it only provides the opportunity.

Geographic constraints: You must be physically present in the IXP's facility (or pay for a circuit extension). If you're not in a city with an IXP, the benefits are harder to access.

Security considerations: IXPs are shared Layer 2 infrastructure. ARP spoofing, MAC flooding, and other attacks are possible if not mitigated. Most IXPs implement protections, but the shared nature introduces risks that don't exist on private links.

Concentration risk: If an IXP experiences problems, all peering relationships there are affected. Networks address this by participating in multiple IXPs.

Remote Peering

Some networks connect to IXPs without physical presence in the facility. A Layer 2 circuit extends from their location to the IXP, letting them participate virtually.

This lowers barriers—you don't need equipment in the IXP's city—but introduces latency (traffic must traverse the extension circuit) and cost (circuit fees). It's a tradeoff between accessibility and performance.

The Architectural Shift

This is the profound thing about IXPs: they change the shape of the Internet.

Without IXPs, the Internet is hierarchical. Traffic flows up to large transit providers, across their backbones, and down to destinations. Power concentrates at the top.

With IXPs, the Internet becomes a mesh. Networks interconnect directly at multiple points. Traffic takes efficient paths. Redundancy emerges naturally. No single provider controls the chokepoints.

This distributed architecture—thousands of networks peering at hundreds of IXPs worldwide—is why the Internet is resilient. It's why a cable cut in one region doesn't bring down the whole system. It's why the Internet routes around damage.

IXPs are invisible infrastructure. Users never see them, never think about them. But every time your video call is crisp, your game is responsive, your page loads fast—there's a good chance an IXP is why.