How an IXP works
An IXP is a shared layer-2 fabric where networks peer directly. Learn about the switching fabric, route servers and bilateral vs multilateral peering.
Last updated:
An IXP (Internet Exchange Point) is a shared layer-2 infrastructure where many independent networks (ASNs) interconnect to exchange traffic directly, bypassing transit from upstream providers. Physically an IXP is usually a set of ethernet switches distributed across several data centres and joined into a single logical network segment.
Switching fabric and member ports
The heart of an IXP is its switching fabric, a shared layer-2 domain built on ethernet switches. Each member connects its edge router to a dedicated port on the fabric and receives an IP address from a common subnet, normally one IPv4 range and one IPv6 range. To the routers it looks as though every member sits on a single large ethernet segment.
Member ports are offered in standard speeds, and a network picks its capacity based on traffic volume and expected growth.
- 1G and 10G for smaller networks and new members.
- 25G and 40G as an intermediate step.
- 100G and 400G for large carriers, CDNs and hyperscalers.
- Link aggregation (LAG) once a single port is no longer enough.
Route servers, bilateral and multilateral peering
On the fabric, networks exchange routes using BGP. Bilateral peering means two networks configure a direct BGP session between their routers and decide for themselves which prefixes to announce. This gives full control, but the number of sessions grows with the number of partners. To simplify connectivity, an IXP provides route servers: a member sets up a single BGP session with the route server and, in this multilateral model, exchanges routes with every other participant connected to that server.
The route server does not sit in the data forwarding path. It only mediates the distribution of routing information, while the packets themselves flow directly between member routers across the fabric. Many networks combine both approaches. They use the route server for broad reach and keep bilateral sessions with key partners where precise routing policy is required.
Benefits and examples of exchange points
The main benefit of being present at an IXP is keeping local traffic local. Instead of sending packets to a distant transit provider, networks exchange them directly, which shortens latency and the packet path. For many operators peering also lowers cost, because traffic exchanged at an IXP is usually not billed by volume the way transit is. Among the best known exchange points are AMS-IX in Amsterdam, DE-CIX in Frankfurt, LINX in London and TPIX in Poland.
Frequently asked questions
What is the difference between bilateral and multilateral peering?
Bilateral peering is a direct BGP session between two networks, giving full control over policy. Multilateral peering uses a route server, so a single session provides connectivity with many members at once.
Does the route server forward network traffic?
No. The route server only distributes BGP routing information, while the packets themselves flow directly between member routers across the layer-2 fabric.
Is peering at an IXP always cheaper than transit?
Traffic exchanged at an IXP is usually not billed by volume, so with large local volumes peering can be cheaper. You still need to account for the port, the link to the IXP and operational effort.
What port speed should I choose when starting out?
Smaller networks usually start with a 1G or 10G port and scale up as traffic grows. Pick a speed with headroom for expected growth.