What Is an Internet Exchange (IX) and How Does Peering Work?

Updated

An Internet Exchange (IX) is a neutral physical and logical hub where separate networks connect directly to exchange traffic with each other. Instead of paying an upstream provider to carry every packet, a network joins an IX, establishes peering sessions with other members, and sends traffic straight to its destination. This matters most when you want lower latency, more control over routing, and reduced dependence on paid transit — but it only pays off if the networks you need to reach are actually present at that exchange.

The core idea: a shared switching fabric

An Internet Exchange Point (IXP) operates a shared network fabric — typically Ethernet switching infrastructure in one or more data centers. Each member connects a port to this fabric and runs BGP (Border Gateway Protocol) sessions with the other members it chooses to peer with.

The IX itself does not route your traffic or sell you reachability. It provides the meeting point. What you do there — which peers you establish, what routes you announce — is up to you and the networks you negotiate with.

DE-CIX, for example, describes itself as operating neutral Internet Exchanges across Europe, the Middle East, Africa, Asia, and the Americas, providing interconnection services to networks of many types. "Neutral" here means the exchange does not compete with its members as a carrier and does not favor one network's traffic over another.

Peering at an IX vs. buying IP transit

These are two different ways to get your traffic to the rest of the internet, and they solve different problems.

Dimension Peering at an IX IP transit
What you pay for Port fee to the IX, plus your own equipment and cross-connect A provider sells you full or partial internet reachability
Who you can reach Only networks that agree to peer with you and are present at that IX Effectively the whole internet, via the provider's upstreams
Routing control You decide which prefixes to announce and accept Largely delegated to the transit provider
Cost model Fixed port and facility costs; traffic volume is not metered per gigabit in the same way Usually volume- or capacity-based
Setup effort You run BGP, manage peering policy, and negotiate sessions Simpler — one relationship, one bill
Redundancy Depends on how many peers and IXs you use Depends on the provider's own resilience

The practical takeaway: peering is not a replacement for transit. Most networks use both. Transit guarantees you can reach anyone; peering lets you reach high-volume destinations directly and cheaply, which reduces how much traffic you push through paid transit.

Who connects to an IX, and why

  • ISPs and access networks — to exchange traffic with content networks and other ISPs, cutting transit costs and improving the path to popular destinations.
  • CDNs and content providers — to deliver content from as close to the end user as possible, which reduces latency and improves delivery quality.
  • Enterprises — to connect their own sites and reach cloud or SaaS providers directly rather than backhauling traffic through a transit provider.
  • Cloud providers and network operators — to offer or obtain direct interconnection as part of a broader ecosystem.

The common thread is that each participant wants direct, controllable paths to specific other networks rather than routing everything through a third party.

The practical benefits

Lower latency. Traffic takes a more direct path between the two networks instead of traversing one or more transit providers. Fewer hops generally means less delay, though the actual improvement depends on geography and how the peers are connected.

Reduced cost. Once you have a port and peering sessions, you can offload a large share of traffic that would otherwise be billed as transit. The savings depend on how much of your traffic goes to networks present at the IX.

Improved redundancy. Peering with multiple networks across multiple exchanges gives you alternative paths. If one peer or one exchange has a problem, traffic can shift — provided you have configured your routing to allow it.

More control. You choose your peers, your routing policy, and your prefix announcements. That control is also a responsibility: misconfigured BGP can affect your own reachability and, in serious cases, that of others.

What to weigh when choosing an IX

  • Location and latency to your users. An exchange near your traffic's origin or destination gives more benefit than a distant one.
  • Neutrality and governance. A neutral exchange does not compete with members as a carrier. This affects how fairly the fabric is run and who is willing to join.
  • Available networks. The value of an IX is the set of potential peers present. An exchange with the content networks, cloud providers, and ISPs you actually want to reach is worth more to you than a larger one that lacks them.
  • Ecosystem and services. Some exchanges offer additional interconnection services beyond basic peering. DE-CIX, for instance, markets "premium interconnection services" alongside its exchanges — check what is included and what is charged separately.
  • Resilience and port options. Consider whether you can connect at multiple locations or exchanges for redundancy, and what port speeds and cross-connect arrangements are available.

How peering is actually set up

  1. Join the exchange. Arrange a port on the IX fabric and a cross-connect from your equipment in the facility.
  2. Configure your router. Assign the IX-provided IP addressing to your interface and prepare BGP.
  3. Establish sessions. Peer with the networks you have agreed to interconnect with, using the IX's route servers where available or direct BGP sessions.
  4. Announce and filter. Announce your prefixes and apply prefix filters and maximum-prefix limits to protect against route leaks and hijacks.
  5. Verify. Confirm sessions are up, routes are being received and advertised as expected, and traffic is flowing over the peering paths.

A common pitfall is skipping filtering. Peering without proper prefix and AS-path filters can expose you to accidental route leaks — a frequent cause of widespread outages. Another is expecting immediate reach: peering only helps with networks that have agreed to peer with you, so the benefit grows as your peering relationships do.