How to join an IXP
Practical steps to join an internet exchange: a PeeringDB profile, ordering a port and LOA, configuring BGP on the router and sessions with route servers.
Last updated:
Joining an IXP (internet exchange point) involves a few concrete steps. You fill in your PeeringDB profile, order a port on the fabric, provide an LOA and cross-connect into the data centre, configure BGP sessions on your edge router, and peer with the route servers and selected bilateral partners. To do this you need your own ASN, an allocated block of addresses (PI or PA) and a router that runs BGP.
Step 1: PeeringDB and prerequisites
Before ordering a port, complete your network's PeeringDB profile. It is the industry database that operators and route servers draw on for information about your ASN: prefix ranges (max-prefix), peering policy, points of presence and NOC contact details. A credible, up-to-date PeeringDB record is a practical condition for most partners to accept your session.
- An ASN (16- or 32-bit) allocated by an RIR through an LIR.
- IPv4 and IPv6 address blocks to announce, with matching route/route6 objects in an IRR.
- ROA records in RPKI for the prefixes you announce, per RFC 6482.
- An edge router with BGP and a free port (1G, 10G or 100G).
- An up-to-date PeeringDB profile with peering policy and a NOC contact.
Step 2: Port, LOA and connecting to the fabric
Submit a port request to the IXP operator at your chosen speed. The IXP assigns you addresses from the shared peering subnet (one IPv4 range and one IPv6) and tells you which switch and data centre to use. If your equipment sits in the same facility, you receive an LOA (Letter of Authorization) that lets the data-centre operator provision a cross-connect between your rack and the IXP port. Where you have no physical presence, the same result comes from remote IXP access over a transport link.
Step 3: BGP configuration and route servers
Once the port is up, configure an interface with the assigned peering address and bring up BGP sessions. The fastest route to broad reach is peering with the IXP route servers: a single multilateral session yields routes to every other member connected to that server. With key partners, add bilateral sessions where precise policy is required. Whichever model you use, apply BGP hygiene: IRR-based prefix filters, RPKI route origin validation (RFC 6811), a max-prefix limit and BCP 38 ingress filtering against spoofing.
After the sessions are established, verify that routes come in and go out as expected and that traffic actually begins to flow over the IXP instead of via transit.
How SkyPass simplifies joining Polish IXPs
AS202520 SkyPass provides remote access to the Polish internet exchanges (THINX, TPIX, WRIX, 1-IX) without you having to place your own hardware in every data centre. From our PoPs in Warsaw and Wrocław we deliver a port on the fabric, sessions with the route servers and support for BGP, RPKI and filter configuration, alongside DDoS protection. If you are starting out with peering or want to offload IP transit, we will help you through the whole path from a PeeringDB profile to working sessions.
Frequently asked questions
Do I need my own ASN to join an IXP?
Yes. Peering at an IXP requires an ASN allocated by an RIR through an LIR and your own IPv4/IPv6 address block to announce in BGP.
What is an LOA and when is it needed?
An LOA (Letter of Authorization) is a document authorizing the data-centre operator to provision a cross-connect between your rack and the IXP port. It is needed when you have a physical presence in the same facility as the fabric.
Do I have to be physically present in the IXP data centre?
No. If you have no equipment there, you can use remote IXP access over a transport link, which delivers a port on the fabric without colocation at that location.
Should I peer with the route server or bilaterally?
The route server gives broad reach fastest through a single multilateral session. Bilateral sessions are added where precise routing policy toward a specific partner is needed. Most networks combine both approaches.