ERA-IX Peering LAN Route Servers


Full Mesh BGP Peering

Computer networks use route reflectors. To ease peering between many routers. When a network BGP network without route reflectors will require N*(N-1)/2 BGP sessions to remain in full-mesh topology. Or even twice as many if all networks operate both IPv4 and IPv6.

A network with 3 routers will require just 3*(3-1)/2 = 3 BGP sessions, but a network with 19 routers will require 19*(19-1)/2 = 171 BGP sessions in a full-mesh topology!

As it would be incredibly difficult to maintain this amount of bilateral peering sessions, a route-reflector can be set up to distribute the routes. In the case of an internet exchange this route-reflector is commonly known as a route-server.

Route Server BGP Peering

When a route-server is utilized, all members can peer and announce their routes to the central route-server. The route-server will take care of filtering the routes and distributing the remaining, validated routes to the other members.

ERA-IX provides two route-servers for redundancy, these route-servers are hosted on two seperated physical machines to ensure continuity.

Peering in the route-server model and setting up direct (bilateral) peering sessions are not exclusive, they can be used to complement each other.

Route-Server filtering

By default, our route-servers are configured with the following filtering policy

  • Drop invalid IRR routes, routes announced to our route-servers must be properly administered in their relevant IRR.
  • (ROV) Drop Invalid, when a route does not match it's ROA, the route is dropped. Routes which do not have a ROA or match their ROA are accepted.

Upon request, our outbound policy can be changed, this will require you to configure your own filtering to ensure secure network operation.

Peering LAN Route-servers

Name ASN IPv4 Address Max-routes (IPv4) IPv6 Address Max-routes (IPv6)
Route Server 1 206221 300000 2001:7f8:12a::1 100000
Route Server 2 206221 300000 2001:7f8:12a::254 100000

Configuring Route-server Peering

To get started with our route-servers you will have to configure BGP sessions towards our route-servers. With default policy in place, the following (example) BGP configuration can be adapted to work inside your network:

ERA-IX route-server configuration example

router bgp 64512
  neighbor era-ix-ipv4 peer group
  neighbor era-ix-ipv6 peer group
  neighbor era-ix-ipv4 remote-as 206221
  neighbor era-ix-ipv6 remote-as 206221
  neighbor era-ix-ipv4 description era-ix
  neighbor era-ix-ipv6 description era-ix
  no neighbor era-ix-ipv4 enforce-first-as
  no neighbor era-ix-ipv6 enforce-first-as
  neighbor era-ix-ipv4 send-community large
  neighbor era-ix-ipv6 send-community large
  neighbor era-ix-ipv4 maximum-routes 300000
  neighbor era-ix-ipv6 maximum-routes 100000
  neighbor peer group era-ix-ipv4
  neighbor peer group era-ix-ipv4
  neighbor 2001:7f8:12a::1 peer group era-ix-ipv6
  neighbor 2001:7f8:12a::254 peer group era-ix-ipv6
  address-family ipv4
    neighbor era-ix-ipv4 prefix-list export-prefixes out
  address-family ipv6
    neighbor era-ix-ipv6 activate
    neighbor era-ix-ipv6 prefix-list export-prefixes out

The most important statements (some of which might be unique):

  • no enforce-first-as This statement disables the check on the first AS in the AS-PATH. If you do not set this, inbound updates will be ignored and you will see zero prefixes from our route-servers.
  • send-community large This statement enables sending communities to our route-servers, allowing you to use them for traffic-engineering.

We recommend applying explicit and strict import and export route filtering, in the example import filtering has been left out. Route-maps and prefix-lists can be used to achieve the desired effect.

Enforcing Policy

Peering with route-servers does not have to mean you lose all control of your peering policy!
With route-servers you remain in control about who your routes are exported to. To learn more, visit our BGP Communities documentation.