One of the most important steps in a small ISP's growth is moving off a single upstream. One link means every subscriber goes offline when it fails, and you have little leverage on price. BGP multihoming fixes both.

Prerequisites

You need your own autonomous system number (ASN) and at least a /24 of IPv4 and a /48 of IPv6. The ASN and an IPv6 PI block come through a RIPE NCC member acting as your sponsoring LIR, or through your own RIPE NCC membership. On the IPv4 side, a new /24 is only available via the RIPE NCC waiting list, the transfer market or leasing. Create route and route6 objects in the RIPE database for your prefixes; upstreams build their filters from them.

Design options

Default route: take only a default route from each upstream. Simplest to set up, but you cannot pick the best exit per destination.

Partial table: each upstream's customer routes plus a default. A middle ground with modest memory needs.

Full table: every prefix on the internet. You can choose the best path to each destination and split traffic by cost and quality. The IPv4 table holds more than a million prefixes, so the edge router needs the memory and processing to match.

Steering traffic

  • Outbound traffic is steered with local-preference: give routes from your preferred upstream a higher value.
  • Inbound traffic is steered with AS-path prepending and the BGP communities your upstreams offer. Communities let you choose where and with what priority your routes are announced in the upstream's network.
  • BFD detects link loss far faster than BGP timers, shortening failover.

IX peering

Most of your traffic goes to a handful of large content networks. Peering with them directly at an internet exchange cuts transit cost and latency. Your multihomed edge can carry the IX link like a third upstream.

RPKI route origin validation

BGP itself does not check who may announce which prefix. If another network announces your prefix, by mistake or on purpose, your traffic can go elsewhere. RPKI reduces that risk:

  1. Create ROAs: in the RIPE portal, create ROAs stating which ASN may announce each prefix and the maximum prefix length.
  2. Enable ROV: your edge router checks each incoming route against data from an RPKI validator and marks it valid, invalid or not-found.
  3. Reject invalid routes: drop invalid routes in your policy. Most large transit providers already do.

With CoreGate BGP Edge Gateway

CoreGate BGP Edge Gateway brings IPv4 and IPv6 full tables, RPKI route origin validation, BFD, community-based policy and FlowSpec onto one platform. It can run alongside your current edge for a staged migration, and we help from the RIPE NCC application through to the first configuration.