What BGP is

BGP exchanges routes between Autonomous Systems (AS) on the internet. Every ISP with its own ASN needs to run BGP to announce IPv4/IPv6 prefixes and receive routes from upstreams and peers.

Bad BGP can take you off the internet, leak third-party routes (Pakistan/YouTube 2008 style) or create traffic loops that take down the backbone.

What RASYS does with BGP

  • IX.br peering — onboarding, session setup and curation of routes received via route servers.
  • Private peering — technical negotiation and session setup with CDNs (Google, Netflix, Meta, Cloudflare), IP transit and other ISPs.
  • Prefix filtering — prefix-list, RPKI, max-prefix per session. Protects against leaks and unexpected more-specific routes.
  • Communities and local-preference — traffic engineering to exit through the right upstream, with automatic fallback.
  • eBGP/iBGP redundancy — full-mesh or route reflectors, BFD for sub-second session failure detection.
  • Multihoming — two or more upstreams with inbound/outbound engineering and automatic fallback on transit failure.
  • Monitoring — Zabbix/LibreNMS tracking sessions UP/DOWN, prefixes received/announced, BGPmon for leak alerts.

Equipment we work with

Huawei NE8000/NE40/NE20, Juniper MX, Cisco ASR/IOS XR, MikroTik RouterOS, Datacom DM, FRR/BIRD on Linux.

When it makes sense to talk to us

You don't have an ASN yet; want to reduce IP transit cost via IX; live with BGP sessions flapping; need to add a new upstream; suspect a route leak; want to deploy RPKI; merging networks after acquisition.

Talk to us — initial conversation, no commitment. See also: OSPF, MPLS, BGP in glossary.

FREQUENTLY ASKED QUESTIONS

Is RPKI mandatory to announce a prefix at IX.br?

It is not mandatory under IX.br rules, but it is highly recommended. Without RPKI, your route may be rejected by peers that validate ROA, leaving you vulnerable to hijacking. ROA registration is free at NIC.br's Stat portal and takes just a few minutes.

How do you add a second upstream without causing flap?

We bring the new session up in parallel with conservative prefix-lists and max-prefix limits, validate the received table outside production BGP (testbed), then activate it with low local-preference to act as backup. We promote it to primary only after it stabilizes.

Can you migrate BGP from MikroTik to Huawei/Juniper without losing sessions?

Yes. We bring the new equipment up in parallel, configure all sessions (peers, route servers, upstreams), validate the received table, then cut over via OSPF/iBGP by adjusting next-hop. The eBGP sessions need to reconnect (a few seconds of drop), but traffic does not go to zero because the internal routing has already been redirected.

What BGP convergence time is acceptable?

For session failure (BFD enabled): 50-300ms. Without BFD: 9-180s (default keepalive 60s + holdtime 180s). For intra-AS route change: 1-3s. Full table convergence after reboot: 5-15min, depending on the number of sessions and prefixes.