Make Public Peering More Resilient with BFD and BGP Add-Path on MegaIX

By Gavin Tweedie, Senior Interconnection Director

Standard BGP can be slow to react. See how BFD and BGP Add-Path help improve uptime and path diversity on MegaIX.

When peering on a public Internet Exchange (IX), standard BGP behavior can introduce delays during topology changes. But if you use MegaIX, there are workarounds.

In this blog, we explain two lesser-used features available on MegaIX which maximize uptime and path diversity: Bidirectional Forwarding Detection (BFD) and BGP Add-Path.

How to enable BFD with MegaIX route servers

Standard BGP keepalive and hold timers (usually 60 and 180 seconds) are slow for modern production traffic. If an intermediate Layer 2 transport failure occurs across an IX fabric or over your transport to MegaIX, your router might continue blackholing traffic for up to three minutes before declaring the neighbor dead and withdrawing routes.

Enabling BFD solves this by introducing microsecond or millisecond keepalives directly in the forwarding plane. Many peers already run BFD with each other when setting up bilateral sessions directly, but there’s no reason not to run BFD with your route servers, too.

Why BFD with a route server matters

On an IX, your router establishes BGP sessions with the Route Servers to learn prefixes from hundreds of peers via a single session. If a peer drops off the fabric or experiences a localized issue, the route server updates its routing table.

However, if the Route Server itself (or your direct path to it) experiences a quiet failure, you have to wait for BGP timers to expire before removing those paths as viable best paths, and the Route Servers will delay in withdrawing your routes from other peers.

By running BFD to the MegaIX Route Servers, your edge routers can detect a loss of connectivity within milliseconds, quickly tearing down the BGP session and shifting traffic to alternate transit suppliers, IXs, or private interconnects.

Because not all peers run BFD on an IX, MegaIX implements BFD Passive Mode.

• The MegaIX Route Server will never initiate a BFD control packet. It simply listens on the BFD standard port UDP port 3784.
• Your edge router must be configured as BFD Active. It must actively send the initial BFD control packets to bootstrap the session. If your side is left at default or configured as passive, the BFD session will never establish.
• Once BFD is established it operates like a normal BFD session.

How to eliminate path hiding with BGP Add-Path

By default, BGP is designed to advertise only its single best path to a neighbor. In a traditional Route Server environment, this can cause path hiding.

The problem: best path masking

If three different networks (Peer A, Peer B, and Peer C) all advertise 192.0.2.0/24 to the MegaIX Route Server, the Route Server runs its own standard BGP “best-path” algorithm. It selects one best path—say, Peer A—and reflects only that single path to your router and all other peers.

Even though Peers B and C offer valid paths to that destination, your router has zero visibility into them. If Peer A drops, your router must wait for the Route Server to recalculate, select the next best path, and send a new BGP update.

The solution: BGP Add-Path

This is why MegaIX has its BGP Add-Path capability. This feature allows the Route Server to advertise multiple paths for the exact same prefix to your router by appending a unique Path Identifier to each route.

Without Add-Path:

MegaIX RS ---> [Only Best Path: Peer A] ---> Your Router

With Add-Path:

MegaIX RS ---> [Path 1: Peer A, Path 2: Peer B, Path 3: Peer C] ---> Your Router

Advantages of Add-Path

1. Local policy control

Your router receives all available paths across the IX fabric from the route server, allowing your local BGP engine to make its own routing decisions based on your specific Multi-Exit Discriminator (MED), local preference, or AS-path length constraints, rather than inheriting the Route Server’s choice.

2. Load balancing

In the example above, where three equal paths are supplied to your router, you could also choose to balance traffic across all three next-hops for better performance.

3. Fast reroute

By holding backup paths in your router’s Routing Information Base (RIB), you can pre-compute a backup next-hop in the Forwarding Information Base (FIB). If the primary peer fails, failover happens at line rate without waiting for a new BGP convergence cycle from the RS.

Public peering doesn’t have to rely on slow default failover or a single visible path. With BFD and BGP Add-Path on MegaIX, you can detect failures faster and make better routing decisions from your own edge, resulting in a more resilient peering setup with greater control over uptime and performance.