“Standard BGP keepalives take 180 seconds to detect a link failure. Configuring BFD 50ms polling enables instant sub-second failover for active virtual gateways.”

While architecting resilient cloud connectivity at IBM, link drops on physical ESXi uplinks caused up to 3 minutes of packet loss while BGP hold timers expired.

Tuning BFD Timers for Microsecond Failure Detection

We enabled Bidirectional Forwarding Detection (BFD) on the eBGP session between NSX-T Edge Gateways and physical Arista ToR switches.

BFD operates independently of BGP routing logic, sending fast UDP probe packets (port 3784) at 50ms intervals with a multiplier of 3. If 3 consecutive probes are lost (150ms total), BFD instantly notifies BGP to switch active routes.

[!NOTE] Ensure physical switch hardware supports hardware-offloaded BFD processing to prevent CPU overload when polling multiple BFD sessions at 50ms intervals.

! # Arista ToR Physical Switch BFD Configuration
interface Vlan200
 description Transit-SDDC-Lab-Uplink
 ip address 192.168.200.1/24
 bfd interval 50 min_rx 50 multiplier 3

router bgp 65000
 neighbor 192.168.200.2 fall-over bfd

The Verdict

Key Takeaway

Pair Every BGP Gateway Session with BFD Sub-Second Probing.

Enable BFD with 50ms polling on all SDDC gateway peering links to achieve sub-second failover during physical network disruptions.