BGP Convergence Optimization

I’m exposed to an incredible variety of topics in my ExpertExpress engagements, but there are always a few recurring themes, one of them being “we’re experiencing long convergence times and high packet loss after our primary Internet link fails.” Almost always the root cause turns out to be full Internet routing table being received on inadequate hardware.


  1. Why in the world would you want to receive the full internet prefix list in an active/standby uplink design? Have your providers advertise a default route to you if you're not going to do intelligent egress selection.
  2. In principle I agree with you ... unless your provider happens to be someone known for peering spats and unilateral de-peering actions.

    Some people think it's safer to get full feed from both providers just in case one of them is missing a prefix or two.
Add comment