Saturday, May 11, 2019

BGP or OSPF for MPLS PE-CE routing?

Having an internal discussion at present with my boss. We have a mpls core with our corporate vrf running a l3vpn. We have a bunch of new offices and data centres coming online this year and we are working out how we connect these back to the mpls.

His preference is to go with ospf between the mpls and data centre/office. I prefer bgp.

He's had experience of ospf in large environments, but to me it seems like a massive pita, with added risks, and not much benefit. As a trial we connected a new switch stack up as a DC core to our mpls and instantly I had to fudge my way around it. First enabling capability vrf-lite as we run vrfs within the DC. This then caused a loop (obviously), which meant I had to filter routes on the pe devices. His feedback though is it was worth it for faster failure recovery.

I want to go with bgp. It doesn't come with these dn-bit caveats. It has loop prevention due to as-path. And is just easier to work with.

Anyone else had experience in the same? For a bit of extra information, each new site/DC will have 2uplinks back to different pe's



No comments:

Post a Comment