Routing

Good morning all,

I'm wondering what the commonly used/best practice is for Layer 2 transport to BNG routers?   I've looked at EVPN-VPWS and so far that looks like the best solution, but I'm curious what other providers are using?  Or if there is an industry standard here?   

Any thoughts?   Any best practices to share?

Thanks

Hi Chuck,

As with most things, there is no one-size-fits-all answer or industry standard here. I have seen more and more that the de-facto choice is, in fact, EVPN-VPWS. Assuming that you're not running a large network with many PEs that must first aggregate access domains before the BNG, PWHT and EVPN-VPWS look like frontrunners. I'd consider a couple of key factors for starters.

Are you running redundant BNGs? If so, what is the redundancy mode? Some redundancy modes require that both BNGs be reachable from the RGs at all times, such as PADO delay. In this case, both PWHT and EVPN-VPWS Single-Active might not be the best option. Others, such as Packet Triggered Based Recovery, require that the backup is not reachable from the RG unless the primary BNG is down. In this case, either PWHT or EVPN-VPWS would seem a good fit.

The next question is what your reversion looks like. For example, if you're running PPPoE, a reversion will be just as impacting on subscribers as a failover, and all RGs will need to disconnect, initiate a new session via PADI, and fully reconnect to have service. You probably want to do this during a maintenance window. Even in the IPoE case with Packet Triggered Based Recovery, the recovery mode, and therefore reversion mode, is degraded (or widely open) depending on the policies you set. EVPN-VPWS cannot practicably be non-revertive, so you don't have the control to bring those sessions back to your primary BNG in a maintenance window, which might drive you toward a PWHT solution. This may be far less of a concern with approaches like ALQ.

If you're not running redundant BNGs, you may have a strategy of a cold standby or even another BNG that could take the load in the event of a catastrophe. In these scenarios, it's probably a lot easier to be able to load the primary's EVPN-VPWS config on the cold standby (along with any other required parameters such as IP routing, pools, etc) and be up and running as opposed to needing to touch every PE that carries traffic from your access to reconfigure l2circuits, so I'd gravitate in that general direction (EVPN-VPWS) lacking a redundant BNG. I might even take some extra steps (set a lower priority on your primary both regarding routing and a single-active ESI) to ensure that if you do need to activate the cold standby, you don't experience unexpected reversion when the situation is resolved.

------------------------------
Jason Rokeach
------------------------------