Redundant QinQ IPoE (DHCP) Network Design

Hi guys,

​

I own an ISP and was hoping to get some ideas and thoughts from everyone on how they would go about designing a redundant QinQ IPoE (DHCP) Network Design. Here is my current Network Design to help visualise things. At the moment I have x1 cross-connect into x1 Router (WA-RTR-01) that has the S-TAGs and C-TAGs configured on the cross-connect interface. For the customers I add a /32 IP address to the C-TAG VLAN (The customers circuit) and it all works fine. The authentication is based on a unique DHCP option 82 Agent-Circuit-Id attribute. As you can imagine only having x1 cross-connect into x1 Router makes me live in constant fear if the cross-connect, fibre patch lead, SFP module, or Router were to go down.

add address=123.123.123.254 interface=C-TAGx network=123.123.123.1

add address-pool=123.123.123.0/24 disabled=no interface=C-TAGx lease-time=12h name=NBN-DHCP-C-TAGx use-radius=yes

​

I understand that if I changed to a HA PPPoE design with PADO delay like in this article a lot of my issues would go away (although CPE PPPoE failover times can be bad on a lot of Home Routers). The reason I haven't done this is because with IPoE when a customer moves over to my network they don't have to re-configure their Home Router due to majority of the ISPs in my country are using IPoE as well. As you know walking non-technical people through changing their Home Router settings to use PPPoE can be quite a bit of support overhead, which is why I am doing my best to stick with the IPoE design.

​

I have thought about getting x2 cross-connects into a stacked Switch running LAG (LACP), but the control plane would be a single point of failure then (Cisco/Other uptime bugs have made me not want to go down this route). MLAG would be the solution in my head, but it is proprietary to the vendor and the "L2 NBN Aggregation" wouldn't support it so that isn't an option. Maybe I can do something with VRRP, VPLS, MPLS, Scripts, or something? Maybe plugging the x2 cross-connects into x2 Switches and manually changing the tagged interface to the 2nd Router (WA-RTR-02) on failure is an option? I would be okay with Active-Active or Active-Passive. I am open to changing the way I am doing the IP addressing on the C-TAG if that is restricting me from a solution. I would be okay with a solution that fails over to the other cross-connect in a few seconds or even minutes. At this point in time if the cross-connect, fibre patch lead, SFP module, or Router dies then I would have to physically go to the Data Centre to fix it which would be a few hours of outage.

​

Thank you for taking the time to read this and any input to help steer me in the right direction would be greatly appreciated, cheers!

​

TLDR; How would you redundantly design a QinQ setup into redundant Routers?