OSPF: scaling routers in area or areas per se
Dear all, I'm currently designing a new network, it has 2 DC's with each having a VPN cluster. Each VPN cluster has a Firewall Cluster behind and it's set up in a circle, so VPN in DC1 has a DCI to VPN in DC2 and a dedicated link to FW in DC1. Same as in DC2 and FW in DC2 has a DCI to FW in DC1. In this circle I'm running area 0.0.0.0 and wherever I add a new interface it's known to the network. Now comes the fun part. We have 100 branches and we use OpenVPN to connect to one of the VPN concentrators with the second as backup. The challenge is, that it shouldn't matter where the branch currently is connected to, BUT, it's only allowed to be connected to one DC to prevent async routing. Technically this is also easy, but I'm not sure which one scales better: - Have only one area 0.0.0.0 also on the branches, so over 100 routers in one area - Have for each branch one area, so over 100 areas having only one router - Running external via BGP and having 100 neighbors with 100 private AS Anyone here experienced regarding scaling this size? Thx! Michael
On Tue, 21 Sept 2021 at 15:34, Muenz, Michael <m.muenz@spam-fetish.org> wrote:
- Have only one area 0.0.0.0 also on the branches, so over 100 routers in one area - Have for each branch one area, so over 100 areas having only one router - Running external via BGP and having 100 neighbors with 100 private AS
Anyone here experienced regarding scaling this size?
I can't comment on FRR specifically, but the last option is the general approach taken in the service provider world. Note; you can use the same ASN at each branch site with BGP Site of Origin + "allow-as in". Both of these BGP methods are very common in the SP world, and scale much higher than OSPF for this purpose (CE to PE connectivity). If the end site LANs ranges raley change then there is no more scalable option than static routes. If you can - go with static routes. Cheers, James.
participants (2)
-
James Bensley -
Muenz, Michael