My understanding of MPLS LDP is that it allocates a label per destination prefix in the FIB. So, If I advertise the global internet table between routers connecting over LDP enabled interfaces, will I exceed a label limit or set fire to my router?
Some context; I have several routers iBGP peering with each other, enabled for the IPv4 and VPNv4 address families. The global table on these routers connects to the internet, learns a default route which is redistributed, all happy. There are several MPLS VRFs, VPNv4 routes being exchanged successfully, also happy. Several (virtual) devices act as gateways to their respective MPLS networks, eBGP peering with both their VRF and the global table - no issues.
Now rather than just a default route, I need to start ingesting the full internet table and redistributing it into my iBGP. My concern is that my router will try to allocate a label for every single global destination prefix, and exceed some capacity. I'm doing some reading on LDP merging but it sounds like it's only relevant for P routers.
So what's the solution? And if any of my understanding is incorrect please let me know. - Run MPLS VPNs and Internet tables on different hardware? - Different iBGP peering(s) for IPv4 and VPNv4, and keep IPv4 off LDP interfaces? - Run the internet table in it's own VRF so there are many fewer outer MPLS labels? Not even sure if this would work because there's still the inner label? - Maybe iBGP next-hop-self reduces the transport labels down to one per router?
No comments:
Post a Comment