1

If I have the following L3vpn topology:

CE --- PE1 ---- PE2 --- CE

*PE1 and PE2 are directly connected.

*PE1 and PE2 exchange VPNv4 routes.

*If PE1 and PE2 were to exchange transport labels for their loopback address and PHP was enabled, then traffic within backbone would have only a VPN label.

1) So my question is, in such a scenario, is an mpls tunnel between PE1 and PE2 really needed for VRF customer traffic to pass via the provider backbone?

2) A non-default VRF route can point to a NH in deafault VRF table only if it has a transport label associated with it (even if it means no label)?

3) Can other types of tunneling be used instead of mpls? (GRE for example?)

2

You can also use MPLS over UDP, which was defined in RFC 7510 and is arguably a better choice than GRE as it has better hashing properties.

Note that tge VPN label is still necessary as it defines the egress VRF or interface,

1

Found it. Backbone can be non-MPLS.

From RFC4364: If the backbone does not support MPLS, the MPLS packet carrying only the VPN route label may be tunneled to the BGP Next Hop using the techniques of [MPLS-in-IP-GRE]. When the packet emerges from the tunnel, it will be at the BGP Next Hop, where the VPN route label will be examined.

  • I'm glad you found your answer, but I'm having trouble understanding it. Practically speaking, you can configure anything you want, but If your backbone is MPLS, why would you configure something different, especially if the PEs are adjacent? – Ron Trunk Sep 10 '18 at 14:58
  • Just was curious about the mechanics of BGP next-hop reachability for mpls VPN. Tge case of adjacent PEs confused me a bit, since no transport label is actually present in this case - however the tunnel is still signaled. In any case, according to RFC4346 some form of tunneling is a must. – sergeyrar Sep 10 '18 at 15:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.