BGP control plane information in MPLS VPN

Question

I am learning about MPLS VPN networks. From my understanding an IGP runs on all core routers (P and PE), while BGP runs on all PE routers. Once the IGP has conveyed reachability information to all routers, and all routers have converged, the exact labels to be used to transfer packets are assigned using LDP.

My doubt is, how are BGP control packets transmitted between PEs.

There are two options.
1. To use the Label switched paths between PEs.
2. To use normal IP forwarding through the P routers.

Which of these two methods is actually used?
If both can be used how does the PE router make a decision on which one to use?
Do we have to manually configure it in the router?
Do these answers vary for different routers like Juniper, Cisco etc. ?

Answer 1

My doubt is, how are BGP control packets transmitted between PEs. There are two options:

1. To use the Label switched paths between PEs.
2. To use normal IP forwarding through the P routers.

If both can be used how does the PE router make a decision on which one to use?

LSPs are preferred over per-hop IP forwarding, if an LSP is available.

Do we have to manually configure it in the router?

By 'it', do you mean configure use of the LSP for BGP control-plane information? It happens automatically on a Cisco IOS box

Do these answers vary for different routers like Juniper, Cisco etc. ?

Cisco will send BGP information through an LDP LSP, as long as the BGP endpoint prefix has an label binding.

I can't remember Juniper's behavior off-hand, they offer somewhat more granular control over LSP behavior.

Answer 2

BGP uses TCP te setup its connection and to send their packets to his neighbors. This means that your neighbors need to see each other on layer 3 (ip) level.

I hope this is the info that you needed.

see: http://en.wikipedia.org/wiki/Border_Gateway_Protocol section "operation" for more details on this matter.