Hi,

we are running a 40 nodes network routed by bgp, where each node is viewed like an AS.
the AP are link to several pppoe server through the woip links.

Now we are starting to think about mpls.
But i am very new to mpls so i need some advice.

Is it possible to mantain the same logic each node is an AS and the propagate the ldp through the bgp?
In fact the wiki exemple is about only ibgp and so i don't know it is tue also in a ebgp scenario.

Hope to receive some hint about it.

regards
Ros

Now we are starting to think about mpls.
But i am very new to mpls so i need some advice.

My main advice is to wait. MPLS is not completely stable in ROS yet, there are still a number of bugs which need to be ironed out.

By all means, start planning, designing and testing, but I would avoid putting it on a production network for now.

i am starting to think about it not putting it in production.

My idea is:
- the vpls should connect the AP to the PPPoE concentrator so i shoul duse static VPLS configuration.
- but i am running eBGP protocols thrugh the nodes and so LDP doesn't work throiught it.

So my question is:
how to easly modify my design to accomodate the ldp distribution without the network outages?

regards
Ros

What do you mean by "LDP does not work through it"? LDP is completely separate protocol and although it is pretty common to use IGP routing protocols in MPLS network, it is not a requirement, you can even run LDP in statically routed network. I do not understand exactly how you use eBGP in your network (and what are the benefits in doing so), but when considering MPLS make sure that you understand the difference between label assignment to IGP and BGP learned routes, probably you will find out that you should change over to some IGP.

Wiki examples talk about iBGP because BGP has extensions for services (BGP based VPLS or IP VPNs) on top of MPLS enabled network core, it is not required to "MPLS enable" your network core.

we configured the bgp as ebgp between the node to have the possibility to prepend the aspath to do traffic engineering on the hops.

Is it possible to have LDP propagated on such behavior?

regards
Ros

If you use eBGP as IGP protocol (meaning - you run it on all routers in network and do not have any other IGP protocol operating), you will not have any benefits from enabling MPLS due to how BGP learned routes are handled:
1) LDP router does not distribute labels for BGP learned routes
2) for BGP routes mapping for IGP prefix "through which" BGP NextHop is looked up is used, instead of mapping for route prefix (like for static or other IGP routes). In your case this would mean that BGP nexthop is always reached by connected route that obviously will not use any mapping.

So I guess you have the following options:
1) start using some IGP - either replacing your BGP setup or running in parallel for some time
2) request a feature that would enable handling of BGP routes the same way as IGP routes (probably based on some property set by route filters)
3) use non-CSPF TE tunnels

I think the best way is to start to use ospf in tnadem with ebgp.

As soon as i have the IGP running i will test the VPLS static config between my ap and pppoe concentrator.

Regards
Ros