L3VPN LSP selection process

Answered Question
Sep 23rd, 2013
User Badges:

Dear all,


Please, imagine two PE routers directly connected by 1GE and one 10GE interfaces and there is one RSVP-TE LSP per link connecting both PE routers. Thay are running MP-BGP from their respective loopback interfaces.When a L3VPN is configured in both PEs:


* Which is the criteria to select the LSP for a L3VPN between both PEs? The LSP preferred by the IGP to reach the peer's loopback interface, so the LSP on the 10GE interface?

* Is there any way to manually impose the LSP used by a L3VPN?

* Is there any way to load balance L3VPN traffic between both LSPs?


Sorry if my questions are pretty basic, but honestly I couldn't find any document able to explain L3VPN's behaviour on this topology.


Thank you very much


Octavio

Correct Answer by Harold Ritter about 3 years 10 months ago

Hi Octavio,


I realize that having to specify the alternate next hop on all VRF can be cumbersome but this way you have a greater granularity. For instance, you could potentially have a different next hop for each and every VRF, which in turn could use a different tunnel. Not scalable but flexible.


If what you are trying to achieve is to use the same tunnel interface for all VRFs, you could simply establish your VPNv4 session using a separate loopback address and have that loopback address reachable via a tunnel interface via a static route.


Regards

Correct Answer by Harold Ritter about 3 years 10 months ago

Hi Octavio,


The cost of the RSVP TE signalled LSP is determined by  the best IGP cost to the tunnel tail end. In your scenario, both LSPs  would have the exact same cost, which would be the IGP cost via the 10g  interface. Assuming you are using "autoroute announce" to inject traffic  into the LSPs, the l3vpn traffic would be load balanced over the two  links by default.


You can change the default behavior  to prefer the LSP going over the 10gig link by using "tunnel mpls  traffic-eng autoroute metric". You could also load balance between  unequal bandwidth links using "tunnel mpls traffic-eng load-share".


If  you want to manually set the tunnel used for a given VRF, you can do  this by using a different next hop for that VRF and have a static route  for that specific next hop point to the given TE tunnel. To change the  next hop for a specific VRF, you would use "bgp next-hop" under "ip vrf  " .


Regards

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (2 ratings)
Loading.
amabdelh Tue, 09/24/2013 - 23:38
User Badges:


Octavio
Ldp creates labels for all routes or permitted routed in the cef table, so if there are two routes to a particular destination and one of them has netter metric it will be installed and a label will be created for this one only. To have more than one LSP for the same prefix, you need to have ECMP first in your cef
Sent from Cisco Technical Support iPhone App

Octavio Alfageme Fri, 09/27/2013 - 00:06
User Badges:

Thank you Amjad,


Even though you refer to LDP, I see your answer applicable to RSVP-TE signalled LSPs too. Thank you.


The only thing that I've been unable to find is if  there is any way to manually impose the RSVP-TE signalled LSP used by a L3VPN. Do you know how to do it?


Thank you


Octavio

Correct Answer
Harold Ritter Fri, 09/27/2013 - 05:42
User Badges:
  • Cisco Employee,

Hi Octavio,


The cost of the RSVP TE signalled LSP is determined by  the best IGP cost to the tunnel tail end. In your scenario, both LSPs  would have the exact same cost, which would be the IGP cost via the 10g  interface. Assuming you are using "autoroute announce" to inject traffic  into the LSPs, the l3vpn traffic would be load balanced over the two  links by default.


You can change the default behavior  to prefer the LSP going over the 10gig link by using "tunnel mpls  traffic-eng autoroute metric". You could also load balance between  unequal bandwidth links using "tunnel mpls traffic-eng load-share".


If  you want to manually set the tunnel used for a given VRF, you can do  this by using a different next hop for that VRF and have a static route  for that specific next hop point to the given TE tunnel. To change the  next hop for a specific VRF, you would use "bgp next-hop" under "ip vrf  " .


Regards

Octavio Alfageme Sun, 09/29/2013 - 23:26
User Badges:

Thanks a lot, Harold. This is precisely what I was looking for. Manually setting the tunnel for a given VRF seems burdensome. I have to tell my Cisco's SE that we need a procedure like the one for pseudowires, for instance. ;-))


Once again, thank you for your invaluable help.


Regards


Octavio

Correct Answer
Harold Ritter Mon, 09/30/2013 - 06:33
User Badges:
  • Cisco Employee,

Hi Octavio,


I realize that having to specify the alternate next hop on all VRF can be cumbersome but this way you have a greater granularity. For instance, you could potentially have a different next hop for each and every VRF, which in turn could use a different tunnel. Not scalable but flexible.


If what you are trying to achieve is to use the same tunnel interface for all VRFs, you could simply establish your VPNv4 session using a separate loopback address and have that loopback address reachable via a tunnel interface via a static route.


Regards

Octavio Alfageme Mon, 09/30/2013 - 23:02
User Badges:

Thank you, Harold. You are right. The second option adds scalability and the first one allows you to manage some exceptions you might have. I really appreciate your comments.


Thank you very much for your help


Octavio.

Actions

This Discussion