Running EIGRP/OSPF inside customer MPLS Network

Answered Question
Mar 1st, 2010

Greetings, this may seem strange but i shall explain, I have been asked by a customer who already has an MPLS network in place for 10 business sites if they can run their own interior routing protocol between sites for the purpose of advertising additional routes to the internet via local DSL connections at some locations, the DSL circuits are used for access to partner/supplier extranet sites.


At present each site has one connection to the MPLS network, each router is configured with a single default static route pointing to the MPLS PE or CE interface ie: ip route 0.0.0.0 0.0.0.0 serial0/0/0


Dynamic routing between the CE and PE routers was never implemented due to the simple nature of the design.


Now im quite happy to get EIGRP or OSPF up and running instead of manually adding additional static routes to the CE routers but can anyone see a problem with doing this?


For the record the ISP wants to charge the customer if they should want to enable dynamic routing between the CE and PE routers hence why im looking at going down this route.


Regards

I have this problem too.
0 votes
Correct Answer by Giuseppe Larosa about 6 years 9 months ago

Hello Mark,

there is no real drawback the EIGRP or OSPF default route, that would be advertised by a CE node/CE nodes, would be advertised into that MPLS VPN context but only on that.

>> For the record the ISP wants to charge the customer if they should want to enable dynamic routing between the CE and PE routers hence why im looking at going down this route.

This is reasonable: more complexity is required and some resources on PE node are used like a dedicated EIGRP address-family (it was worse for OSPF at the beginning, it required a dedicated process per customer that was counting against total OSPF process count on PE node)

Actually customer IGP is emulated over MP-BGP by using appropriate BGP extended communities to carry key information like EIGRP AS number and EIGRP routes data structures so that they can be built again on remote PE nodes.

To be noted that from customer point of view the next-hop is always a PE IP address in any case being MPLS L3 VPN peer-to-peer model.

The only change would be at the site where the CE should advertise the default route: with static routes the roles would be inverted from current ones.

Instead of having a default route pointing to PE ip address there would be a static route on PE in VRF pointing to CE IP address. The CE would need some summary route for all other sites.

From the point of view of remote sites, there is little difference between implementing static routes or a dynamic PE-CE protocol because in any case there is the SP signalling plane in the middle.

If you had multihomed VRF sites, site-to-site links to manage (out of MPLS) then a dynamic protocol (OSPF for its support of sham-links) could be a more appropriate choice.

Hope to help

Giuseppe

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (1 ratings)
Loading.
Correct Answer
Giuseppe Larosa Mon, 03/01/2010 - 07:22

Hello Mark,

there is no real drawback the EIGRP or OSPF default route, that would be advertised by a CE node/CE nodes, would be advertised into that MPLS VPN context but only on that.

>> For the record the ISP wants to charge the customer if they should want to enable dynamic routing between the CE and PE routers hence why im looking at going down this route.

This is reasonable: more complexity is required and some resources on PE node are used like a dedicated EIGRP address-family (it was worse for OSPF at the beginning, it required a dedicated process per customer that was counting against total OSPF process count on PE node)

Actually customer IGP is emulated over MP-BGP by using appropriate BGP extended communities to carry key information like EIGRP AS number and EIGRP routes data structures so that they can be built again on remote PE nodes.

To be noted that from customer point of view the next-hop is always a PE IP address in any case being MPLS L3 VPN peer-to-peer model.

The only change would be at the site where the CE should advertise the default route: with static routes the roles would be inverted from current ones.

Instead of having a default route pointing to PE ip address there would be a static route on PE in VRF pointing to CE IP address. The CE would need some summary route for all other sites.

From the point of view of remote sites, there is little difference between implementing static routes or a dynamic PE-CE protocol because in any case there is the SP signalling plane in the middle.

If you had multihomed VRF sites, site-to-site links to manage (out of MPLS) then a dynamic protocol (OSPF for its support of sham-links) could be a more appropriate choice.

Hope to help

Giuseppe

Actions

This Discussion