cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
1916
Views
30
Helpful
6
Replies
MMCDOW14
Beginner

OSPF over L3 MPLS

Was just wondering when using an MPLS Layer 3 backbone, when connecting the CE to the PE using OSPF will they both be in Area 0?

So each site (site1, site2, site3..etc) connecting will be config in Area 0 and therefore these routes will show up as "O" in each of the sites routing tables.

Or is the MPLS backbone routers configured as Area 0 and then each of the sites can be configured in Area 1 and will then show up as "O IA" in each sites routing table.

Also do the site routers (CE) and the MPLS backbone router (PE) have to be in the same OSPF ID proccess? and do all the sites have to be in the same OSPF ID proccess to connect? e.g (site1) router OSPF 100, (site2) router OSPF 100

I'm not interested in the config from the SP side i am just trying to design the sites LAN design and connect them through MPLS.

Don't know if anyone will understand that horribly described question but cheers for any help.

6 REPLIES 6
blau grana
Rising star

Hello Michael

- by default OSPF routes will be redistributed from BGP as LSA type 5.

- on PE-CE peering routers ID process has to match (requirement for establishing neighborship)

- no, not all sites have to have same process ID, process ID has to match only on neighboring routers

I would suggest to read section from MPLS Foundamentals where is described OSPF as PE-CE protocol.

MPLS Fundamentals, Chapter 7 -> PE-CE Routing Protocols -> OSPF page 209

http://www.amazon.com/MPLS-Fundamentals-Luc-Ghein/dp/1587051974

Best Regards

Please rate helpful posts and close solved questions

Best Regards Please rate all helpful posts and close solved questions
Giuseppe Larosa
Hall of Fame Master

Hello Michael,

OSPF process-id must match on PE side of each VRF site, because the OSPF process-id is sent as a BGP extended community in MP BGP VPNv4 advertisement of a route learned from CE. This is part of the emulation of an OSPF super backbone that actually is still MP BGP communication.

In this way O routes an O IA routes learned by CE node will be rebuilded as O IA on remote VRF site.

O E1 or O E2 routes will be seen as O Ex routes on remote VRF site.

The only way to see O routes between sites is the use of sham-links that are a sort of specialized virtual links between PE nodes. Their use should be limited to topologies where there are some direct inter-site links.

The OSPF process-id on the CE nodes can be different as actually in OSPFv2 this information is not exported in Hellos or LSAs.

So from the CE side you are free to use the OSPF process-id that you like. The OSPF areas that can be used are more important I would say that either the PE node  either the CE node can act as ABR connecting a non zero area to backbone area.

Keeping the ABR role on CE node can give more control on what routes are propagated to the MPLS backbone and what routes are taken from MPLS superbackbone and injected in local non zero area. (this is valid for internal routes only as LSA type 5 cannot be filtered).

Hope to help

Giuseppe

           

Thank you both for your fast and detailed replies.

I'm currently studying CCNP and MPLS is only lightly covered so had a lot of questions regarding it.

I asked the question because i am currently doing a uni project (so only concerned about the client configuration for now)  and MPLS is the WAN technology that i what to use for site to site connections.

I have attached a photo of the ospf layout, wondering if that would be acceptable? So the MPLS PE routers would be area 0 configured by the Service provider?

Thanks again for any help

Hello Michael,

You can use that design, but routes from core segment of each site will not be in routing tables on PEs. You will need to configure virtual-links.

For example: routes from area 9 will not arive to PE7&PE8 so they will not be redistributed into BGP and advertised to other branches.

To overcome this behaviour you will need to do this:

on CE:

area 4 virtual-link IPaddr-OSPF_RID_of_PE

on PE:

area 4 virtual-link IPaddr-OSPF_RID_of_CE

Best Regards

Please rate helpful posts and close solved questions

Best Regards Please rate all helpful posts and close solved questions

Hello Michael,

>> I have attached a photo of the ospf layout, wondering if that would be acceptable?

No, it is not a good design as explained by blau grana it would require the use of virtual links. This is something to be avoided as a permanent solution in real world.

I strongly recommend to move PE-CE links in area 0 to correct the design in this way you will have an area 0 and all non backbone areas directly connected to area 0 as it is required in a sane OSPF design.

Edit:

don't be confused by the term MPLS superbackbone you are still dealing with standard OSPF following the usual design rules.

Hope to help

Giuseppe

MMCDOW14
Beginner

Thanks to you both!

Yeh I dont know why i didnt realise that the design would require virtual links (My mistake).

Will make all the CE-PE connections in area 0.

Thanks for clearing this up.

Michael

Content for Community-Ad

This widget could not be displayed.