Can we implement mpls(MP-BGP) in this network setup

Unanswered Question
Sep 10th, 2009

Let me explain you the network. There are two 7200 series routers at each core location(we call it DC).One is to be considered to be primary and the other secondary. At each loaction there is one has 2Mbps MPLS based link from the service provider to the primary and a ISDN line to the secondary.Means if the primary 2Mbps link fails it shld fail over to ISDN.

From the service provider end, we have BGP public AS number.

Now my question is:

Can we implement MP-BGP in this network ?,othe prime requirement is to segregate the networks.

If yes will the service provider BGP will convey(carry) all the VRF information to each PE router.This is the primary concern, because not sure how it works on BGP confederation boundaries.

One more concern is that:

If there is another PE under PE(PE1->PE2).Do i need to define all the neighbours under PE2 OR is there another way of doing it?(beacuse i need to decrease on bandwidth consumption as it will create individual TCP sessions for each neighbour).I think route reflector is the way to do it but i need to confirm.

Another concern is bandwidth consumption:

How much bandwidth does a MP-BGP eat up?

Happy to help!

Nishant

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (1 ratings)
Loading.
Giuseppe Larosa Fri, 09/11/2009 - 02:07

Hello Nishant,

I try to provide some answers:

about MPLS and MP-BGP: setting up the MP BGP sessions is not a problem.

The issue here is how to provide a forwarding path.

First of all it depends from the nature of of the MPLS link:

if the service provider is giving you an MPLS L3 VPN service you are already in a VPN from the provider point of view.

Supporting MPLS services including your own VPNs is possible in the Carrier Supporting Carrier model where you and the service provider should cooperate.

Another possibility given the relative slow bandwidth and type of platforms you use is to use GRE tunnels between your devices with MPLS enabled on them: in this case service provider is not involved.

Finally if the current service is a L2 service like a leased line you are free to do what you want. (this happens if the two PE nodes share a subnet over the link).

To be noted you can configure also MPLS VPN access over ISDN without using L2TPv2 we were able to do this.

You need to consider what to do when primary link fails.

About BGP sessions BW usage: in a steady network BGP uses very low BW just to exchange the BGP keepalives that are small packets so you shouldn't be worried by this.

However, the suggestion is to use QoS to protect BGP traffic putting them in a dedicated queue in a CBWFQ scheduler.

Hope to help

Giuseppe

nishantmj Fri, 09/11/2009 - 15:52

Thanks Giuseppe!

You have been of great help to me.

Let be more specific about details:

First Issue:You need mean that if it is MPLS L3 VPN service we need to implement GRE tunnels between endpoints, on which we need to establish MP-BGP sessions.

In the other case where service provider agrees then, they need to enable VpnV4 address family and need to activate send communtiy vpnV4.

Correct me if am wrong on this?

Let me take one by one issues.because i am not sure if L3 MPLS VPN service might work on ISDN line.If you have some document supporting this, would be really appreciated.

Happy to help

Nishant

Laurent Aubert Fri, 09/11/2009 - 18:02

Hi Nishant,

If it's a true hub&spoke topology, you can use VRF-lite with one GRE tunnel per VRF and use mGRE tunnel on the DC to aggregate all the remote tunnels if they share the same VRF.

Don't know if it it will make your life easier as it depends of the number of VPNs you need.

HTH

Laurent.

nishantmj Fri, 09/11/2009 - 18:20

Hi Laurent

That was a good suggestion!However the number of vrf's i am considering over here might definately be more than 15.

That makes this even more tedious if we consider the redundancy(ISDN line) into consideration.You can imagine i need to create tunnels for each of the VRF.Even if i am successful in doing this, redundancy will be left out.

Happy to help!

Nishant

Actions

This Discussion