shivlu jain Tue, 09/08/2009 - 21:56

Hi Mukar

Definately RT is required. RT is an extended BGP community. With the help of RT the communication is possible in single VPN. If you are not using RT, In that case how you will receive the routes of same vpn on different PEs. The routes are imported from different PEs for same vpn in vrf occurs only with the help of route target. So it is mandid to add route target in that.


shivlu jain

Mukarram Jah Raheel Thu, 09/10/2009 - 10:40

Hi Shivlu,

When we say two sites are in same VPN, does it mean that they are sharing routes thru RT...

What are the parameters that need to match for two sites to be in same VPN...

Harold Ritter Thu, 09/10/2009 - 10:43


The import RT need to match the export RT for the route to be imported.


Giuseppe Larosa Wed, 09/09/2009 - 01:50

Hello Mukarram,

as Shivlu has explained route targets are used to decide what vpnv4 prefixes to import in a specific VRF.

without route targets you just get a collection of isolated VRF sites not a working MPLS VPN.

Advanced usage of RTs allows for extranets or hub and spoke topologies.

Being a BGP extended community like standard community multiple RTs can be associated to a single vpnv4 route (for example for implementing extranets) but at least one has to be present.

any-to-any connectivity between VRF sites require only a single RT.

It's also important to understand differences between RT and RD:

RD is prepended to the IPv4 prefix to build the vpnv4 prefix sent in address-family vpnv4.

RD is there to allow for address overlapping in different or same PE node: with distinct RDs we can distinguish between two same IP subnets in two different VRFs.

RTs are associated as extended BGP communities to the vpnv4 routes.

RTs are used to build the IP routing tables of each VRFs and provide IP connectivity between VRF sites.

Often simple examples use the same value for RD and RT because both are 64 bits and they can use the same notation but they are different.

Also RDs can be different in each VRF site without causing problems.

Actually in some complex topologies with multi PE connections the usage of a different RD is mandatory to achieve the desired effects in terms of fault tolerance and load balancing.

Hope to help


Mukarram Jah Raheel Thu, 09/10/2009 - 08:19

Following is an extract from MPLS Fundamentals:

"If RDs were just used to indicate the VPN, communication between sites of different VPNs would be problematic. A site of Company A would not be able to talk to a site of Company B because the

RDs would not match. "

This makes me think :

1. RD needs to be same for sites of same VPN. RD is used to differentiate between IPv4 prefixes... and there are no chances of overlapping prefixes within and organization..

2. RDs are not required to help the sharing of routes between sites of same VPN...

Next Extract:

" The concept of having sites of Company A being able to talk to sites of

Company B is called extranet VPN. The simple case of communication between sites of the same

company-the same VPN-is called intranet. The communication between sites is controlled by

another MPLS VPN feature called RTs."

Giuseppe Larosa Thu, 09/10/2009 - 13:01

Hello Mukarram,

>> RD needs to be same for sites of same VPN.

this is not true I contributed to develop a semi automatic system for MPLS VPN configuration for a service provider.

We have chosen to have unique RDs in each VRF site.

This is very important for multihomed site where using a different RD makes routes of both PE nodes to propagate through Route Reflectors.

>> The communication between sites is controlled by

another MPLS VPN feature called RTs.

this is correct

Hope to help



This Discussion