route distinguisher question

[Tahir Ali]

Hi,

I need to clear some concept regarding rd, i had a topology from one of the COD in which instructor connects 2 CE sites to each other using VPN_A but at one PE the configured RD was different and at another PE RD was different. The thing which connected both the CE sites was the same import/export RT.

Previously i knew that RD should be unique per customer then y did the instructor configured different RD for the same VPN. i.e VPN_A.

Next question is that is the Vrf name i.e VPN_A in this case locally significant on the router?

[Tahir Ali]

what about connecting 2 customer sites of the same customer at different PEs...will the RD for that customer be same ? In my scenario the RD for the same customer was different at each PE with vrf name VPN_A at both PE-CE interfaces?

[lee.reade]

Hi,

The RD does not need to be the same for a customer VPN across any number of PE's, this can any value you wish.

Its only purpose is to make the customers IPV4 prefix unique within the MPLS core, (across all PE's) so that customers can have the same internal IP addressing, otherwise we would need somekind of NAT involved.

The RD appends 8 bytes onto the existing 4 byte IPV4 prefix, to give a unique 12 bytes VPNV4 address.

HTH

LR

[lee.reade]

Hi,

The RD is only used to make a specific customer learned prefix unique within the MPLS core, ie you have two vpn customers that each use the internal network 192.168.1.0/24, then by assigning each customer vrf a different RD, then these two VPNV4 routes will now be different from each other.

There is no requirement to use the same RD for a specific customer vpn on two PE routers.

The VRF name is case sensitive and is only locally significant.

HTH

LR

[mohammedmahmoud · Accepted solution]

Hi,

RD is used for the sake of address space overlapping (by constructing unique VPNv4 routes; in the RD:IPv4 format) and also as a VRF identifier (on the PE and the RR the VRF is identified by the RD, accordingly the RD is a mandatory attribute for the VRF configuration), and it is only significant on the PEs and RRs if exists.

Accordingly, the RD for a certain VRF needs not be the same allover the MPLS domain, RT will be used to import and export prefixes between VRFs on different PEs, lets discuss how it really happens in order for you to be able to connect the pieces together; MP-BGP exchanges the VPNv4 (RD:IPv4) prefixes between the PEs (P routers are not VPNv4 aware) not IPv4 routes - When a PE receives a VPNv4 route it discards the RD attached to it and attaches its local RD (according to the VRF that the route is going to be injected into according to the attached RT and the local VRFs import RT) - Since for the local router routing table database the VRF is identified via the RD prepended to the IPv4 route to construct the VPNv4 routes.

With the same logic the VRF name is locally significant to the PE (it maps to a RD on the local PE, while having no significance on a RR, the RR only understands RDs), it is introduced to make things user friendly, rather than using the RD solely to identifying the VRF when doing configurations and with show commands - Most commonly the same VRF name is used on all PEs (with each PE having its own RD for this VRF in the format IP:number, where IP is the PE loopback and the number is a locally unique number).

I hope that i've been informative.

BR,

Mohammed Mahmoud.