How is Route target working?

[annayuzhao]

If route target export 100:1 on one PE1 , must I config route target import 100:1 on PE2 to make them in one VPN and then the PE2 is able to receive the routes through BGP? Vice versa for the route target import value on PE1 has to same as route target export value on PE2? If that's the case, can I use diffrent value for the other direction? like 100:2 as route target export on PE2, and 100:2 as route import on PE1, as long as they are same?

[Harold Ritter]

You are absolutely right. Whatever you export on one side as to be imported on the other side and vice cersa if you want bidirectional connectivity. You can use different route target extended communities on each PE. Using differnt route targets is often required when you want to create hub and spoke type of topologies where you have a central office with a bunch of remote offices. You want the route target exported by the PE connected to the central office to be imported by all of PEs connected to the remote offices. In the other direction, you only want the PE connected to the central office to import the route target exported by PEs connected to the remote offices. This forces inter remote office traffic to transit via the central office.

Hope this helps,

[annayuzhao]

Thanks.

So what's the purpose of RD?

[Harold Ritter]

The sole purpose of the RD is to make VPN routes unique in the SP core. On the PE router, the IPv4 prefixes are prepended with the RD assigned to the VRF, from which the prefixes are exported, to create unique 96 bits VPNv4 prefixes (64 bit RD + 32 bit IPv4 prefix) . This allows different customers to use the same address space without causing any confusion as these prefixes are carried over the SP shared infrastructures.

For more information refer to Section 4.1 of RFC2547.

Hope this helps,

Hope this helps,

[annayuzhao]

Does this vpnv4 routing table only show up in the PE router, or also be in all the P routers?

If they are not in the P routers, how is MPLS lable progagate ? But if they are in the P routers, how can these P routes receive this VPNv4 routes? I thought these vpnv4 routes are only updated via the BGP peer. I guess I have some confusion here.Thanks

[ruwhite]

Typically, LDP is used to proagate the MPLS labels through the P routers. MPLS extensions to OSPF and IS-IS can also be used for this, I believe, but I'm not certain how common this is in comparison to using LDP.

:-)

Russ.W

[Harold Ritter]

VPNv4 routes are exchanged via mBGP between the PEs only. The mBGP update also carries a label (referred to a inner or service label) associated with the VPNv4 prefix. When a packet is received from a given VRF interface, the FIB for this VRF is looked up and the appropriate label is pushed in the packet. Additionally another label (outter or IGP label) is pushed in the packet allowing it to get from the ingress to the egress PE. The IGP label is associated with the BGP next hop for the VPNv4 prefix and is learnt via LDP, TDP or RSVP. Once the packet reaches the egress PE, the inner label is used to forward the packet to the appropriate interface.

Hope this helps,

[annayuzhao]

I think I almost understand here after you explained . I initially mixed the vpn routes and IGP routes. Last thing I want to double check with you is how the PE route know which next hop should use.

In the following example:

A-B-C-D--E-F(10.1.2.0)

A and F are CE routers

BCDE are PE and P routers

B and E are IBGP peer.

If B received F's 10.1.2.0 entry from E via IBGP. Now A wants to ping 10.1.2.1, how does B know what next hop it should use.My guess based on what you explained is that B knows 10.1.2.0 received from E so its next hop is C and will add the inner label of E's loopback address(which is the ibgp identifier). Is this right?

[Harold Ritter]

The label corresponding to E's loopback address is actually the outter label (or IGP label).The inner label is the one that is received as part of the VPNv4 updates and is only being used once the packet gets to the egress PE. You are correct about E's loopback address being the BGP next hop though.

Hope this helps,