On a transit service provider, we have as a principal target give to the client ISP the access of all world network available content as quickly as possible. The other target is to provide the maximum simplicity of the necessary actions to provision new clients, giving free time to NOC agents to realize the monitoring and preventive actions to better control the network.
Problem:
Typically, the first service offered by an ISP to your clients when it enters at Transit Market is the transport with L2VPN, where it configures a tunnel between the client access and your own BGP router, configuring the transit only on it.
This solution will work very well while the complexity of the network is low and the distance between clients and them with BGP edge router remains low.
When this network increases its complexity, it starts offering high geographical availability and long distances between its POPs, the provisioning of new customers, the management and maintenance of this network exponentially increases the time loss of NOC agents.
Another issue of this configuration is that all the traffic exchanges between the clients need to pass through BGP edge router, even if there are shorter routes on the MPLS backbone.

[PE2]
#
interface GigabitEthernet0/0/2.20
vlan-type dot1q 20
mpls l2vc 10.99.99.1 20
#
interface GigabitEthernet0/0/3.10
vlan-type dot1q 10
mpls l2vc 10.99.99.1 10
#
[PE1]
#
interface GigabitEthernet0/0/1.10
vlan-type dot1q 10
mpls l2vc 10.99.99.2 10
#
interface GigabitEthernet0/0/1.20
vlan-type dot1q 20
mpls l2vc 10.99.99.2 20
#
[BGP ISP]
#
interface GigabitEthernet0/0/0.10
vlan-type dot1q 10
ip address 192.0.2.2 255.255.255.254
#
interface GigabitEthernet0/0/0.20
vlan-type dot1q 20
ip address 192.0.2.0 255.255.255.254
#
[BGP CLI1]
#
interface GigabitEthernet0/0/0.10
vlan-type dot1q 10
ip address 192.0.2.3 255.255.255.254
#
[BGP CLI2]
#
interface GigabitEthernet0/0/0.20
vlan-type dot1q 20
ip address 192.0.2.1 255.255.255.254
#
Solution:
The better architecture to ensure the easy growth of this ISP is the adoption of MPLS VRF (L3VPN). With this configuration, all the PE routers of this MPLS network works like a big unique router, that way allowing that the traffic passes through the shorter way signalized by IGP.
In addition, VRF enables more simplicity on geographical redundancy activation of edge BGP, because it allows adding new edge BGP routers without the necessity of creating new L2VPN for each client.

[PE2]
#
interface GigabitEthernet0/0/2.20
vlan-type dot1q 20
ip binding vpn-instance VRF_IP
ip add 192.0.2.4 255.255.255.254
#
interface GigabitEthernet0/0/3.10
vlan-type dot1q 10
ip binding vpn-instance VRF_IP
ip add 192.0.2.2 255.255.255.254
#
[PE1]
#
interface GigabitEthernet0/0/1.30
vlan-type dot1q 30
ip binding vpn-instance VRF_IP
ip add 192.0.2.0 255.255.255.254
#
[BGP ISP]
#
interface GigabitEthernet0/0/0.30
vlan-type dot1q 30
ip address 192.0.2.1 255.255.255.254
#
[BGP CLI1]
#
interface GigabitEthernet0/0/0.10
vlan-type dot1q 10
ip binding vpn-instance VRF_IP
ip add 192.0.2.3 255.255.255.254
#
[BGP CLI2]
#
interface GigabitEthernet0/0/0.20
vlan-type dot1q 20
ip binding vpn-instance VRF_IP
ip add 192.0.2.5 255.255.255.254
#



