Time to implement L3VPN? Design considerations for small core of MX5 and EX4200 / EX4550

‎11-19-2014 06:50 AM

Hello all,


My current topology looks like "Sites tripod topology current". It has served us well for a good while. The current rationale behind this is, amongst other things;
- We want to use the EX stacks as default GW's for the servers at each site (,, due to their redundant nature and NSSU.
- Only use OSPF in the EX's since BGP requires a software license.


In the months to come, I have a need to setup BGP to external networks over dedicated, redundant WDM connections. Please see figure 2. These are to be private, ie. non-internet and need to be transported across the network in a redundant way. If I was to do this in the current fashion, I would have to establish separate VRF:s over separate VLAN:s across the PE routers (where I terminate the WDM's) and over the EX-stacks. That would be a lot of configuration and a lot of IP-addressing. For one "circuit" it would probably be manageable but my need is for at least three of these.


I have no previous experience in MPLS but from what I've been able to grasp so far, this is a common use case for L3VPN's? If that's the case, what is a good starting point and what considerations should be made? To what extent is an MPLS L3VPN topology able to coexist with the current setup, of course I'd like to change as little as possible in the current pure-IP topology and the bit by bit (and as I learn along the way) move more and more from L2 and IP to MPLS. The EX's would in this case clearly be P routers, and the MX's would be PE routers, but I'd like to keep using the EX's as default gw's so they would have to do a bit of both (P and PE) - how do I go about that?






Re: Time to implement L3VPN? Design considerations for small core of MX5 and EX4200 / EX4550

‎11-24-2014 09:52 PM

1)You can run INTER_ AS MPLS VPN  option B with you carrier . if carrier provides the BGP peering at all sites or minimum 2 sites for redundancy , You will run MPBGP with you carrier  and import / export the extended communities for you DC / Cloud customers .


2 ) if you choose to have the 3 different carriers for all three sites &  carriers might need to run the option B or C amongst them to pass your target communities in case of link failures. This is cumbersome way ( not recomended) . this is like carrier supporting carrier . 


3) you can have the VPLS Mesh also for the layer 2 customers .