[j-nsp] VPLS design - dual homed

[Luca Salvatore]

Thanks for the links.
As I mentioned to Chip - my MX routers will be directly connected via fibre.
Do the rules still apply for this type of topology?

Luca


-----Original Message-----
From: Ben Dale [mailto:bdale at comlinx.com.au] 
Sent: Tuesday, 30 October 2012 11:17 AM
To: Luca Salvatore
Cc: juniper-nsp at puck.nether.net
Subject: Re: [j-nsp] VPLS design - dual homed

Hi Luca,

> MX1---------darkfibre----------MX2
>  |                              |
>  |                              |
> MX3---------darkfibre----------MX4
> 
> So above you see that there are dual  links which will create a loop.
> 
> How does VPLS handle these types of topologies?  Do I need to just use spanning tree and have one link in blocking?
> Or perhaps use MSTP and send some VLANs down one link and some down the other?
> I will be spanning around 1000 VLANs across these links. They are 10Gb links, so it seems a shame to have one in a blocking state.

VPLS will handle the above just fine using full-mesh and split-horizon - eg: A PE will never forward frames received on one LSP to another LSP.

The dual-homing from your MXs (assuming they are PE) to your CE/Edge devices is where you'll need to be concerned about loops.

There are mechanisms built into VPLS to deal with this though.  A quick search on "junos vpls multihoming" provides plenty of detail on the knobs you'll need:

http://www.juniper.net/techpubs/en_US/junos12.2/topics/example/vpls-multihoming-convergence-example.html
https://www.juniper.net/techpubs/en_US/junos12.2/topics/task/configuration/vpls-multihoming-bgp-signaling-solutions.html

I've not actually tried dropping in MC-LAG into a VPLS but I can't see why this wouldn't work.

Ben