[j-nsp] Policies for VPLS traffic

[Rafał Szarecki]

Monika,

Let's look for simple cast: 1 ingress PE one egress PE, and multiple ECMP
path between them. I will not elaborate how to get 2 or more equaly good
transport lsp. But this is possible e.g with LDP but not only.

The from ingress PE perspective there are two equaly good pathe to BGP
NEXT-Hop revived with VPLS/L2VPN NLRI. So, multiple next-hops (physical)
will be installed in FIB. This is exactly like for L3VPN secnario.
Then ingress PE has to decide how to balance incomming traffic. This is done
base on hash function. the key for this function is ingres IFL, source and
destination MAC.
Se picture below.
     P--P
    /      \
PE       PE
   \       /
     P--P


The other problem is when You have multiple equal path between P and P
routers, somwhere inside cloud. Then P router do not distinguish VPLS form
other traffic types -> make hase base on ingress IFL and few top labels (two
of them).
PE--P====P--PE


And last consideration - multihoming. _This is not recomended_, but if you
tweak BGP to instal multiple path and configure both PE with same site ID,
then load ballancing should also be possible... Never try.
2007/7/10, Monika M <monika.vpls at gmail.com>:
>
>
> http://www.juniper.net/techpubs/software/junos/junos83/swconfig83-policy/id-10522406.html#id-10522406
> Above section explains about configuring the Load-Balance Per-Packet
> Action.
> It talks about installing ECMP routes in the forwarding table.  It is all
> perfectly fine. It also indicates that "You can configure per-packet load
> balancing to optimize VPLS traffic flows across multiple paths.".
> How do we get multipaths in a VPLS cloud? We establish pseudo wire to
> different edge routers and L2 flood/forward the traffic based on the FDB
> table.
> Even for redundancy if we multi-home, only one PE is chosen based on the
> BGP
> selection algorithm. If we allow the traffic to flow in both the nodes,
> won't that be against split horizon rule?
>
> Regards,
> Monika
> _______________________________________________
> juniper-nsp mailing list juniper-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/juniper-nsp
>



-- 
Rafał Szarecki JNCIE
+48602418971