[afnog] MPLS traffic ....

Mark Tinka mark.tinka at seacom.mu
Sun Apr 1 13:30:36 UTC 2012


On Thursday, March 29, 2012 05:24:26 PM ibtisam jamal wrote:

> Traffic on the fiber is reaching 99% utilization ,i was
> wondering if there is a way ,to configure the mpls such
> that when utilization on link 1 reaches may be 90%
> ,additional traffic moves to link 2.Because we do not
> have possiblility of getting a second fiber anytime
> soon.

A couple of options:

	a) For a non-MPLS solution, have you checked to make
	   sure your IGP is seeing these 2x paths as equal-
	   cost for traffic flowing across it? Of course,
	   the challenge here is to with the hosts sitting
	   behind both links that have to view them as
	   equal-cost.

	b) For an MPLS solution, MPLS-TE is your best bet.
	   Since you're already running MPLS, this shouldn't
	   be a huge deal to add. You have 3x options for
	   MPLS-TE:

		i) Static routing into the MPLS-TE.

	       ii) IGP Shortcuts (a.k.a. Autoroute
	  	   Announce).

	      iii) Forwarding Adjacencies (FA).


If you have a simple network, I'd be okay suggesting option 
i), above. What this means is you would write static routes 
on the routers that are connected to these 2x links, and 
make the next-hop the MPLS tunnel. However, you'd need to 
redistribute these routes into your IGP so that downstream 
routers can see the path via the routers connected to the 2x 
links (IGP redistribution is necessary because the routes 
will previously have been in your IGP prior to you making 
them static routes - remember, static routes are better than 
IGP routes, so without redistribution, you'd cut off access 
to those routes for downstream routers).

If your network is fairly large (or if, like me, you want to 
scale the solution from Day One), I'd recommend option ii), 
above. IGP Shortcuts (commonly also known as Autoroute 
Announce) is there the MPLS tunnel is used as a one-hop 
interface. Any routes crossing the tunnel are automatically 
populated into your IGP, making the tunnel a viable path 
across the network. This is much easier and known to operate 
rather well. One caveat, in Junos platforms, IGP Shortcuts 
are not support if CSPF is disabled (unlike in IOS, IOS XE 
or IOS XR).

While FA's are a reasonable option, I generally wouldn't 
recommend them unless your topology requires it. The reason 
I say that is because FA's really "mess up" the actual 
physical topology, and in most cases, traffic can bounce 
between routers multiple times before actually going the 
right way. That said, FA's are great if you need to keep the 
number of TE tunnels to a minimum, and can't afford to run 
tunnels between every head and tail.

Hope this helps.

Cheers,

Mark.



More information about the afnog mailing list