[rc] Layer 2 MPLS - which routing protocol to use? WAS RE: anyone alive out there?

[jon.hartman at verizon.com]

I guess my only concern with that would be that since many topologies use
eBGP externally and OSPF internally, to use BGP for your site-to-site
routing, you'd have to redistribute between iBGP and OSPF and raise the
complexity proportionally. If you're just talking about a few branch
offices connected to HQ that might be OK, but when you're talking about
connecting major data centers, those concerns are elevated.

I have to admit, while I know a bit about the technology, I know
little/nothing about the practical implementations yet of MPLS. So, why is
it that OSPF has issues? Could any issues not be overcome by also playing
with it's timers? Granted, you may have some multicast issues, but I know
you can also manually configure OSPF neighbors.

-Jon

-----------------------------------

Date: Mon, 24 Mar 2008 03:58:00 +0000 (GMT)
From: Chris Morrow <morrowc at ops-netman.net>
Subject: Re: [rc] Layer 2 MPLS - which routing protocol to use? WAS
	RE: anyone alive out there?
To: Joseph Jackson <jjackson at aninetworks.net>
Cc: "routing-chat at puck.nether.net" <routing-chat at puck.nether.net>
Message-ID: <Pine.LNX.4.61.0803240354420.14506 at arb-h2.bcf-argzna.arg>
Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed



On Sun, 23 Mar 2008, Joseph Jackson wrote:

> Well lets talk about routing then. :)  We are planning on moving away

okey-dokey!

> from some DS3's to a layer 2 MPLS setup.  We have a little debate going 
> on which routing protocol we should use to connect our sites.  Over the 
> DS3 we use OSPF but I've been pushing to run iBGP since the routers also

> do eBGP.  Of course fast failure detection is a major goal so I'm just 
> wondering what others might think about that.
>

I've seen a bunch of private-ip/mpls provider's give their speeches to 
potential customers, in almost all they spout off about how they 
can/do/will support routing protocols like:

1) bgp
2) ospf
3) eigrp
4) rip

I almost never hear that things other than bgp work reliably... after much

digging often the others are 'well we have plans in Q2...' or 'we have one

customer doing this...'. Never really encouraging :) BGP and linking the 
neighbor-down to the route-withdrawl seem to do well enough. How fast do 
you need convergence to happen if the link drops? Have you tested to be 
sure that it won't drop fast enough?

-Chris