[afnog] Juniper VPLS

[Mark Tinka]

On Tuesday, May 14, 2013 09:09:55 AM ibtisam jamal wrote:

> Is there any way that a loop can be caused by just
> configuring the topology i indicated earlier ?

The VPLS domain uses a Layer 3 control plane (BGP or LDP) to 
build it. To be thorough, loops are always possible here if 
routing is poorly done, i.e., an IP routing loop.

As a service, loops are possible at Layer 2, when your 
customers connect to you with switches and decide to 
unintelligently loop them. 

Apart from those two scenarios, I can't see where VPLS would 
cause a loop on its own.

Personally, I don't like VPLS; I think it's a solution 
looking for a problem. The majority of data networks today 
are IP-based, in which case an l3vpn (which provides an any-
to-any topology) will be ideal. In most cases where a Layer 
2 transport has been required, point-to-point or point-to-
multipoint EoMPLS works much cleaner than VPLS, e.g., old 
ATM machines still running X.25, hotel reservation systems 
still based on IPX, e.t.c.

Two use-cases where VPLS "can" make sense are when used as 
part of the provider core, and not offered as a service:

	1. Backhaul for PPPoE-based or DHCP-based broadband
	   services. This replaces the traditional L2TP
	   tunnels that have been used to backhaul broadband
	   connections from customer modems to a BRAS/BNG.

	2. Mobile backhaul core, as 3G/4G IP networks are
	   typically based on a flat Layer 2 domain so as to
	   support IP mobility across a GSM network.

If it's not for those two cases, I'd steer clear of VPLS. 
Too much hassle for too little benefit.

Mark.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
URL: <http://afnog.org/pipermail/afnog/attachments/20130514/6f89dd96/attachment-0001.sig>