The MPLS-OPS Archive

[Mathew Lodge <mathew@cplane.com>]

At 03:24 PM 1/23/2002 -0800, Daniel Kharitonov wrote:
>L2 VPNs are a good solution to one problem and a bad
>solution to another. Specificly, "fixing" the need for
>the usable L3 VPNs w/L2 VPNs is a bad trend.
>First, L2 VPN needs n^2/2 tunnels to be configured for
>n customer endpoints. Next, hauling full L2 (read,
>Ethernet) traffic is unnecessary and also brings the
>MTU problem on the table. You will be very surprised
>to know, how many brand-name routers and switches are
>simply incapable of taking ethernet packets larger
>than 1536 bytes.

>In a word, I am a strong advocate for the L3 VPNs
>without client's routing tables propagated through
>ISP.

To me, the L2/L3 MPLS discussion is like the "switching vs. routing" 
argument of some years ago. The answer was, of course, "both"... but that 
never makes for a good headline :-)

Specifically, Martini is great for simple point-to-point L2 (a service 
concept that has been selling very well for years in the guises of private 
line, frame relay PVCs and ATM PVCs), and Lasserre-Kompella is great for 
small, simple VPNs (where the n^2 mesh is not an issue). RFC2547 is good 
for large, complex VPNs, but too complicated for smaller VPNs or 
point-to-point. Note that "simple" or "complex" refers to the end 
customer's view of the service. The interesting thing is that these 
technology solutions seem to map fairly well to small, medium and large 
enterprises, respectively, so there's a sort of "natural" segmentation of 
the marketplace.

As far as MTU issues, many switching vendors are requiring a new MPLS card 
or a switch/router card upgrade to get MPLS capability, and are taking the 
opportunity to fix any MTU problems with that upgrade. A limited set of 
card upgrades seems palatable to the SPs that I've talked to.

So, while I agree that IP routing over L3 VPNs could be easier, I believe 
that there are going to be several different L2 and L3 MPLS services that 
will appeal to different segments of the end customer marketplace. This 
would be a path of lower resistance for a service provider rather than 
requiring end customers to adopt a new "multihop aware" routing protocol.

Cheers,

Mathew


| Mathew Lodge                 | mathew@cplane.com     |
| Director, Product Management | Ph: +1 408 789 4068   |
| CPLANE, Inc.                 | http://www.cplane.com | 

-------
The MPLS-OPS Mailing List
Subscribe/Unsubscribe:  http://www.mplsrc.com/mplsops.shtml
Archive: http://www.mplsrc.com/mpls-ops_archive.shtml