[afnog] Problem bringing up MPLS on disimilar interfaces

[Mark Tinka]

On Monday 06 October 2008 17:48:24 John Tully wrote:

> You are right.  Though if you only have interfaces that
> support 1500 byte MTU, then the RFC4623 implementation
> will allow you to get around all these problems --
> considering that the MTU for the MPLS is set up properly.

Well, RFC 4623 defines fragmentation & reassembly for pw's. 
Particularly, this RFC is generalized, so it applies to 
various methods that may be used to signal a pw, i.e., 
MPLS, L2TPv2 and L2TPv3.

Its emphasis is mostly on pw's created for customers 
utilizing the service providers network as an emulated 
circuit-switched backbone. Traffic would get fragmented as 
it enters the network over the attachment circuit, and then 
reassembled as it exits the network toward the customer.

As fragmentation & reassembly is taxing to router resources, 
I'd generally not recommend if it can be avoided, i.e., 
this RFC defines a short-term solution.

But most important to note is that this solution does not 
replace the need to support at least baby giant, or at 
best, jumbo frames, on Ethernet networks in order to 
successfully implement MPLS, in and of itself, before pw's 
can be switched across the network. This solution only 
helps if the resultant MTU, after the customer's pw has 
been factored in, exceeds the backbone's core MTU value.

Cheers,

Mark.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 835 bytes
Desc: This is a digitally signed message part.
URL: <http://afnog.org/pipermail/afnog/attachments/20081006/b28e599b/attachment-0002.bin>