The MPLS WG Archive

["David Allan"<dallan@nortelnetworks.com>]

Title: RE: Clarification on today discussions

George,

There's actually multiple scenarios to consider. If the network is LDP driven, then the PW endpoint is not expected to have to take any action to perform fault recovery, the network will recoverge, and if it is possible to for the network to restore connectivity it will just happen.

If the network is ER-LSPs because I needed to engineer some applications or require high availability, I may also provide recovery paths for the service that the PW layer does not know about, and the MPLS layer is expected to take some proactive action to restore service on failure (e.g. protection switching, re-balancing multilink...whatever) and this may frequently be in the form of the LSP ingress needing to do somthing while the rest of the network does not. In these recovery scenarios, the end point controls it's MPLS destiny. This is a different application of the e2e principle and has similar validity. Recovery takes place at a layer that can do somthing about it. Availablity measurement is not the only function enabled here.

Pushing this functionality into the PW layer does not make sense, it is not expected to be aware of the transport attributes. It is a passive partner.

cheers
Dave

> -----Original Message-----
> From: George Swallow [mailto:swallow@cisco.com]
> Sent: Wednesday, March 20, 2002 6:07 PM
> To: neil.2.harrison@bt.com
> Cc: Shahram_Davari@pmc-sierra.com; mpls@UU.NET; swallow@cisco.com
> Subject: Re: Clarification on today discussions
>
>
> Neil -
>
> > NH=>Well, that means one has to question the viability of
> what is happening
> > in PWE3 and whether anyone is intending to carry ATM, or
> anything else like
> > Sonet (since exactly the same principles apply here), over
> MPLS....or more
> > accurately any PSN.  It can't be argued both ways, which
> seems the current
> > case. So what's the answer from the ADs on this?
>
> I quite agree that an operator who uses MPLS to extend and or replace
> an ATM or FR service may well desire a level of OAM in the MPLS
> protion that matches what exists (or is expected) elsewhere.  But
> following the end-to-end argument, this should be built into PWE3
> solutions.  The vast majority of MPLS which is not offering such
> services should not be "strongly recommended" to adopt such methods.
>
> ...George
>
> ==================================================================
> George Swallow       Cisco Systems                  (978) 497-8143
>                      250 Apollo Drive
>                      Chelmsford, Ma 01824
>
>