The MPLS WG Archive

["Thomas D. Nadeau" <tnadeau@cisco.com>]

At 12:56 PM 3/19/2002 -0800, Shahram Davari wrote:
>Hi All,
>
>During today's MPLS meeting, Ping Pan raised the question that the 
>application should
>derive the solution, and that he is not aware of any application needs 
>that can't be
>met by LSP-ping and GTTP.
>
>My answer was:
>
>ATM/FR over MPLS application requires that upon any failure detection in the
>MPLS network, AIS should be sent from the egress LSR on the ATM/FR connections
>in order to suppress the downstream alarms at the terminating point of those
>connections. This can't be done with ICMP/LSP-ping or GTTP, because in 
>these solutions
>the egress is not the node that detects the failure. While in Y.1711 it is 
>the egress
>LSR that detects the failure and can correctly inject AIS in to the ATM/FR 
>connections.
>
>Talking to Ping Pan after the meeting, I realized where his comments are 
>coming from.
>His argument is that since he doesn't know of any deployed ATM/FR over 
>MPLS network, he thinks
>there is no problem to be solved.
>
>In other words he thinks we should not try to fix a problem which has not 
>occurred
>in the real world yet. Although this might be an acceptable approach to 
>some folks/companies,
>this is not simply true for many others. In my opinion anticipation of the 
>problems and
>providing solutions upfront is a sound engineering practice and leads to 
>easier adaptation
>of the technology in shortest amount of time without frustrations.

         This is a problem that can be solved using the existing tools. 
Same for the PWE
case.  If you set up a bidirectional ping (one on each PE), then one or 
both ends of the connection
can detect the failure (egress or ingress).  Alarms can be suppressed or
alternate actions can be taken appropriately.

         --Tom




------------------------------------------------------------------------
Mathematics is the supreme nostalgia of our time.