The MPLS WG Archive

["Punj, Arun" <Arun.Punj@marconi.com>]

Per Eirik,

I think what you are asking is how to ensure/monitor the "health" of an LSP.
A healthy LSP L1 (from E1 to E2) would be defined as one which has any or
all of the following features ( depending afcourse on how healhy you want
your LSPs to be).

1. LSP is operationally up. ( A frame input at E1 will arrive at E2)
2. No leakage of traffic from LSP. ( A frame sent on LSP L1 will not be
misinserted into some other LSP L2 )
3. No leakage of traffic into LSP. ( A frame sent on LSP L2 will not be
received on the LSP L1 )
4. Traffic flowing on L1 gets the service contracted for.
[This list is not exhaustive, but representative]

Generally the procedures used to ensure the health of LSPs ( or VCs ) are
refered to as OAM procedures. Some feature desired of these  procedures (in
addition to above requirements) are  :
- They require no user complaints
- They be MPLS application independent (  be applicable for non IP LSPs too
)
- They should be independent of control planes ( although control planes
themselves may use these procedures for OAM ).

Traditionally IP networks have used tools like PING and TRACEROUTE to ensure
that destinations are reachable and trace the path to the destination etc.
Variants of these procedures have been defined recently for MPLS. However,
they all have various shortcomings, having been originally designed only for
IP world. 

There were some internet drafts which tackled the issue of OAM holistically
by folks like Neil harrison, Shahram Davari etc but I am not sure if they
are still current or expired.


thanks
Arun Punj
Office of CTO, Marconi ,
Ph  : 724-742-7583 - Web: www.marconi.com
1000 Marconi Drive, Warrendale, PA - 15086



> -----Original Message-----
> From: Per Eirik Heimdal [mailto:perheim@frisurf.no]
> Sent: Wednesday, October 31, 2001 2:28 AM
> To: mpls@UU.NET
> Subject: Performance Monitoring on MPLS
> 
> 
> Hi all!
> Can anybody help with suggestions to what should be monitored 
> on MPLS level to give an impression of the performance? I am 
> thinking of both traffic parameters and operational isssues 
> (indications if an LSP has gone down etc.)
> 
> I would also appreciate references to drafts, RFCs etc. 
> eventually describing this.
> 
> Thanks!
> BR
> Per Eirik
>