The MPLS WG Archive

Cell Relay Retreat>MPLS WG Archive>month:2003-Apr> msg00051



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

[PWE3] MPLS PID

  • From: Curtis Villamizar <curtis@fictitious.org>
  • Date: Wed, 02 Apr 2003 16:53:47 -0500
  • cc: "'curtis@fictitious.org'" <curtis@fictitious.org>, jleu@mindspring.com, Dan Tappan <tappan@cisco.com>, David Allan <dallan@nortelnetworks.com>, "'Lloyd Wood'" <L.Wood@eim.surrey.ac.uk>, pwe3@ietf.org, mpls@UU.NET


In message <4B6D09F3B826D411A67300D0B706EFDE0115C88D@nt-exch-yow.pmc-sierra.bc.
ca>, Shahram Davari writes:
> >Unsignaled E-LSPs seems to be the most widespread.  ECMP is being used
> >on them.  
> 
> If you meant EXP bits are taken to account during ECMP, then that could
> reorder the microflows, because in E-LSP the EXP represents both PSC and
> drop precedence. 

No I don't mean that.  ECMP is being used on E-LSP.  ECMP is not
considering EXP.

> If you didn't mean that, then putting PID in the EXP would not cause any reor
> dering.
> Off course there would be some issues if the bottom label is of uniform model
> .

The EXP bits are already being used on unsignaled E-LSPs.  E-LSPs are
widespread.  They carry both IP and non-IP.  ECMP is applied to the
E-LSPs but do not consider the EXP bits since doing that would
obviously cause massive microflow reordering problems and deployments
don't do things incredibly stupid (for very long).

> -Shahram

Curtis

  • References: