The MPLS WG Archive

Cell Relay Retreat>MPLS WG Archive>month:2003-Nov> msg00181



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

mp2p with ECMP (was on documenting ECMP (was on the mpls oam framework))

  • From: neil.2.harrison@bt.com
  • Date: Thu, 20 Nov 2003 07:10:04 -0000
  • Thread-Index: AcOuudcnzHY5H4TXSkyrJEpaF05iVgAGfW6w
  • Thread-Topic: on documenting ECMP (was on the mpls oam framework)
  • X-MIME-Autoconverted: from quoted-printable to 8bit by cell.onecall.net id hAK7S2n14526
  • X-OriginalArrivalTime: 20 Nov 2003 07:10:05.0239 (UTC) FILETIME=[52B77C70:01C3AF35]

Can someone please explain to me the purpose of having a standardised MPLS layer mp2p construct with proprietary ECMP?  The former is a co-ps mode server layer consruct which is causing *traffic convergence* of its clients, whilst the latter is doing its best to ensure *traffic divergence* of its clients.  In terms of control/determinism this would seem to undermine the routing role of the mp2p server layer construct......but I'll bet some people also do it with p2p TE LSPs (which, if true, would seem to completely undermine their purpose).   That is, if I removed the mp2p server layer then the actual traffic distribution of the clients would not be affected......so exactly what was the point of doing it in the 1st place (other than creating a new and rich set of network problems to address, eg OAM and fault management, traffic, performance).  

Further thought.....just how does all the above fit in with TE DS classes and the notion of constraining routing?  

BTW - just in case anyone is wondering.  I still firmly stand by the observations that a mp2p consruct and PHP are violations of the co-ps mode.  Nothing can change that view as its simply true.

regards, Neil