The MPLS WG Archive

Cell Relay Retreat>MPLS WG Archive>month:2001-Aug> msg00316



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

RSVP-TE packet loss

  • From: David Charlap <david.charlap@marconi.com>
  • Date: Fri, 24 Aug 2001 15:45:55 -0400

> .....RSVP-TE runs over UDP so it is less reliable than CR-LDP who
> runs over TCP......

This is completely untrue.

1: RSVP runs as raw IP (protocol 46), not over UDP.  UDP
   encapsulation is only supposed to be used by nodes that are
   incapable of generating or receiving the raw IP datagrams.

2: Not using TCP does not make RSVP unreliable.  By this logic you
   might as well claim that IS-IS is an unreliable routing protocol
   because it doesn't use TCP either.  It doesn't even use IP, for
   that matter.

   Not using TCP simply means that RSVP must implement its own
   mechanisms for reliability.  The use of soft-state is one such
   mechanism.  The message ID/ack components of the refresh reduction
   extensions are another.

   Use of TCP simply means that LDP is offloading its reliability
   features onto the OS's protocol stack.  This has some benefits
   and it has some drawbacks.

> I read about some enhancements to RSVP-TE for allowing it to
> recover from packet loss in time frame comparable with those of
> TCP.

Read RFC 2961: RSVP Refresh Overhead Reduction Extension.

Among its components, this extension has a mechanism for reliable RSVP
message delivery that greatly improves (over just using soft state) the
recovery time due to packet loss.

-- David