The MPLS WG Archive

Cell Relay Retreat>MPLS WG Archive>month:1998-Sep> msg00194



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

problems with rsvp (fwd)

  • From: Roch Guerin <guerin@watson.ibm.com>
  • Date: Wed, 16 Sep 98 12:30:58 -0400
  • Cc: jh@lohi.eng.telia.fi, mpls@external.cisco.com, rsvp@ISI.EDU
  • X-SMAP-Received-From: outside


> More relevant, perhaps, is that removing the requirement for
> heterogeneity would indeed simplify the RSVP spec, for example by
> removing the need for blockade state.  I don't know enough about MPLS
> to know whether or not mergng of heterogeneous reservations is an
> issue.
> 
>   *> in rsvp the sender may never be sure if the reservation actually
>   *> succeeded.
> 
> That is not quite true either; if the reservation succeeded, a sender
> can be sure of it (by receiving the E-E Resv message).

Bob,

I think that Juha may have meant receiver here ;-)  In that case, a RESV_CONFIRM
will work, but only for unicast or homogeneous multicast reservations.  
If you remove the requirement for heterogeneous reservations, then
that should do the job.  
On the other hand, a sender based system is a little
more efficient when it comes to acking reservations as it takes only 1 
roundtrip vs 1.5.
Another important aspect from Juhas's earlier note is the merging of
sender reservations, which is not something that can be readily done
with RSVP as it stands today.

Roch