The MPLS WG Archive

Cell Relay Retreat>MPLS WG Archive>month:2000-Aug> msg00246



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

LinkSummary in draft-lang-mpls-lmp-01.txt missing fields?

  • From: Jonathan Lang <jplang@calient.net>
  • Date: Tue, 22 Aug 2000 08:42:54 -0700

Spencer,
  See comments inline.

-Jonathan

> -----Original Message-----
> From: Dawkins, Spencer [mailto:Spencer.DAWKINS@fnc.fujitsu.com]
> Sent: Tuesday, August 22, 2000 6:33 AM
> To: 'IETF MPLS mailing list'
> Subject: RE: LinkSummary in draft-lang-mpls-lmp-01.txt missing fields?
> 
> 
> Thanks, this helps a lot. If I understand this correctly, 
> 
> o the active control channel is listed as a working channel in the
> LinkSummary, 
yes

> 
> o any backup control channels are listed as protection channels in the
> LinkSummary, pointing to the active control channel as the protected
> channel,
yes

> 
> o so switching to a backup control channel looks (roughly) 
> like switching to any other protection channel.
The actual switching to a backup control channel looks quite different from
switching to a protection channel because of the procedure that is used in
LMP for the switchover.  For the component links, the switchover to a
protection channel is handled by the signaling protocol (e.g., RSVP or LDP).
Having said this, the mechanism to map the protection channels to working
channels is the same for control channel and component links.
 
> 
> I think I understand how control channel switchover happens 
> (section 4.1.3 is pretty clear), but am a little less clear on how data 
> channel switchover happens. Most of Section 7 deals with fault detection
and 
> isolation, not with signaling that data channel switchover is
occurring/has occurred.
> Actually, this is a good place to start.
> 
> When a data channel fails and the upstream node (right?) 
> sends an updated LinkSummary (including the former protection channel
listed 
> as a working channel), does this mean (1) the switchover has already 
> occurred, or (2) the switchover should occur? I'm guessing (2) from the
end of 
> section 7.2, but I'm not sure I saw this explicitly stated.
The signaling of data channel switchover is not done in LMP, but is
relegated to the signaling protocol that is being used.  LMP is used to
exchange the working/protect channel mappings and to assist in the fault
isolation.  Therefore, using your above example, (1) would be the proper
operation.

> 
> I also noticed that there isn't a section 8...
yes, I had some numbering problems...