The MPLS WG Archive

Cell Relay Retreat>MPLS WG Archive>month:2000-Aug> msg00276



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

questions on mpls-ldp-09

  • From: Eric Gray <EGray@zaffire.com>
  • Date: Thu, 24 Aug 2000 08:49:04 -0700

Yuan,

	Please see embedded comments below.

> -----Original Message-----
> From: Yuan Gu [mailto:guy@sh.bel.alcatel.be]
> Sent: Wednesday, August 23, 2000 8:43 PM
> To: mpls@UU.NET
> Subject: questions on mpls-ldp-09
> 
> 
> Dear All:
> 
> I have some questions on  mpls-ldp-09.txt:
> 
> 1.How many LSPs can be setup in same session?  

As many as the two LSRs can maintain state for.
There is no artificially imposed limit.

> If session fails for some reason, do all LSPs in 
> this session have to be released?

Currently, yes.  See the draft on Fault Tolerant 
LDP - which was accepted as a working group draft
at the last IETF meeting - for how LSP state may
be preserved across LDP sessions.
 
> 2. In section 3.5.2.1 Hello Message Procedures
> Configuration Sequence Number:
> What kind of session parameter can be changed by passive LSR without
> noticed by active LSR? If receiving LSR detects a change in the
> configuration in the sending LSR by detecting the change of
> Configuration Sequence Number, beside clearing the session 
> setup backoff
> delay, what else the receiving LSR does? If receiving LSR playing the
> passive role and detect the change, what does it do?

The intent of the configuration sequence number is
to allow the active peer to be able to find out that
the passive peer has been re-configured.  

Why do we want to allow this?  

Because we can assume that a likely reason why two 
peers are unable to establish a session is that they 
have been configured with session parameters that are 
incompatible.  In this event, each repeated attempt
will fail and - with exponential back-off - each
new attempt will be delayed by an interval that is
twice what it was in the immediate prior attempt.
Obviously, this delay can grow to be quite large.
The inclusion of a changed configuration sequence 
number in a Hello message from the passive peer 
signals the active peer that it might ignore the
current retry delay and try again immediately.

Since the passive peer does not control when an
attempt will be made (it does not initiate LDP
session establishment), it does not matter if it
detects a change in the configuration sequence
number.  If the active peer has been re-configured,
it can initiate LDP session establishment on its
own (without sending a configuration sequence
number to its passive peer or peers).

> 
> Thanks alot.
> Yuan Gu
> 
>