[prev in list] [next in list] [prev in thread] [next in thread] 

List:       mpls
Subject:    Re: [mpls] I-D Action:draft-ietf-mpls-ldp-igp-sync-04.txt
From:       "Ilya Varlashkin" <Ilya.Varlashkin () de ! easynet ! net>
Date:       2008-12-18 9:36:27
Message-ID: 7000E71D8C525042A815432358B2F124020644F2 () paul ! adoffice ! local ! de ! easynet ! net
[Download RAW message or body]

Hi,

I've just went through the draft and the way it deals with broadcast
networks seems to be quite harsh. Section 3 says that on broadcast links
traffic may be diverted from all nodes due to single LDP session. Even
if it's controlled by policy, this is heavy hit and undesirable.
Couldn't following solution work in case of ISIS used as IGP (haven't
thought yet of OSPF)?

1) if node involved in new LDP session setup is non-DR, delay advance
from INIT to UP until LDP sync is achieved (so neighbor state in the
ISIS TLV #240 is set to INIT for new LDP peer)
2) DR seeing HELLO PDU with such state should review all other HELLO
PDU's recently seen on the interface, and:
	a) if there are more PDU's listing that neighbor (LDP peer from
step 1) in non-UP state than PDU's listing it as UP, then set cost from
pseudo-node to that neighbor to 0xFFFFFE (when normally it's zero)
	b) if there are more UP PDU's than non-UP reset cost from
pseudo-node to that neighbor back to zero

Alternatively condition in step 2a could be strengthened to " if there
is at least one PDU with non-UP state set cost to 0xFFFFFE".

Could that work?

Cheers,
iLya
_______________________________________________
mpls mailing list
mpls@ietf.org
https://www.ietf.org/mailman/listinfo/mpls
[prev in list] [next in list] [prev in thread] [next in thread]