On 4/29/12 5:54 AM, Somasundaram Selvaraj wrote:
All
I went through RFC 4861 and found the section on "Neighbor
Unreachablity Detection" to be of less signifance especially on the
links between Router.
Did you read the text in section 7.3 which says
Neighbor Unreachability Detection may
also be used between routers, but is not required if an equivalent
mechanism is available, for example, as part of the routing
protocols.
Keep in mind that ND predates BFD; if the specification which written
today it might as "as part of the routing protocols or BFD".
I don't know if anybody has looked for a detailed description of how ND
address resolution between routers might interact with BFD, but AFAICT
this should be straight-forward. Just disabling NUD on the interfaces
where BFD is being used.
Does that answer your concerns?
Erik
Reason is that, while there are failure detection protocols already
available to detect a liveliness of the neighbor I dont understand
effectivenss of such a mechanism within IpV6.
Moreover, whenever the neighborstate is not in "Reachable state", there
is a bunch of packets that gets trapped to the CPU to invoke IPV6 to
send NeighborSolicitaion message to detect the liveliness of the
neighbor. And Interestingly, the implementations continue to forward to
the neighbor irrespective of what the "Nbr state is" and then trap the
(copy of the packets)packets to cpu to do a Neighbor Unreachablity
Detection check (for the sake of sticking to the standard).
I didnt like Router Doing this especially on the Router-Router link.
Here are my opinion on handling the above issues.
1: Why to go through all the NeighborStates
(Incomplete->Reachable->Stale->Delay->Probe) when Protocols like BFD are
already present to check the liveliness of the neighbor (On a
Router-Router link). Cant we avoid "Neighbor unreachablity detection"
completely?
2: Simplify the NeighborState Machine.Just mainitain two states in the
Neighbor cache when the neighbor is a router. The states would be
"Incomplete and Reachable".
Enhancement as follows:
Initially the Neighbor state would be marked "Incomplete" and a
NeighborSolicitation message would be originated to determine the link
layer address.Once it receives a response move the neighbor state to
"Reachable". Just maintain an ageout timer like the one meant for ARP
(no more complications) to flush an entry from neighbor Cache.
Please share your thoughts.
Regards
Somasundaram
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
