RE: 6MAN Adoption call on raft-chakrabarti-nordmark-6man-efficient-nd-04

I support the adoption of this draft.

Thanks
Suresh

----- Original Message -----
From: "Eric Levy- Abegnoli (elevyabe)" <elevyabe@cisco.com>
To: Ole Troan <otroan@employees.org>, 6man WG <ipv6@ietf.org>
Cc: 6man Chairs <6man-chairs@tools.ietf.org>
Sent: 11/21/2013 3:20 AM
Subject: Re: 6MAN Adoption call on raft-chakrabarti-nordmark-6man-efficient-nd-04

Hi,
I support this draft: besides the operational value already mentioned, I
can see all the cool things we would be able to do, to scale up the number
of node on the link.

Broadcasting on the link is quite fine when you have a limited number of
addresses to deal with, and the dynamic is low. But we plan for layer-2
networks much bigger and much more dynamic (industrial, datacenter, very
large campus) and broadcasting + caching is creating a bottleneck.

I imagine we could also do routing with a cache instead of a table, and
broadcast queries to find the destination prefix. Everybody would agree it
does not scale. It's not just the
numbers; it's also the dynamic.

I ran a simple math: 100,000 nodes, 3 addresses each, 10% of nodes moving
every minute (moves can be physical or virtual), 3 multicast messages sent
per move (DAD, NA, MLD), message size could go from 100B to 1KB (when
using CGA for instance), that would range from 1 Mbit to 8 Mbit/sec.
Definitely an issue for the wireless links, but not only.
Last but not least, there is quite a bit of latency involved when looking
for a target that is not in cache.

A cache is a good response to these problems, as long as the dynamic is
low. But it's just a cache: when numbers grow, it can't keep up.  And it¹s
not designed to be distributed.
Registration and dissemination (this draft, and sail-draft) propose a
shift to a more pro-active approach, "a la" routing. Any node looking for
an address on-link will either have it in its table (not cache) or will
know who have it (that's the distribution part).
The critical piece is not just registration; it's also distribution so
that the information is available quickly where it's needed, along the
forwarding path.

There is a lot to clarify though: co-existence with the legacy model,
distribution scheme (hierarchical, hash-based, Š), nature of the
registrars (Routers? Switches? DHCP servers? Š).

Eric

On 19/11/13 11:45, "Ole Troan" <otroan@employees.org> wrote:

>All,
>
>There was moderate support to adopt this draft at the working group
>meeting in Vancouver.
>This is an adoption call to confirm the result of the hum at the meeting.
>
>Please provide a view with reasons as to whether the WG should adopt this
>or not.
>
>This message starts a one week 6MAN Working Group call on adopting:
>
>    Title           : Wired and Wireless IPv6 Neighbor Discovery
>Optimizations
>    Author(s)    : S. Chakrabarti, E. Nordmark, P. Thubert, M. Wasserman
>    Filename    : draft-chakrabarti-nordmark-6man-efficient-nd-04
>    Pages        : 31
>    Date          : 2013-10-22
>
>
>http://tools.ietf.org/html/draft-chakrabarti-nordmark-6man-efficient-nd-04
>
>The call ends on November 26th, 2013.
>
>Regards,
>
>Bob Hinden & Ole Trøan
>--------------------------------------------------------------------
>IETF IPv6 working group mailing list
>ipv6@ietf.org
>Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
>--------------------------------------------------------------------