Michael Richardson <mcr at sandelman.ca> Sent by: roll-bounces at ietf.org
11/20/2009 01:39 PM
To
roll at ietf.org
cc
Subject
Re: [Roll] updating DAO caches (was
Re: Something to ADD)
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
>>>>> "Jerald" == Jerald P Martocci <Jerald.P.Martocci at jci.com>
writes:
Jerald> As probably the strongest proponent of the P2P
requirement
Jerald> on this mailing list, I will readily admit that
even in
Jerald> building control only about 25% of the router
nodes need to
Jerald> pass intra-LLN P2P messages. As I understand
the DAO
1/4 of the nodes need P2P messages, but how much traffic does that
represent? Is it a big percentage of overall traffic? Does
it have
particular characteristics that make a trip through the root bad
(i.e. latency)
The traffic generated
by these 25% of these intra-LLN is also about 25% of the overall traffic.
As an example, a room controller may need to access the 'outdoor
air temperature' and 'relative humidity' sensors so it can properly calculate
its control algorithm. These two points are common to many rooms
and will only be instrumented once. All other room control will need
to reference these points. As for latency, this is important since
the room control algorithm cannot be calculated until all input variables
have been accessed. We need to complete the read request in under
500 ms.
Jerald> So yes, I think we either need to design outward
traffic
Jerald> into the DAG at the same level as we do inward
traffic; or
Jerald> devise an on-demand scheme where a node can easily
Jerald> 'discover' another node in the DAG as needed.
Richard's
Jerald> comment below is profound saying that right now
establishing
Jerald> a good outward link is a matter of luck, not design.
When you say "discover", I think you mean:
discover a path through cousin nodes that avoids the root
Yes. Maybe 'discover'
may mean something different in IETF parlance. Maybe 'find' a path
would be better. The point I am trying to make is that right now
using DAOs, we need to create and maintain a link for every node in the
DAG on the chance that there may need to one day be a message that must
traverse the link. So, on a 200 node network where every node might
need to talk to any other node, there would need to be 39,800 (i.e. 200*199)
links created and maintained. In reality only 50 links (i.e. 200*25%)
are required. I'm suggesting that if we had a means to find a P2P
path as needed, we could save significant processing power and DAG overhead.
I too, would like to have this.
good!
I think that this could be a different protocol --- that it could
leverage the communication used by the DAG to implement some kind of
flood-fill algorithm to learn specific P2P routes that are important.
I think the constraints of least-capable nodes excludes this from the
base protocol -- but it's really want I think I care about.
- --
] He who is tired of Weird Al is tired of life!
| firewalls [
] Michael Richardson, Sandelman Software Works, Ottawa, ON
|net architect[
] mcr at sandelman.ottawa.on.ca http://www.sandelman.ottawa.on.ca/ |device
driver[
Kyoto Plus: watch the video <http://www.youtube.com/watch?v=kzx1ycLXQSE>
then sign the petition.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Finger me for keys
iQEVAwUBSwbwP4CLcPvd0N1lAQIOjAf/VrBNnnxuVjhoV7CVZL2vuT1lJhJiMOH3
iDWPlL5OCjwjpkjnC2HLYIxxIYh+kMKrqUvea2BnEoMYIAgfG7MsjuU6eFIY5hMe
gywIbHhf5Zs69HkWxO6uUj4h4F2eiJAJjrTp8b207DWEY+oR/Ky0m1ffH1TCGuOz
jlwqsSMI94M/RLPSwCw/iGD/zLNDyjvI9Sird16YQ7TP2FM2JTRTqJPOpt74dNCR
KPeWOUIfK5kKd3S9tYOYC1BRyWUNUFlDO1iApA5YCIcFhQbi3rCqtfJWQqqKcumC
Ojw1sPchsmdfLb7ScYgYKOgdimkzG4Ga+jOdtCv0HqvrviwNJy+A3Q==
=Ujy3
-----END PGP SIGNATURE-----
_______________________________________________
Roll mailing list
Roll at ietf.org
https://www.ietf.org/mailman/listinfo/roll
