+1 - except I think a good policy would be for join priority/preference to override rank
One way to do this would be to run the objective function, have the OF produce a list of potential parents, in OF priority order, and then walk the list and find a parent with the best preference. Just spit-balling here...
Randy
On July 21, 2017 at 7:29 AM "Pascal Thubert (pthubert)" <[email protected]> wrote:Hello Mališa
In my view:
- A join preference expressed as binary is too restrictive:
o We need to express a relative weight in a normalized way, because it will express things like the load of the neighbor cache and balancing can be valuable (see Hou’s work).
o We need something akin to an objective function to compute the join preference based on various sources of load, DAG wide, node local etc…
- also,
o You may use a node that is a preferred join parent for joining. RPL Rank may be a tie breaker.
o Once it ihas joined, the node would select its parents as prescribed by RPL between the candidates of least Rank. We may use RFC 4191 or an extended version in DIO of it as tie breaker.
Take care,
Pascal
From: 6tisch [mailto:[email protected]] On Behalf Of Mališa Vucinic
Sent: vendredi 21 juillet 2017 13:02
To: Michael Richardson <[email protected]>
Cc: [email protected]; Xavier Vilajosana <[email protected]>
Subject: Re: [6tisch] draft-richardson-6tisch-roll-join-priority-00 --- RPL DIO message
According to RFC8180, Join Priority/Metric is already calculated as a function of the RPL rank. See Section 6.1. My suggestion was to use that as-is *if* the join is allowed by the binary option in this new IE.
On 21 Jul 2017, at 12:44, Michael Richardson <[email protected]> wrote:
I have also a question, why not a simple binary option to decide if a
node accepts Join or not? Having a range and having the nodes calculate
That's what the previous revision had.
Malisa suggested it should be a priority.
_______________________________________________
6tisch mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/6tisch
_______________________________________________ 6tisch mailing list [email protected] https://www.ietf.org/mailman/listinfo/6tisch
