[Roll] NCE vs DAO state saturation [was: roll-wg/draft-ietf-roll-enrollment-priority] Propose changes to reflect major points of our discussion (PR #14)]

Pascal Thubert <pascal.thubert@gmail.com> Fri, 11 February 2022 15:19 UTC

Content-Type: multipart/alternative; boundary="Apple-Mail-155CF975-F1EB-40FA-ADCD-5028402AF666"
Content-Transfer-Encoding: 7bit
From: Pascal Thubert <pascal.thubert@gmail.com>
Mime-Version: 1.0 (1.0)
Date: Fri, 11 Feb 2022 16:19:22 +0100
Message-Id: <D5C694FF-5A06-4135-9344-0780955D0371@gmail.com>
References: <6C895AAD-89D5-4D59-8990-4491FE012B74@gmail.com>
To: ROLL WG <roll@ietf.org>
Archived-At: <https://mailarchive.ietf.org/arch/msg/roll/47eTeAbHK3XTk6Nll31HAk94Hds>
Subject: [Roll] NCE vs DAO state saturation [was: roll-wg/draft-ietf-roll-enrollment-priority] Propose changes to reflect major points of our discussion (PR #14)]
Precedence: list

Replying from my phone gets me surprising destinations. Retry to the list:

> De: Pascal Thubert <pascal.thubert@gmail.com>
> Date: 11 février 2022 à 14:20:54 UTC+1
> À: roll-wg/draft-ietf-roll-enrollment-priority <noreply@github.com>
> Cc: roll-wg/draft-ietf-roll-enrollment-priority <draft-ietf-roll-enrollment-priority@noreply.github.com>, Push <push@noreply.github.com>
> Objet: NCE vs DAO state saturation [was: roll-wg/draft-ietf-roll-enrollment-priority] Propose changes to reflect major points of our discussion (PR #14)]
> 
> Dear all 
> 
> It appears reasonably easy and much needed to adapt rfc 8505 for sleeping nodes and NCE saturation, both of which should start at 6lo though we must impact rfc 9010 as well.
> 
> OTOH the saturation of the DAO state in storing mode is a hard problem. Basically a distributed asynchronous protocol that would overall balance a tree and even worse a dodag with possibly non equal loads. 
> 
> Worst case is that any router may be the single child of the root. Which means that that node should have as much state as the root itself. If that’s so then any node should be able to store the whole network.
> 
>  We designed storing mode with all that in mind and left it at that.
> 
> If there’s such a hard value for the size of the dodag then this draft is already sufficient for discouraging joins beyond the hard value.
> 
> The DAO projection work enables a hybrid mode where the amount of state in each node is supervised by a controller to optimize the value of the state that is installed, within the capabilities of each node, though the capabilities are for now know by an external mechanism. 
> 
> We can probably do more and I’d love to see research on that, but let us first complete the steps that we have taken, what do you think ?
> 
> Pascal
> 
>>> Le 10 févr. 2022 à 22:47, Konrad Iwanicki <notifications@github.com> a écrit :
>>> 
>> 
>> @konrad-iwanicki pushed 1 commit.
>> 
>> 8710275 editos and typos from Konrad Iwanicki
>> —
>> View it on GitHub or unsubscribe.
>> You are receiving this because you are subscribed to thi

[Roll] NCE vs DAO state saturation [was: roll-wg/… Pascal Thubert