Re: [Roll] Question on MPL: parameter update on runnning MPL forwarders

[Yusuke DOI <[e-mail]>]

(2013-11-02 5:51), Michael Richardson wrote:
> peter van der Stok <[e-mail]> wrote:
>> Hi Yusuke,
>>
>> For the moment the parameters are general for al interfaces, seeds and
>> messages.
>> In the case of changing per message, the parameter sets need to be
>> attached to each message entry. (for me that is a large change)
> 
> I don't think he wants to do it per message
> He wants to use some out-of-scope management interface to tweak the values
> as the network grows, changes, or as performance data is gathered.
> 
> The question could be restated as:
>      What is the affect on MPL stability is nodes differ on Imin, etc. by up
>   to
>      10% while data transmission is active.

Michael, thank you for question clarification. Yes, we expect changing
large-scale mesh network with low traffic (somewhat around 1,000 node or
even more). I think carefully-selected 'default parameter set' should
work good enough for 95% of deployment. However we need some 'knob' to
control MPL parameters for the last 5% (or even more). The parameter
update shall be moderate and 10% is a good threashold.

(c.f.
http://tools.ietf.org/html/draft-doi-roll-mpl-parameter-configuration-03
is intended to make this happen)

I personally don't expect parameter update should happen frequently
(say, once a day or so). I'd like to analyze deeper on this topic soon,
but no clear result for this IETF meeting yet. (also depends on the
implementations -- if the code does not expect parameters update,

I also have interestingly read draft-roux-roll-mpl-eval-00. I'm trying
to follow their method (by my own code. I need to build it up from
scratch). If there are good codebase to start with, I really appreciate it.

Thanks,

Yusuke