Re: [Roll] ETX metric

[JP Vasseur <jvasseur at cisco.com>]

Hi Henning,

On Oct 13, 2009, at 10:11 AM, Henning Rogge wrote:

> Am Dienstag 13 Oktober 2009 04:48:34 schrieb JP Vasseur:
> > > I am saying that when using wireless links, a node should be aware
> > > that his communication could interfere with other communication.
> > > Less powered is the transmission signal, less noise is seen by
> > > neighbours.
> > > A low powered node will also need less energy to communicate in a
> > > less noisy environnement.
> >
> > Bear in mind, that you do not know this environment so this may lead
> > you to an incorrect conclusion.
> > The real metric is the link reliability, which reflect the properties
> > of the environment.
> We have a similar discussion about metrics in the Manet-WG.  
> Different kind of
> metrics seem to be useful in different enviroments. Signal strength  
> seem to be
> a good indicator for a "bad" link, but in praxis (especially with  
> incorrect
> signal strengt estimation of consumer hardware) sometimes links with  
> low
> signal strength are better than others with higher signal strength.  
> On the
> other side ETX might be easy to implement on any kind of layer-2  
> hardware, but
> is only a rough estimation of the "link quality" which does not  
> consider long
> time link statistics (stability, variance, ...) or transmission speed.
>
> Metrics for wireless networks are still being researched without  
> "final
> perfect metric" in sight, so ROLL should keep metric IDs for futher  
> metric
> types.

Adding metric will not be an issue, just a new codepoint. What you  
said, what we need is
a "good" enough mono-dimensional reliability metric to start with.

> But you should put at least ONE easy to implement metric into the  
> basic
> WG document.
> The problem of OLSRv1 (as an example) was that it did NOT include
> any metric except hopcount in it's basic document, so most people  
> using OLSR
> for their projects (who do NOT research metrics for mesh networks)  
> just use
> hopcount metrics and think that OLSR does not work.
>
> So the question about a metric for ROLL should not only be "what's  
> the best
> metric for our networks" but "what metric is simple enough that  
> everyone can
> implement it on any kind of hardware".

I fully agree with your approach.

Yes, what we have done so far (in the new revision to be posted) is to  
define one mandatory
metric (hop count) and have all other ones optional. May be "hop  
count" is one the one that
we will keep as mandatory but the spirit is to have a set of metrics  
people can decide to use
based on the OF.

We still need to define the one reliability metric that we will be  
using.


> > > As said somewhere else, the final metric should be a computation of
> > > different metrics.
> >
> > This is the objective function.
> Yes, a good metric "post processing" (including long term  
> statistical analysis
> of a links different metrics) can be a good thing to enhance the  
> quality of
> the "link cost" estimation.

Thanks for the useful feed-back.

JP.

> Henning Rogge