Re: [Roll] FW: New I-D on composite routing metrics

[C Chauvenet <[e-mail]>]

Hi,

See inline.

Le 5 août 2011 à 01:58, Panos Trakadas a écrit :

Dear Cedric,

We believe that the definition of "order relation" is correct: the order relation is used by the OF to compare link/node weights along the path (equivalently, metric values) and select parent node.


[C.C] I Agree with your definition.

As an example, consider the case where the ETX is aggregated along the traversed path. If node A receives DIO messages from nodes B and C (potential parents) with ETX values of 2.5 and 3.2, respectively, then the order relation will help selecting node B (advertising the minimum value of ETX) as the parent of node A.
In this concept, your comment is right; the order relation of LQL must be minimum (and not maximum). Please consider it as a typo.


[C.C] OK


Regarding link-color comment: Apart from being used as a 10-bit flag field (e.g. for indicating the encrypted links, as described in rpl-metrics draft), link color may be used to indicate the number of nodes along the path transmitting at a specific radio channel; in this way interference can be reduced by selecting transmission in the "least occupied" channel. In this case, the order relation must be "minimum" and the domain is integer.

[C.C] I think that the confusion is greatly due to the very wide range that the link color metric may cover. You example of the usage of this metric is interesting, but many others usage may be possible, and so an global order relation seems difficult to find for this metric.

In general, we will re-format Table 1, following the metric description in rpl-metrics draft. Also, we will replace words "minimum"/"maximum" in the "order relation" column by "less-than" and "greater-than" in accordance to the order relation definition.

[C.C] I agree that it would increase the readability.

Before submitting this draft, there was a discussion on whether we should include examples and formulas in this initial version of the document; it seems that our decision was wrong... We will be happy to include examples and formulas to improve readability of this draft.

[C.C] Nice. I think this would provide precious guidelines for readers, and speed up some RPL design.

Finally, we think that "hysteresis" can provably improve RPL performance, and thus it will be included in this draft. To be honest, there is a "hidden" phrase: "...and if the first component values are equal *or differ less than a predefined threshold*..." (section 4.1, first sentence) indicating our effort to include hysteresis in this draft. We will give more emphasis on that point.

[C.C] I noticed that reference, but it should be worth to explicitly speak about it. I also agree that the Hysteresis mechanism defined in the MRHOF draft greatly improve the stability of the topology and should be highly recommended. BTW, some guidelines about how to determine the good threshold value would be interesting.

Thanks for your valuable comments,
Panos.

Cédric.


--- Στις Τετ., 03/08/11, ο/η C Chauvenet <[e-mail]<mailto:[e-mail]>> έγραψε:

Από: C Chauvenet <[e-mail]<mailto:[e-mail]>>
Θέμα: Re: [Roll] FW: New I-D on composite routing metrics
Προς: "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>
Κοιν.: "TRAKADAS PANOS" <[e-mail]<mailto:[e-mail]>>, "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>
Ημερομηνία: Τετάρτη, 3 Αύγουστος 2011, 18:46

Hi Theodore,

I found your draft interesting and it gave me some useful guidelines for metric's design.

I've several comments on this :

I fail to understand well the "Order Relation" you use in Table 1.
The definition of "metric order relation" given on page 4 doesn't really helped me.
Could you help me understand the Order Relation column in the table?

In this table in particular (Figure 1):

    - Could you precise your view of the  Link-Color metric (Why is it in the integer domain, why the order relation in "minimum" ?).

    - As far as I understand the Order Relation, it seems that the LQL order relation should not be "maximum" but rather "minimum" (same relation as ETX, excluding the 0 value).


More generally, I would be happy to see in the document some formulas examples for metric computation, or maybe some guidelines.

Some example of metrics composition would also be great (You mention one in your draft with is ETX + RE (remainging Energy). Do you have some idea about other set of metric that would be interesting to bind ?

I finally think that it should be worth to add some content about the mechanisms that aim to deal with metric's dynamics (Multi Threshold, EWMA, Low pass Filters, ...). Some of them are mentioned at the end of the rpl metric draft.

Best,

Cédric.

Le 2 août 2011 à 02:55, Theodore Zahariadis a écrit :

> Dear all,
> My colleague Dr. Panos Trakadas and myself are working on an EC-funded
> research project, called VITRO (Virtualized Distributed platform of Smart
> objects - http://www.vitro-fp7.eu/), which has adopted RPL as the primary
> routing protocol and we are currently performing some simulations and tests.
>
> By monitoring roll, we have noticed that three (sorry if we have missed any)
> I-D are discussed in the ROLL WG (draft-ietf-roll-minrank-hysteresis-of,
> draft-ietf-roll-of0, draft-gnawali-roll-etxof), defining RPL Objective
> Functions related to a single routing metric (either hop-count or ETX).
>
> In VITRO, each sensor may be part of many virtual networks (even from
> different administrative domains) and must be able to concurrently support
> multiple user requests even with contradicting QoS characteristics. It is
> obvious that in such cases, a composite routing metric is of great
> importance.
>
> In draft-zahariadis-roll-metrics-composition-00
> (http://datatracker.ietf.org/doc/draft-zahariadis-roll-metrics-composition/)
> , we aim at specifying the guidelines for designing efficient composite
> routing metrics to be applied at RPL routing protocol.
>
> We would be happy to receive your comments
>
> Best Regards,
> Theodore Zahariadis
>
>
> ------------------------------------
> Ass. Prof. Theodore Zahariadis, PhD
> Technological Educational Institute of Chalkida
> Psachna, Chalkida, Greece
> Tel: [phone]
> Skype: theodore.zahariadis
>
>
> _______________________________________________
> Roll mailing list
> [e-mail]<http://gr.mc296.mail.yahoo.com/mc/compose?to=[e-mail]>
> https://www.ietf.org/mailman/listinfo/roll


_______________________________________________
Roll mailing list
[e-mail]<http://gr.mc296.mail.yahoo.com/mc/compose?to=[e-mail]>
https://www.ietf.org/mailman/listinfo/roll