Re: [manet] DYMO and other routing protocols

["Charles E. Perkins" <charles.perkins at nokia.com>]

Hello Philippe,

I really want to understand your point, but I confess
that I am somewhat mystified and not yet getting it.

How does OLSR adapt to "space"?  Do you mean
Euclidean space with constant radio range?  I can just
see the point about topology (because of MPRs),
but I think we should insist that both DYMO and
OLSR be equally able to utilize some SMF (or, lately for
us, SMURF) ways of flooding over reduced relay sets.

As I understand this, the common dependence on a
reduced relay set will give both proactive and reactive
approaches equal adaptability to topology.  But this
still doesn't enlighten me about adaptability in space.

Regarding the adaptability in time...

DYMO has the following time-based dependencies
- route timeouts
- shifting route discovery anywhere up to and
    including application launch
- some second order effects about sequence
   numbers.
I can definitely see a role for a new extension which
enables nodes in a path to specify a non-default
value for the route timeout.  But it is not clear to me
that this is what you are referring to.  Could you be
a bit more explicit?

Anyway, this note is a long way of asking for more
details, and I hope you won't mind having a longer
discussion about it.

Finally, if you would care to point out cases where
DYMO performs poorly, but which are not able
to be explained by reliance on brute-force flooding,
it would be appreciated.  I know about the case of
fully-connected traffic patterns, and a few other
corner cases, but somehow I feel you may have
something else in mind.

Regards,
Charlie P.


ext Philippe Jacquet wrote:
> I would say that from a very broad perspective, on-demand protocols 
> discriminate routes with respect to time and proactive protocols 
> discriminate routes with respect to space. Personnally I have 
> preference to proactive approach because it is richer and gives better 
> internet legacy. Objectively the two approaches have their advantages 
> and drawbacks and the two should be considered.
>
> This said, I don't think that on-demand and proactive approaches are 
> represented at the same level in MANET. OLSRv2 has "space" 
> adaptability: i.e algorithms which automatically adapt to space and 
> topology variety, including for instance the MPR mechanism. In DYMO I 
> don't see an equivalent "time" adaptibility that goes beyond the 
> default values in timers. I think this may explain why the protocol 
> performs great in some scenario and very poorly in other.
>
> Some adaptability feature in DYMO would be welcome, not necessarily 
> proactive, but at least some memory based feature could help.
>
> Philippe
>
> David Murray a écrit :
> > Hi,
> >
> > I am wondering how DYMO fits in with the other routing protocols. I 
> > have read a number of papers that discuss how on-demand routing 
> > protocols like DSR/AODV work better in highly mobile environments 
> > where movement is fast, CPU and memory are low and batteries are 
> > limited. Equally, in situations where movement is very low, and there 
> > are no power limitations, protocols like OLSR/TBRPF/STAR perform better.
> >
> > So, I have a mental picture of OLSR/TBRPF being predominantly used in 
> > stationary 802.11 mesh devices and AODV/DYMO being used to connect 
> > users mobile devices ush as phones and PDAs. Is this correct or are 
> > things not quite as simple as this? (I know RFC 2501 discusses MANET 
> > applications and characteristics but the discussion is quite general)
> >
> > If this is correct, it seems to me that DYMO and AODV are used in 
> > very similar situations (the ad hoc interconnect between users 
> > devices). I am aware that DYMO is a simplified version of AODV both 
> > in code and network operation. It seems like the major difference is 
> > the path accumulation feature in DYMO which allows nodes to append 
> > their information to a RREP to give other nodes better knowledge of 
> > the topology. It also seems that the hello feature has been removed 
> > in DYMO. So, is DYMO likely to be a replacement for AODV or do they 
> > have different uses/applications?
> >
> > Thanks for your time
> >
> > Dave
> >
> >
> > _______________________________________________
> > manet mailing list
> > manet at ietf.org
> > https://www1.ietf.org/mailman/listinfo/manet
>
>
> _______________________________________________
> manet mailing list
> manet at ietf.org
> https://www1.ietf.org/mailman/listinfo/manet



_______________________________________________
manet mailing list
manet at ietf.org
https://www1.ietf.org/mailman/listinfo/manet