Re: [manet] Comments on "Problem Statement for OSPF Extensions ..." draft

["Madhavi W. Chandra" <mchandra at cisco.com>]

Hello Justin and Charlie,

On Thu, Apr 08, 2004 at 12:45:06PM +1000, Justin Lipman wrote:
> Hi Charles,
> 
> > Your observation is very important to understanding
> > how to improve MPR flooding.  We have, indeed,
> > measured a drop in packet delivery ratio when the
> > number of broadcasts is kept to the theoretical
> > minimum.  This occurs because broadcasts are not
> > reliable, and when some are lost during MPR flooding
> > their absence is more damaging than with blind flooding
> > (I assume not totally blind!).
> 
> Very true...
> 
> I have some results in the following paper which shows the performance of 
> Blind flooding, MPR, LMST Flooding and a reliable variant of RMST that 
> utilises unicast transmission instead of less reliable broadcast 
> transmission.
> 
> http://www.cse.unsw.edu.au/~justinl/rmst.pdf
> 
> This approach takes optimisation to the extreme using MST and is probably not 
> acceptable for route discovery in reactive routing protocols. Also it places 
> much emphasis on transmission power control. However, for a proactive routing 
> protocol like OLSR where link state dissemination is the requirement, it may 
> be most applicable.
> 
> We have a paper in preparation which looks at ways of improving MPR 
> reliability for use in both reactive and proactive routing protocols.
> 
> Ill be happy to send that to you once its ready.
> 
> > >What is your definition of "reliable" flooding?
> >
> > I thought it meant that a broadcast source gets an acknowledgement
> > from each of its neighbors that its broadcast was received.  This
> > acknowledgement can come in various forms, of course.  If I am
> > using the term in some way that is not compatible with the discussion
> > about OSPF extensions, I hope someone will tell me what the correct
> > meaning should be.

I've been operating on this understanding of reliable flooding as
well...that the recipient signals back in some form that it has indeed
received the transmission.  To me, the reliable flooding definition
below sounds more like best effort.  Justin, thanks for the pointers
below...I'll have to look into them.

Regards,
Madhavi

> Im not sure either if its compatible with the OSPF extensions as I havent kept 
> up with that. Im interested to know though.
> 
> However, my definition of "reliable flooding" is that the message being 
> flooded is delivered to as many nodes as possible. This is irrespective of 
> local acknowledgement of received packets. For example Wei Lou's work:
> http://www.cse.fau.edu/~wlou/publications.htm
> 
> W. Lou and J. Wu, Double-Covered Broadcast (DCB): A Simple Reliable Broadcast 
> Algorithm in MANETs, accepted to appear in Proc. of IEEE Conf. on Computer 
> Communications (INFOCOM 2004), Hong Kong, China, April, 2004. 
> 
> W. Lou and J. Wu, A Reliable Broadcast Algorithm with Selected 
> Acknowledgements in Mobile Ad Hoc Networks, Proc. of IEEE 2003 Global 
> Communications Conference (GLOBECOM 2003), San Francisco, California, USA, 
> December, 2003. 
> 
> > >Blind flooding can be seen as reliable and my own work shows it performs
> > > very well (if one ignores the broadcast storm).
> >
> > Why is it legitimate to ignore the broadcast storm?  That's the
> > entire motivation anyway for trying to create a better broadcast
> > mechanism.  Congestion due to broadcast can significantly
> > degrade packet delivery ratios in many scenarios.
> 
> I "dont" think it is legitimate to ignore the broadcast storm.
> I was just saying if one were to ignore the broadcast storm problem then Blind 
> flooding does provide very reliable packet delivery as shown in my paper.
> 
> My approach, RMST, addresses the broadcast storm problem and has comparable 
> reliability to blind flooding (in terms of packet delivery).
> 
> > > Sadly, MPR and other optimised
> > >flooding algorithms do not perform well at all!
> >
> > I thought it depended on the scenario!
> 
> Hmmm... not sure what you mean?
> 
> In my paper I introduced increasing CBR traffic into the network from 3 source 
> destination pairs. The optimised approaches performed very badly. Blind 
> flooding performed very well as did RMST.
> 
> I can think of two scenarios:
> 
> - If there is heavy network load (traffic) the packet loss in the network 
> increases, resulting in reduced packet delivery by the optimised flooding 
> mechanism.
> 
> - If nodes are mobile and the neighbour information being used is outdated 
> this may lead to bad rebroadcast decisions by the optimised flooding 
> mechanism, possibly resulting in reduced packet delivery. Examples of this 
> are nodes being selected as rebroadcasting node, or in self-selecting 
> approaches nodes not selecting themselves to rebroadcast...
> 
> 
> Gezzz ad hoc networks are a fascinating problem! :-)
> 
> Kind regards,
> Justin
> 
> -- 
> Research Associate
> Smart Internet Technology Cooperative Research Centre
> http://www.titr.uow.edu.au/~justin
> 
> _______________________________________________
> manet mailing list
> manet@ietf.org
> https://www1.ietf.org/mailman/listinfo/manet

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