RE: [Ospf-wireless-design] Re: consensus

["Joe Macker" <[e-mail]>]

Emmanuel:

You have some interesting thoughts on analysis specifics and perhaps some of
this should taken to the manet-dt list if we talk about flooding
effectiveness, especially fragility as you mention. 

I have been leaning towards adjacency reduction as a key factor for
OSPF-MANET work, but as I have mentioned we have ongoing work looking at
flooding optimizations only in the context of SMF.  Adjacency reduction and
flooding effectiveness can be different subjects.  

See draft-ietf-manet-smf-01. We have been examining several CDS algorithms
for this purpose. including modifications of ones being considered in
manet-ospf.  There is some common ground here between ongoing manet SMF work
and OSPF-manet.  However, in the routing control plane adjacency reduction
is a scalability winner and I think that has been an appropriate area of
concentration.  If there is a significant win in the flooding effectivenesss
by protocol A vs. B (when mobile), I think that should be pointed out. So
far we see similar performance between several algorithms considered when
looing at packets delivered vs. packet sent in a variety of mobile
scenarios.  

Justin Dean has done some analysis, both analytical and mobile simulation of
CF,MPR, MPR-CDS (shared set), MDR (without redundant sets).  He will be in
Vancouver and could talk off-line about some of what he has done.


-Joe



>-----Original Message-----
>From: [e-mail] 
>[mailto:[e-mail]] On Behalf 
>Of Emmanuel Baccelli
>Sent: Friday, November 04, 2005 5:29 AM
>To: [e-mail]
>Subject: [Ospf-wireless-design] Re: consensus
>
>Hi all,
>
>i want to point out that it could be beneficial to include 
>Philippe Jacquet in this debate as he and Richard already had 
>many discussions about related subjects on the MANET mailing 
>list (and these may not be totally taken into account yet in 
>this design).
>
>I also want to clarify my point of view, because it seems i 
>was not clear enough. I think that flooding schemes and 
>topology reduction schemes do not have to be tied together. In 
>particular it may not be the best approach to evaluate a 
>solution only based on the topology reduction performance it 
>achieves. For example, flooding can be done with an MPR-based 
>mechanism, while topology reduction can be done based on a CDS 
>approach.
>I think we can have a flooding optimization design that is 
>more independent from the topology reduction design than the 
>recent discussions seem to point out.
>
>One thing that must be considered though, is the stretch 
>factor that is implied by the topology reduction scheme going 
>with MDRs. We are currently at INRIA running some simulations 
>to evaluate this stretch factor. It seems it may be quite 
>substantial, and this should be taken into account in the 
>design, as route stretching introduces an overhead that is 
>proportional to the data traffic. On the other hand using 
>MPR/MPRselector adjacencies does yield more adjacencies, but 
>no route stretching whatsoever. So here (in the topology reduction
>design) there is a trade-off that was not yet fully discussed 
>in my opinion.
>
>As far as the flooding optimization design is concerned, it 
>was not discussed so much lately, as the focus was rather on 
>topology reduction design. It should be beneficial to continue 
>discussing this point, and we would like to get some feedback 
>on our report that we published this summer about the 
>stability of CDS-based flooding schemes. It seems that when 
>the ad hoc nodes are not static, MPR flooding performance is 
>much steadier than CDS-based flooding which is rather fragile, 
>especially when the CDS is reduced to the fewest nodes. ( 
>ftp://ftp.inria.fr/INRIA/publication/publi-pdf/RR/RR-5609.pdf )
>
>We also ran some quick simulations to evaluate the number of  
>MPR adjacencies (for the same domain/range parameters as 
>Richard), and we found the numbers shown below:
>
>Number    Number of
>of nodes    MPR links
>10             10.50
>20             32.72
>30             57.56
>40             84.14
>50           114.56
>60           143.88
>70           178.64
>80           211.04
>90           241.80
>100         273.90
>
>We were therefore very surprised to compare these with the 
>simulation results given by Richard, as there is a rather 
>drastic difference?
>
>Emmanuel
>
>
>
>_______________________________________________
>Ospf-wireless-design mailing list
>[e-mail]
>https://www1.ietf.org/mailman/listinfo/ospf-wireless-design
>



_______________________________________________
Ospf-wireless-design mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/ospf-wireless-design