Re: [Ospf-wireless-design] Design team outbrief (draft slides) posted

[Emmanuel Baccelli <[e-mail]>]

I think there is something rather confusing in the latest discussions: 
the use of the term adjacency, especially when compared to its 
traditional meaning in OSPF. For example, some simulation results tend 
to show that the MDR approach features a reduced number of adjacencies 
without featuring any route stretching -- if "min-hop" LSAs a used.  
However, this supposes that routing is not only done over adjacencies 
(indeed, if routing was only done over adjacencies, the MDR approach 
leads to more than 50% route stretching in most cases). This is thus 
very confusing, since in traditional OSPF routing is supposed to happen 
only over adjacencies. So there is here at least a vocabulary issue, 
that needs to be clarified in my opinion.

Emmanuel


Richard Ogier wrote:

>> Emmanuel Baccelli stated that MPR flooding offers better properties 
>> than CDS flooding, with at least as good topology reduction 
>> capabilities, and better routing stretch performance
>
>
> After further consideration, I partly agree with the above statement
> depending on how it is interpreted.
> The part about "better route stretch performance" is true if we are
> talking only about *flooding* and not *routing* (since routing can
> occur along non-adjacencies).  Min-hop routing can be achieved either
> with MPRs or with the min-cost LSAs described in the MDR draft.
>
> The part about "good topology reduction capabilities" depends on how
> adjacencies are defined. (By topology reduction I am talking about
> adjacencies, keeping in mind that LSAs can also include non-adjacent
> neighbors.) In our simulations, Phil and I showed that this
> is certainly not true if each node forms an adjacency with each of
> its MPRs and MPR selectors, and Emmanuel has agreed with this:
>
>> On the other hand using
>> MPR/MPRselector adjacencies does yield more adjacencies,
>> but no route stretching whatsoever.
>
>
>
> If adjacencies are defined in another way, then the method needs
> to be specified. One possible way (which I plan to investigate by
> modifying OSPF-MDR) is to select the CDS/MDRs by pruning MPRs
> and then defining adjacencies as in the MDR draft, but Emmanuel
> did not specify such a method.
>
> Richard
>
>
>
>
> _______________________________________________
> 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