Re: [Roll] RPL Metric I-D

[Alexandru Petrescu <alexandru.petrescu at gmail.com>]

JP Vasseur a écrit :
> On Oct 12, 2009, at 4:00 PM, Alexandru Petrescu wrote:
>
> > JP Vasseur a écrit :
> > > Hi Alex,
> > > > [snip]
> > > > > > We could struggle to compare a path with high energy cost yet low
> > > > > > hopcount cost to a path with low energy cost yet high hopcount 
> > > > > > cost. Is
> > > > > > P1 preferable over P2 or vice-versa?
> > > > > >
> > > > > > I think this is a question to which one can't answer without 
> > > > > > knowing the
> > > > > > deployment context.
> > > > > Well at least we need one example where it could be useful. I can't 
> > > > > find of any personally.
> > > >
> > > > Here is one:
> > > >                    --        --
> > > >                   |R1|      |R2|
> > > >                    --        --
> > > >                     |        |
> > > >                     |        |
> > > >                     |        |
> > > >                     |        |
> > > >                      \ ---- /
> > > >                       |Host|
> > > >                        ----
> > > >
> > > > R1 and R2 send two different RAs.  Each pretends to be a default 
> > > > router (non-zero Router Lifetime) and send different "link energy 
> > > > metrics" to Host.
> > > Link energy relates to which link ?
> >
> > Link energy metric in RA relates to the link on which that RA is sent. 
> > One RA is typically sent on a single link.
>
> Yes but how would you expect "host" to use that information to select 
> the router ?

By using some local intelligence knobs.  If the knobs are up saying "use 
the default router on the link involving the least energy consumption" 
then do so.

> It may very well be that the router offering the lower energy cost end 
> up following a high energy path.

Yes.

But in very many cases the end node can't choose paths based on any 
parameter, not only energy (well, except RSVP).

The notion of 'default' router itself is such.  When the host receives 
an RA from a router pretending to be a default router there is no 
guarantee for the host that that router offers a shortest path to a 
certain destination (IP hopcount metric).

My PDA with a 3G and a WiFi interface: I have no idea which offers the 
shortest path to gmail.com.  I just pick the interface which is sure to 
lead to the Internet.  If I need an interface to a link which consumes 
less then I pick the interface which has just received an RA telling it 
how little it is supposed to consume.

Same with MTU: when host reads an RA telling it the MTU is 1280 it has 
no guarantee that the entire path is 1280 - often the intermediary links 
to the final destination offer larger-than-1280 MTU, typically the 
backhaul links.

Thus I propose to have the same kind of behaviour for the energy metric 
for link in RA, as we have for default routers and with MTU.

Later we can see about link energy metric for routing protocols. (as 
Path MTU Discovery is for MTU of paths).

Alex



> > In the above figure there are two links.
> >
> > Is this ok?
> >
> > Alex