RE: WG next steps

["marcelo bagnulo" <marcelo@it.uc3m.es>]

Hello Christian,


> There are really two classes of solutions. I would categorize them as
> "host multi-homing and network multi-homing".

I think there is at least another type of solution, which we can call hybrid
solutions, that involve both the routing system and the hosts. This type of
solutions require the cooperation of both elements.

The rationale for this type of solutions would be something like this:

At least, a multi-homing solution must provide a way to store multiple
alternative routes to a destination and it also must provide a way to use
this information when the route used stops working properly.

Network based solutions store the information about alternative routes in
the routing system. This presents scalability problems, as you mention.
However this type of solution is attractive, IMO, because the elements that
are forwarding packets have the alternative route information, so they could
rapidly react when an outage occurs. I mean, on-route packets can be
re-routed through the alternative route, since the info resides in the
routing system. This, however is currently not always possible, since
re-convergence time of the routing system is too long, but i guess this
could be possible.

OTOH, host based solutions have much better scalability properties, since
alternative route information is stored in hosts, so the routing system is
not overloaded. However, this solutions imply that host must recover from
path outages, imposing new mechanisms to detect path availability, and
furthermore, en-route packets do not survive an outage, since the elements
that forward packets do not know about alternative routes to the given
destination.

So, here is where an hybrid solution could make sense, since it could store
alternative route information in the hosts, providing good scalability
features, and it could use that information in the routing systems, so that
when a router has no destination for a given address, it can use an
alternative route.
The problem here is that the information about multiple alternative routes
is stored in hosts and used in routers. A possible solution for this can be
to include alternative route information in packets. In this case, multiple
alternative route information is stored is hosts, carried in packets and
used by routers when is needed.


> I suggest we charter two follow-on efforts, one to explore a network
> based solution and one to explore a host based solution.

I suggest also to  explore hybrid solutions.

Thanks, marcelo

>
> -- Christian Huitema
>