> From: <[email protected]>
> My problem is that in traditional routing systems, the IPv4 address
> serves both the identifier and locator, and the routing table growth
> and routing dynamics happens to the routing system. When identifier and
> locator separates, the routing system will have much less pressure, but
> what about the mapping system. It seems that the pressure has been
> transfered from the routing system to the mapping system. Can the
> mapping system handle the large number of mappings and dynamics?
This is a good observation. My answer has two parts.
First, the path selection (i.e. routing) system has a hard enough time as it
is - trying to handle the network growth, dynamics etc is already hard,
without supporting other things too. So we _have_ to split things up into two
subsystems, because the network has gotten so large.
However, you are correct that if we want to support things like a certain
rate of change in identity->location bindings, the mapping system has to be
designed to handle that. I think the question is not so much 'can we build a
system that supports a given rate of change in bindings', or 'can we build a
system in which after a binding change, the change is propogated to everyone
who has a copy of that binding'. The question really is 'can we meet a given
desired performance target with an acceptable level of cost'. There's no way
to answer that without actually designing the mapping system in detail,
though.
Noel
_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg
