The IP Over NBMA (ION) Archive

[Masataka Ohta <mohta@necom830.hpcl.titech.ac.jp>]

> > > Masataka, OK, so you don't see the need to subdivide any given campus
> > > into multiple IP subnets.
> > 
> > Wrong. There are a lot of reasons to subdivide campuses into multiple
> > IP subnets, none of which needs cut-through.
> 
> What do you assume as the ratio of ATM switches to IP routers, roughly?

My opinion?

*IF* configuration of ATM bridges with full SVC capability WERE
easier than that of IP routers, which is completely possible
by having no PNNI hierarchy, 10 or 100 ATM bridges per an IP router
could be a reasonable figure in campus environment. If each switch
has 10 hosts attached, it means 100 or 1,000 hosts in a non-broadcast
subnet. Not a bad figure.

If it is not and is likely to continue to not to be, don't have
ATM bridges.

*I* think such configuration makes management effort minimal.

But, if you think you can manage NBMA with 1,000,000 hosts, I can't
stop you try. But, note that management of a single LIS with ATMARP
is much easier than that of a multiple LISes with NHRP. So, if
scalability were not an issue, ATMARP will be able to support,
perhaps, an NBMA with 100 or 1000 times more hosts than NHRP.

You may still have some administrative reasons to have more routers
for firewalling or things like that which means no cut-through.

							Masataka Ohta