The IP over ATM Mailing List Archive by date

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

> We first said VCs are cheap, so we don't need the additional demultiplexing.

I still think they are cheap.

Though Q.2931 signalling of long haul VC's may not be so cheap,
it is a problem of large cloud model, not IP over ATM in general.

> Then we found that VCs are not cheap and use resources to setup and maintain
> at all points
> in the network.

Of course, several bytes of couters are necessary for leaky bucket
algorithm, which is almost as cheap as a VC conversion table entry.

Does that matter?

RSVP needs a lot more state (16 bytes for an IPv6 address only)
setup in each router.

So, ATM switches may need some amount of storage for RSVPed communication.

Of course, we should NOT try to use reserved communication automatically
for all the TCP connections. But, we can't because we can't estimate
the necessary QoS, which may vary as time goes by, of the connection.
Reservationless reservation is just non-sense.

> So now we are looking for techniques to reduce VC useage.

We are not.

> If we had chosen to use AAL3/4, at least we would have a mechanism to avoid
> the need to 
> queue packets at arbitrary points in the network where mp-to-p flows join.

Maybe, it's time to reduce AAL5 IP MTU to 512 byte. :-)

						Masataka Ohta