Subject: Re: The future of ATM
To: Eran Peled <epeled@ibm.net>
From: Jason Thorpe <thorpej@nas.nasa.gov>
List: tech-net
Date: 10/11/1996 08:47:33
[ CC'd this to tech-net@NetBSD.ORG, since there may be some interest
  on this topic there, as well. ]

[ disclaimer: I'm not an expert on Hippi ... rather, these are conclusions
  I've reached from my experience with ATM and from reading papers and
  attending talks by, and having conversations with Hippi experts. ]

On Fri, 11 Oct 1996 11:11:28 GMT 
 Eran Peled <epeled@ibm.net> wrote:

 > At 10:39 AM 10/10/96 -0400, Ron wrote:
 > >atm is dead. At least here in the us. Here's some interesting pricing 

I must agree with Ron on this one... My job is to work with ATM, and
I'm becoming more and more convinced that the kinds of things people want
to use ATM for just don't make sense for ATM.

In terms of really high performance LAN connectivity, Hippi-6400 is going
to be a clear winner, and will, if the companies can keep their promises,
be available 1Q97.  Hippi-6400 provides 6400 Mb/sec, and has solved 
the switch/bus contention issues associated with Hippi-800.

The Hippi-6400 switches will also bridge to Ethernet.

Hippi-6400 will also win in price/performance; to achieve the speeds
of Hippi-6400, ATM interfaces will need quite a lot of fast memory
on the interface cards to do SAR.  The way Hippi-6400 works (multiple
stage ACKs done by hardware along the entire path), keeps the per-VC
buffer small (there are exactly 4 VCs used by Hippi-6400, as opposed
to N where N may be very large in ATM) ... the memory required to keep
state for up to 1km of distance fits within the ASIC.

Now, I'm not endorsing Hippi-6400 outright, but it certainly looks
interesting, and better than ATM for high-performance local area
networks, at least on the surface.  Need to evaluate Fibrechannel
next :-)

The places where I see ATM being useful for real are replacing high-cost
trunk circuits with bandwidth purchased from a cloud service.  Shuck
those costly T3s, and buy 45Mb/sec from {carrier of your choice} (I'm
not about to endorse one over the other...)

Fast Ethernet can already make for a decent backbone, and with gig
Ethernet is out, ATM, watch out.

 > >Ok, so looking at performance/price there's already a very serious 
 > >problem. Now let's look at management. To install a 100 mbit ethernet you 
 > >buy the parts and plug them together and it works. To install atm, well, 
 > >you know what a chore that is :-)

Esp. with all of the (IMO) really lame protocols that have been invented
to make ATM behave like Ethernet.

 > >I'm not so happy about this either, i'm just finishing up a gigabit ATM 
 > >card. Such is life.

Know how that goes... I've spent the last year working with ATM, only to
reach the conslusion that it's probably not worth spending much time with.

 > This is very convincing, yet, isn't it  only true for the LAN market case?
 > What about WANs?

See my comment above...

 > And, if Enet will be the winner, will it support Voice as well as Data? 
 > If so, would it need synchronization,. due to delays and jitters?
 > What about real time Videoconferencing? What do you think will be the fast
 > Internet solution? 

Not "natively", no.

The one thing that ATM can really claim is that it's an "end-to-end
soluition", in that you can have a full ATM path connecting two
workstations located 3000 miles from each other (indeed, we have this;
performance using the ttcp test tool is terrific, but I'll be damned
of anyone can find a _real_ use for a usable ~130Mb/sec.).

Hippi and Ethernet cannot claim this.

Part of ATM's "end-to-end" solution is its notion of Quality of Service.
I think ATM is going to have a run for its money when RSVP is available
on IP routers (are any vendors shipping this yet?).  And, for the
end-to-end problem, I can see Hippi or fast Ethernets connecting to an
RSVP-savvy router, which maps the RSVP parameters to QoS parameters
when the connection is established across the ATM WAN.

Well, I should probably quit now for ATM bashing on an ATM list :-)

Jason R. Thorpe                                       thorpej@nas.nasa.gov
NASA Ames Research Center                               Home: 408.866.1912
NAS: M/S 258-6                                          Work: 415.604.0935
Moffett Field, CA 94035                                Pager: 415.428.6939