Re: ATM comes and goes

["James E. [Jed] Donnelley" <[e-mail]>]

>from Paul Ferguson <[e-mail]>:
>> Will it happen? Who knows. With the IP-over-ATM cell tax and the
>> advent of gigabit-ethernet, ATM as we know (or envision) it
>> could be a thing of the past in the future. (sic)

(chuckle...)

Responce from Albert Manfredi <[e-mail]>:
>Gigabit Ethernet might be a threat to ATM LANs. Then again, ATM in the
>LAN was never such a super idea.

You have that right.  As you know I am not sure it (ATM) is such a
super idea in the WAN area either.  Since you seem to be engaged
in this "debate", it might be interesting to discuss some
alternatives with you.  When I was in Germany last year I
worked on something that I called "Just In Time" switching.
This concept is designed to provide fast (both low latency
and high bandwidth) "virtual circuit" (though more 'real' than
ATM's SVCs) connectivity making use of Wavelength Division
Multiplexing with very low overhead - but still electronic,
in fact bit serial per wavelength - switching.  If you are
interested in such things I would be happy to discuss them with
you.  I might even be able to dig up something in writing on it,
though nothing has been published on this concept.  I am not wedded
to this particular approach, but I do think it is time to begin
looking at alternatives to ATM for high speed wide area networking!
As you note, alternatives for the LAN area seem to be abundant
(including HIPPI ["Super 6400 or no"], Fibre Channel, and SCI/SCX,
even beyond the noted and promising gigabit Ethernet developments).

Except for his disparaging comment about the bandwidth
capabilities of todays PCs, I essentially agree with
the comments from Mike Gawdun responce to the note from
Dharmesh Shah:

>Dharmesh Shah <[e-mail]> wrote:
>>I have heard that the next killer-app for ATM will be the internet.
>>Sounds good.  This should include internet from home.  I would hope that
>>the telcos, the cablecos, the utilcos would provide me with access to the
>>internet via ATM on my PC.  This could be at flat-costs depending on the
>>MCR I ask for when subscribing to the service.  There are many bandwidth
>>mongers for whom NISDN is not enough (BTW, I still use a 2400 baud modem).
>>ATM could enable ISPs to provide very flexible data-rate vs. pricing
>>options.  Is this happening now?

Responce from Mkie Gawdun <[e-mail]>:
>You've got to be kidding! Right?

Sounds like a joke to me too!  For one thing I can't imagine
even "browsing" the Web with a 2400 baud modem.  I find that
with a 28.8 modem my performance at home is nearly always as
good as my performance at LLNL where we have a very high speed
feed (T3+) to the "Internet."  While I am definitely a bandwidth
junky, I have very little motivation to upgrade even to ISDN
because I don't expect much of a performance improvement (yet).

I also agree that there are so many practical problems with
effective use of ATM SVCs (e.g. as below) for Internet application
that it seems unwise to me to deploy ATM for such use - except
perhaps as an easier way to provision pipes between ISPs routing
IP packets (accept the "cell tax" in return for 'circuit'
provisioning that is quicker and more flexible than provided by
SONET with ATM CBR VCs in place of SONET Telco provisioned SONET
circuits).

From Mkie Gawdun <[e-mail]> again:
>I just returned from the ATM Forum meeting in San Jose. There are many
>issues unresolved about running IP traffic over ATM. Also, the IP world
>views ATM completely differently than the ATM Forum, so these issues need
>resolution also.

Agreed!!

__________
From Mkie Gawdun <[e-mail]> again:
>You have another problem, the bus speed of your PC is incapable of
>supporting 100 Mbps Ethernet at throughputs above 20 Mbps, so how useful
>will it be to connect to an ATM network when the PC can't handle the firehose.

The above comment seems generally irrelevant to me.  PCs can
get faster if need be.  DSPs or DSP equivalents can be put in
to support fast multimedia if the needed network bandwidth
is available.  Also hardware protocol support can be put in as
well.  I just don't think the above is important except in a very
narrow and essentially irrelevant time window.
___________

From Mkie Gawdun <[e-mail]> again:
>If I can't get past the Internet backbone congestion and Web site port
>contention that I experience everyday on a 128 kbps ISDN connection, then I
>will certainly not get better throughput with an ATM connection to the home.

Agreed!!!

From Mkie Gawdun <[e-mail]> again:
>If you still use a 2400 bps modem today, you will see a dramatic improvement
>with an ISDN line. Forget the ATM stuff, it is too far off. Most of the ATM
>vendors today are building gigabit Ethernet switching products and others
>are building IP switching devices to overcome the astronomical cost of ATM.
>If the vendors can't get ATM costs down to the level of existing competing
>technology, then ATM may be put on the backburner even further into the
>future. Technologies like Frame Relay and Fast Ethernet are outselling ATM
>by a factor of 20:1, so it will be a while.

Agreed!

>Mike Gawdun

Nice comments Mike!  They certainly dovetail (with the one
exception) with my thinking!


--Jed   http://www-atp.llnl.gov/atp/jed-signature.html