[2193] in Commercialization & Privatization of the Internet
Re: Are We to Buy Speed for a Few or Connectivity for Many?
daemon@ATHENA.MIT.EDU (Karl_Kleinpaste@N2.SP.CS.CMU.EDU)
Wed Jan 29 15:00:43 1992
From: Karl_Kleinpaste@N2.SP.CS.CMU.EDU
Date: Wed, 29 Jan 92 14:49:11 EST
To: sean mclinden <sean@dsl.pitt.edu>
Cc: com-priv@psi.com
In-Reply-To: sean@dsl.pitt.edu's message of Wed, 29 Jan 92 13:15:29 -0500
Date: Wed, 29 Jan 92 13:15:29 -0500
From: sean@dsl.pitt.edu (sean mclinden)
"Researchers need gigabits, NOW."
I'm curious to know:
1. Who, besides the people building gigabit network testbeds, are these
researchers and what are they doing?
2. What is the social, technical, economic, and political impact of their
use of clear-channel gigabit bandwidths in comparison to that of the
hundreds/thousands of people who have an aggregate need for this band-
width (or rather, *might* have a need for it if they ever got connected).
1. For the Nectar testbed's applications, I refer you to this paper:
"Parallelizing a New Class of Large Applications over High-Speed Networks"
HT Kung, Peter Steenkiste, Marco Gubitoso, Manpreet Khaira
Carnegie-Mellon Univ, School of Computer Science
This is an overview paper regarding parallelization methodology, and
includes discussion on 3 applications running on the Nectar prototype.
They are
COSMOS, a logic simulator used, e.g., to simulate the
650K-transistor iWarp chip
NOODLES, a geometric modeling application
Los Angeles air pollution (e.g., weather forecasting issues)
The whole paper including bibliography is only 11 pages; it's a good
start to see what Nectar and its applications are up to. It can
probably be ftp'd from somewhere around campus, but I don't know where
or how just now. Sean, since you're in town, if you like you can
wander over to CMU campus sometime soon and I'll just hand you copies.
Other applications of which I'm aware (and mind you, I'm pretty
distant from the applications up to this point, both due to my work
[kernels] and my lack of exposure [only been here ~6 months]) include
chemical flowsheeting to simulate processes in a chemical plant, and
civil engineering problems involving pressure analyses and so forth.
Frankly, my math background is not up to proper explanation of what's
being done here. Perhaps other Nectar folk are on this list and will
contribute a short synopsis of the problems.
2. Hard to quantify your request regarding these applications. Not
that it can't be done, but I am probably not close enough to the
applications to do it properly. Economic issues of highly
parallelized, network-based applications typically involve drastically
reduced time-to-solution for very large problems, such as the chemical
flowsheet issue -- go from need for a new chemical product to working
solution that can be implemented in a plant much, much faster. Things
like COSMOS have rather obvious application, getting chip
manufacturers closer to their solutions in new chips faster. Social
effects are sometimes easier to describe than economic or technical:
For the chemical flowsheet problem, one of the major requirements of
any solution revolves around environmental impact, so use of (Gigabit)
Nectar will involve improved abilities to avoid environmental problems
-- more solutions examined, better solutions chosen.
If I had anything that _could_ be said about the politics of it all, I
doubt I'd say it. But I don't, which obviates the situation for me.
There are days when being a peon is good. :-)
Note, I _work_ for Gigabit Nectar, but I don't _speak_ for either of
CMU's Nectar or iWarp groups.
--karl
