[11099] in Commercialization & Privatization of the Internet
Re: uniform policies at the NAPs
daemon@ATHENA.MIT.EDU (Hans-Werner Braun)
Tue Mar 22 00:03:11 1994
From: hwb@upeksa.sdsc.edu (Hans-Werner Braun)
To: cook@path.net (Gordon Cook)
Date: Sun, 20 Mar 94 15:00:57 PST
Cc: swb1@cornell.edu, com-priv@psi.com
In-Reply-To: <9403202215.aa24465@pandora.sf.ca.us>; from "Gordon Cook" at Mar 20, 94 10:15 pm
>gee, hans werner aside from the issue of where you work and you own
>intrenched
>interests in keeping funding for your research projects coming, you mean to
>tell
I have never denied that. Not to you or anyone else. I thought I was
even pro-active letting people know where my interests are. Just read
the Journal article:
"NSF Implementation Plan for Interim NREN", May 1992 B. Aiken (NSF),
H.-W. Braun (SDSC) and P. Ford (LANL). Editor: K. Claffy (SDSC). SDSC
Report GA-A21174, Journal on High Speed Networking, Vol. 2, Num. 1,
1993
for example. I am working with NSF for quite a number of years. Over
the years I have received NSF funds for all of research, engineering,
and operations.
> me you have now solved the latency problems associated with the speed of
>light over wide area networks?
Not sure where you got that idea from, but I believe that latency in
and out of itself is not as much a problem. Latency jitter is much
more of an issue, and that has to do with resource contention. Read:
"Measurement Considerations for Assessing Unidirectional Latencies" K.
Claffy, H.-W. Braun, and G. Polyzos, Jan 1993 Journal of
Internetworking, v4n3. September 1993, UCSD Report CS92-252, SDSC
Report GA-A21018.
for example.
>I talked over some of these issues with a PhD physicist whom I have knwon
>since 1987 and who for longer than that has been a systems programmer for
>supercomputers... first for control data and since may of 1990 for Cray.
>
>He said he considered it ludicrous to assume that supers could do meaningful
>simultaneous parallel processing with each other over WANs. That is one
>machine working one part of the algorithm while the other works the remainder.
Check the diverse references about superlinear speedup for chemistry and
global climate modeling research on the CASA gigabit testbed before
going further on that train of thought.
>What would the centers use the vBNS for? Well center a could transfer a major
>database via the net to center b so that someone there could use it in a
>simulation and whne done return it to center "a"
I think the massive transfer of com-priv email will be sufficient to
justify two vBNS, don't you think? Again, read up on the CASA stuff.
>what about runnning a visualization of a process in real time and being able
>to "steer" the simulation based on the results of the visualization?
Yeah, that were ideas from years ago. You really have to do more for
integration today. E.g., Mosaic is quite multimediaish. NCSA gets at
times on the order of ten requests a second. That's not packets, those
all full blown application service requests. Typically lasting for at
most a few seconds. Wanna fire up SVCs for each? Just an example,
nothing special. The Mosaic stuff is just also something I looked at
yesterday for flow interarrival time graphs as accumulative
histograms. I can send you a postscript of a graph if you like. May be
able to do better by tomorrow, if I get off email. The change between
1993 and 1994 on the graphs is mind boggling. As is, for example, a
similar histogram graph for an about one hour long single Mosaic
session that a colleague of mine did for comparison reasons. For more
details on the inter supercomputer center collaboration check into the
MetaCenter stuff, of which some is available via Mosaic.
>His answer....very difficult but might become doable by the vbns if it had
>sufficiently few hops. Of course the same effect could be achieved by having
>the computational scientist on site. More than a few airplance tickets could
>be purchased for the $50 million I think.
Ask Steve Wolff how hard it is to get me to fly across the country for
a three hour meeting. I much rather spend three hours in San Diego than
in DC. Especially in winter. And in summer too, for that matter. May be
$50M is cheaper than the amount of neurons it costs? Especially if
done many times? Money just ain't everything. Besides, what's $50M for
enabling technologies compared to todays telecommunications
infrastructure. For comparison, how much did you say are AT&T and
Sprint getting out of the FTS2000 deal with the feds about providing
the federal telephone system?
I also see from my office often those Tomcats flying by on the way out
to the ocean from Miramar (Top Gun). Also nice free air shows on the
way to work. Sometimes also almost feels like you have to duck on I15.
Just for calibration purposes, how much did you say each of them
costs? An F-14D ain't cheap.
Besides, the issue is more laying the groundwork of how high
performance applications would work in the future. In the ARPAnet days
the packet switches and mailbridges were quite expensive, but the
research in the end paid off, as you were able to send your email,
which may have been received by hundreds or even thousands of people.
What was high performance then is nothing any more today. But it was
enabling technology.
>Now I know you will have technical knowledge that you will likely blow my
>obsvervations away with. What's more i won't wind up replying because I'll be
>off line for three days in a few hours at a meeting in DC.
Is that good or bad? Who are you meeting with?
>I know you and Scott can have fun doing more research. What I am less sure of
>is why you expect the taxpayerto underwrite you considering all the other
>problem this country is suffering from.... Just my 2 cents worth.
I doubt Steve would dedicate it to the two of us. But you sure seem to
be enjoying the fallout of the ARPA research while sending email and
making a living off it. And even using tax payer paid NSFNET
infrastructure while you are making your living and sending your very
note. And I think it is even acceptable according to NSF's AUP, as your
email is related to advancing the networking agenda.
Hans-Werner
