[clue-tech] distributed.net

[mike havlicek]

Hello,

At least a number of years ago the University of Wisconsin CS dept was doing something like this with a project called Condor. What I remember
hearing about it or what stuck out as one use was that they were using on campus student lab machines off hours for scientific computing.

-Mike 

--- On Thu, 7/15/10, Nate Duehr <nate at natetech.com> wrote:

> From: Nate Duehr <nate at natetech.com>
> Subject: Re: [clue-tech] distributed.net
> To: "CLUE technical discussion" <clue-tech at cluedenver.org>
> Date: Thursday, July 15, 2010, 12:57 AM
> On Jul 14, 2010, at 10:51 PM, Jason
> Ash wrote:
> 
> > About six months ago I discovered a way to donate my
> idle CPU cycles to science through distributed.net. The two
> current projects are breaking the RC5-72bit encryption
> algorithm and Optimal 27-Mark Golomb Rulers. Basically, they
> are building the world's largest supercomputer cheaply
> through distributed computing.
> 
> Building?  Haven't they been around for about 10
> years?  If I remember correctly they were the first to
> crack DES at some level of bits... and then again at a
> higher level of bits... making the push to AES a reality for
> panic-stricken government folks... 
> 
> Mostly I stopped "donating" my CPU cycles to this and
> learned to have the machines go to sleep when not in use,
> after a) realizing that many of these applications are
> horrible about releasing your own CPU *back* to you when you
> actually start using it, and b) because electricity is well,
> used heavily by a machine (or five or six like I once had
> running distributed.net, SETI, who knows what other "fad"
> distributed platform prior to them firing up the new and
> improved marketing term for such things... "Cloud
> Computing"...) and I decided I'd stop payin' their power
> bill for things THEY wanted to find with computers... and
> pay the power only when I needed my computer to do things I
> wanted... 
> 
> I also started thinking deeply about... "Why would I want
> to escalate the crypto arms race?  It just leads to
> more bits of crypto, and requirements for never-ending
> faster processors, and that costs me money too... I have to
> keep upgrading machines 'round here to keep up!  Give
> me a break.  HELP prove that modern crypto will
> continue to fall again and again, and giving the researchers
> hard numbers as to just how easy/fast it'll be?  Not
> really all that smart from a personal finance point of
> view.
> 
> Of course, nowadays, the cool kids are doing all of this in
> GPUs (graphics cards), stuffing multiples in their machines
> and using them for insanely fast parallel number/factoring
> processors, since they do factoring far far FAR faster than
> the typical CPU/general purpose RISC or CISC
> processor.  Imagine row after row of machines stuffed
> with graphics cards crunching numbers, not displaying
> things, at your local "interested agency"... whoever you
> like to name for that job... 'cause they're doing it. 
> A lot of it.  And then look at the prices of graphics
> cards, and think "government is buying these"... and you'll
> see the cost curve typical of a product being bought by
> government... and you'll see the ever-increasing
> performance... 
> 
> You didn't think the graphics card makers were dumping
> millions into R&D so we could all get 60 fp/s in 3D
> games, now really, did you? ;-)  (Just look at their
> 10K filings at EDGAR...)
> 
> (Side-note: "Cloud Computing" as a general rule makes me
> laugh at bit, and also throw up in my mouth a little at the
> same time, like all good Marketing campaigns probably should
> do.   Pssst... hey you... Mr. IT manager...
> Want a mainframe?  I hear they're makin' a
> comeback!  Hahaha... again... it's only recently that
> I've gotten former Sun employees to tell the ghastly stories
> of running those "thin client" things Sun was pushing a
> while back... at Sun themselves... network hiccup, main Sun
> boxes down, whatever... you'd walk in and everyone's
> terminals were down, as well as whatever work you hoped to
> do that afternoon...)
> 
> Fun stuff, watching software like distributed.net do its
> thing, but it's been around a great long while... nothing
> wrong with helpin' them out if you feel so-inclined... but I
> think I'd rather help someone out that's doing research on
> helping humans with diseases, etc... today... over cracking
> crypto... again... and again... and again... just from an
> ethics standpoint.
> 
> --
> Nate Duehr
> nate at natetech.com
> 
> facebook.com/denverpilot
> twitter.com/denverpilot
> 
> _______________________________________________
> clue-tech mailing list
> clue-tech at cluedenver.org
> http://cluedenver.org/mailman/listinfo/clue-tech
>