Archive of UserLand's first discussion group, started October 5, 1998.

Re: 100 websites (SETI@Home)

Author:Steven C. Den Beste
Posted:1/20/2000; 12:26:58 AM
Topic:100 websites that will change the world
Msg #:14631 (In response to 14600)
Prev/Next:14623 / 14633

While this kind of distributed computing has been done before, no-one has done it at the kind of scale involved here. By the way, a much earlier distributed project was done at Apollo by "the Midnight group" IIRC which used spare cycles on the workstations at the company to raytrace a short animation called "a Long ray's journey into Light". For its time it was spectacular, and far beyond anything seen before. This happened in 1985.

From a technical point of view, it was a superb application of parallel computing. Ray Tracing is almost infinitely parallellizable; at the most extreme case you'd use a separate computer for each pixel of each frame in the film.

This particular effort involved a couple of thousand computers running for several months. For its era, it was immense. But the computers (mostly 25 MHz 68000's) were hardly what we'd call fast nowadays, and there really weren't all that many of them. Nonetheless, it was vastly larger than anything like it done before that.

Even earlier, there was Iliac IV, a 64-processor behemoth in the late 1960's which collectively had about as much compute power as the 450 MHz PII I'm typing this on. It's amazing how far we've come. Iliac IV's great claim to fame was computation of thermal problems [calculation of how objects heat up and achieve equilibrium when unevenly heated, a significant problem in the aerospace industry], also a good problem for parallel processing.

distributed.net has at most a few tens of thousands of active participants. According to their web page, they're at about 44,000 active participants on the RC5-64 crack, 225,000 total. Seti@home has had upwards of two million people contribute at least one work unit, and there are nearly 450,000 regular contributors (who have contributed at least one work unit within the last two weeks, which considering the immensely longer amount of time a work unit takes, is a fair comparison; also, that number has been stable for weeks now). We achieved break even a long time ago; we're now actually processing data far faster than they collect it (about twice as fast: 200K work units per day collected, and we're processing over 400K per day and have been for a long time) and we're burning off the backlog. (In fact, I think we caught up.) As a result, they're going to be introducing a new version of the client soon which actually spends *more* time and looks even harder at the data. It's most amazing.

Sometimes a difference in degree becomes a difference in kind. This is such a case. SETI@Home is at least ten times bigger than the RC5-64 crack. They've been breaking a lot of new ground on how to manage these kinds of projects, particularly in dealing with scaling issues. Whether it actually creates any scientific result (i.e. discovers an anomaly in space) or not, it is already a success in terms of creating new technology in computer science.

On the other hand, it's not clear that something more esoteric like modelling proteins, or simulating atom bomb explosions, which are equally capable of sucking teraflops, would attract the kind of audience participation that something sexy like SETI has attracted.

To some extent, that's why distributed.net didn't get this big; to the average computer owner, "Want to help crack RC5-64?" gets a bewildered stare in response.

Ah, but looking for ET; that's something people understand. Even non-technical users are willing to participate in something like that, just because of the romance of it all. (And it's no accident that the graphics on the screen saver are whizzy even though it triples the time spent per work unit while the graphics are being displayed.)

I'm a participant in SETI@Home, though I never had any urge to participate in cracking the various encryption challenges. I tend to crank a work unit about every 11 hours, and so far I've finished 680 of them, which puts me well into the 99th percentile of all participants. I started a couple of days before the official kickoff, and for a while I had three computers running simultaneously. Because of this project, I started leaving my home computer on 24/7. Since it has a cable modem, it's always connected to the net. Right now I have more than a year of cumulative CPU time (9628 hours, about 385 days).

From the point of view of a mathematician, SETI@Home is a "good bet", which means that the payoff if we win multiplied by the chance of success is higher than the cost of playing the game. (Partly that's because we've *already* paid off more than it cost to play. From here on out, we're playing with the house's money.) While I consider the chance of the project actually finding an extraterrestial civilization to be remote in the extreme, I also think that the scientific, cultural and sociological ramifications of a positive result would be incalculably large. Quite simply, it would be the single most important scientific discovery of all time, if it was confirmed. (The religious ramifications alone are mind boggling; did Jesus visit the Rigellians after he left America? )

So it's unlikely that any other project following the path that SETI@Home is blazing will have the success at attracting participants. But even at a much lower participation rate, it's still an amazing concept and will be extremely valuable. Future projects will, interestingly enough, succeed or fail on their marketing. Since they are basically attempting to attract swarms of volunteers, the only way they can pay them is with glory. That's why atom bomb simulations and protein analyses aren't going to do well.

One of the most amazing aspects of the SETI@Home project is that it's being run on a shoestring. The total amount of money which has actually been invested is preposterously small for the result they've gotten out of it.

Distributed.net pays with cash; if your computer is the one to luck into the key, you get a share of the prize. The evidence is that SETI@Home is attracting more people with romance than Distributed.net attracts with a long shot chance of winning a few thousand bucks. Someday companies will stop offering cash prizes, and the incentive model being used by Distributed.net will collapse. But glory is available in infinite quantities, if we on the SETI@Home project win.

[By the way, it's a damned good thing that "@Home" hasn't objected to the use of their trademark. Of course, it would be a major public relations fiasco to pick on SETI@Home, e.g. etoys.com recently.]




This page was archived on 6/13/2001; 4:54:09 PM.

© Copyright 1998-2001 UserLand Software, Inc.