Distributed Computing

[mrnoodle]

Distributed computing is a neat idea, I think. Not sure if it's been covered in the Cellar yet, but what do you think the implications of it are?

The fine folks at Stanford U. are trying to understand protein folding and how a misfolded protein can result in things like Alzheimer's disease, Parkinson's, etc. The science is beyond me, but apparently, simulating the protein-folding process is quite a resource hog. So, a prof at Stanford has implemented Folding@Home, wherein you can donate your unused CPU cycles to their timescale simulation, maybe even helping cure a disease in the process.

How cool is that? It's like the photo-negative of a pyramid scheme.

[breakingnews]

The concept has been around for a while now - probably even at the root of the Web's early spinnings. No time to research specifics, but I remember there being a couple of sites - some even offering to pay - asking to donate idle CPU time to calculating complex mathematical equations and the like. Seeing that college was the first time my computer was hooked up to a high-speed line, it raises the obvious issue of bandwidth. Broadband penetration is soaring - something like 40 percent of US households by 2010 - but things like shared computing could significantly congest networks run by cable and DSL operators, where most people are still getting 1.5 mbps rates or so. Companies and schools will likely ban it, and the average American probably won't know enough to consider installing the program, or they will consider it a backdoor for spyware and hackerz.

Not a bad idea, but will computer makers resist? Will the nation's infrastructure get built up enough to adequately support what's needed? Does it pose security risks?

[Griff]

Good topic. Quite a while back we put a cellar team together to provide comp power for umm.... something UT actually put a lot of resources into it. I was thinking about suggesting it again for the right organization.

[BigV]

Originally Posted by breakingnews

The concept has been around for a while now - probably even at the root of the Web's early spinnings. No time to research specifics, but I remember there being a couple of sites - some even offering to pay - asking to donate idle CPU time to calculating complex mathematical equations and the like. Seeing that college was the first time my computer was hooked up to a high-speed line, it raises the obvious issue of bandwidth. Broadband penetration is soaring - something like 40 percent of US households by 2010 - but things like shared computing could significantly congest networks run by cable and DSL operators, where most people are still getting 1.5 mbps rates or so. Companies and schools will likely ban it, and the average American probably won't know enough to consider installing the program, or they will consider it a backdoor for spyware and hackerz.

Look, the projects just want your CPU cycles, not your bandwidth. Sure you have to transmit the results, and get the project elements, but each of these parts is trivially small compared to the time required to calculate the results. I participated in the SETI project years ago and processed many hundred of work units.

Basically, you connected to their site and downloaded a small program. This program was designed to grind out the calculation for a given set of parameters.

It was tunable to permit the user to decide how much of the cpu load could be consumed by the process. I could configure it so that it only worked when my screensaver was on. That meant NO impact to my computing power since the cpu cycles at a constant rate, and most of the time it's doing nothing. And when the machine is sufficiently idle that the screen saver has time to kick in, it means that I'm not needing the cpu cycles either. So it computes away, sometimes taking a day or more to complete one work unit. Then, when it's finished, it connects to the site again, uploads the finished work unit (the results, if you will), and downloads another work packet and disconnects.

It takes lots of CPU cycles, not lots of bandwidth.

Originally Posted by breakingnews

Not a bad idea, but will computer makers resist?

Why in the world would a carmaker resist increased use of cars?

Originally Posted by breakingnews

Will the nation's infrastructure get built up enough to adequately support what's needed?

Uh, like, do we have enough computers? Some problems are more ameable to parallel processing. The ones that are will benefit from greater distribution of machines running their computing client. Other problems don't fit this mold and would not be solved any faster even if there were more machines available.

Originally Posted by breakingnews

Does it pose security risks?

That's a good question. This program does have some similar characteristics that other malicious programs have, like autonomously connecting to a given site, but so does Windows Update. Same old answer here, it depends. You should carefully examine what you're inviting into your system first.

[breakingnews]

Originally Posted by BigV

Look, the projects just want your CPU cycles, not your bandwidth. Sure you have to transmit the results, and get the project elements, but each of these parts is trivially small compared to the time required to calculate the results.

I guess I was just thinking fatalistically here: How big will shared computing get? Yes, transferring a small packet (or several small packets) of data is minute relative to your typical Internet usage (or is it?). But as more and more computers come online to serve such a function, there's more opportunity to bottle up what's available. What if a program decides your computer can handle two, three or 500 simultaneous calculations, or staggers them so there's a constant stream of data flowing in and out?

Why in the world would a carmaker resist increased use of cars?

It's not increased use of computers, it's more efficient use. What if everyone in a community shared a handful of cars? Carmakers could only make their money by building more expensive, powerful vehicles able to tolerate the extensive use (hypothetically 24 hrs a day) and reducing car life to, say, 2 or 3 years, but that would severely restrict their annual volume growth. Or, it could resemble an erie outsourcing scenario: why buy a Cray when you have thousands of cheap Dells - which you don't even have to pay for - at your disposal via network?

These are extreme examples. Admittedly I do not know much about this subject. I was bored at work and dreaming of the apacolypse in hopes that at the very least, my editor might burst into flames and die. Just being a royal pain in the ass.

[BigV]

Originally Posted by breakingnews

I guess I was just thinking fatalistically here: How big will shared computing get? Yes, transferring a small packet (or several small packets) of data is minute relative to your typical Internet usage (or is it?). But as more and more computers come online to serve such a function, there's more opportunity to bottle up what's available. What if a program decides your computer can handle two, three or 500 simultaneous calculations, or staggers them so there's a constant stream of data flowing in and out?

A continuous flow would be ok, given that it did not impair the main function of the computer. If it did, then I would have to step in and readjust it's allocation of resources, namely the cpu cycles. I could also adjust the connection method by only permitting it to connect when I gave it explicit permission to do so.

You see, the cpu cycles "cost" the same whether they're doing an actual calculation or not. Might as well put them to some useful purpose. There isn't a perfect analogy I can think of, but this is close. Imagine an escalator. It's moving up, or down, continuously. It moves at a constant rate irrespective of the number of people on it. Within it's design limits, it costs the same to run it full or empty or anything in between. The cpu works that way too. It checks it's "pipeline" for any work every tiny fraction of a second and does what it's asked to do. By the way, all those dazzling megahertz and gigahertz refer to the number of times per second that the cpu checks to see if there's some math to be done. Millions (mega) or Billions (!) (giga) of times each second it checks. The VAST majority of cpus are starving for work.

Originally Posted by breakingnews

It's not increased use of computers, it's more efficient use. What if everyone in a community shared a handful of cars? Carmakers could only make their money by building more expensive, powerful vehicles able to tolerate the extensive use (hypothetically 24 hrs a day) and reducing car life to, say, 2 or 3 years, but that would severely restrict their annual volume growth. Or, it could resemble an erie outsourcing scenario: why buy a Cray when you have thousands of cheap Dells - which you don't even have to pay for - at your disposal via network?

Bold prediction: Won't happen.

NO trend in technology has shown any dimunition in the rate of acceleration wrt speed, size, quantity, capacity, etc. Certainly, there is a validity to the concept of "a sufficiency of computational power" (I guess) but what has always happened is that when a previously unreachable boundary has been crossed, it has always led people to wonder what was over their "new" horizon. There will be an steady desire to know the answer to ten or a hundred more decimal places of accuracy, or to have the answer faster. As those limits in turn are reached, new "must reach" horizons are revealed.

Newton's aphorism, "If I have seen farther, it is by standing on the shoulders of giants." has some miles in it still.

Originally Posted by breakingnews

These are extreme examples. Admittedly I do not know much about this subject. I was bored at work and dreaming of the apacolypse in hopes that at the very least, my editor might burst into flames and die. Just being a royal pain in the ass.

It could happen...
http://theshadowlands.net/spon.htm

But it would be a mixed blessing, considering the additional carbon load. I looked everywhere and could not find a calculator to do the math. But overall, I think I could reduce our family's load by a little more than a tonne.
http://www.climatechange.gc.ca/oneto...lish/index.asp

[russotto]

Originally Posted by BigV

NO trend in technology has shown any dimunition in the rate of acceleration wrt speed, size, quantity, capacity, etc.

CPU speed and transistor density seem to be hitting a wall about now.

Removable media hit a really solid wall a while back, and now it's far behind, leaving lots of people with drives they are unable to easily back up.

[tasadar24]

Actually, my school district prohibits use of Distributed Computing programs on it's computers. My school's ECS said that the reason was that they didn't want the load that thousands of Distributed Computing computers would place on our network. It makes sense, I believe CCSD is the largest school district in the country, and we're all running through the same proxy into the internet at large.
-Clark County School District

[mrnoodle]

The entire internet is a shared computing model, really. And if we run out of resource, we'll make more. Yay capitalism.

As to security, I would think that something like this would require that you tinker with your firewall or reduce the security setting on your virus protection, which i'm not comfortable with. The person I got the link from suggested to our company that, since we don't power down our desktops at night, we corporately decide to contribute to this project. I don't see it happening...

[BrianR]

Grid computing is not new, I participate in this (or some very similar project) through Grid.org. I watch proteins folding and whatever analysis is done on them on my screensaver all the time. My old computer takes the better part of a week to finish one block, but so what?

I have no problem with this because I know what the program is doing, I can configure it to use only certain ports and allow those through my firewall. I can help research without any real cost to me. All I need to do is leave the computer run all the time (I do anyway) and let it connect to the Internet at will. Simple enough.

Others may have a different opinion though.

Brian

[Wombat]

blu-ray schmu-ray.

Witness the Holographic Versatile Disc, storing more than a terabyte!

http://www.theregister.co.uk/2005/02...iance_founded/

[BigV]

Originally Posted by Wombat

blu-ray schmu-ray.

Witness the Holographic Versatile Disc, storing more than a terabyte!

http://www.theregister.co.uk/2005/02...iance_founded/

Woof!

[wolf]

Not to denigrate your district at all, but how could a school district that I can't even place in a state be larger than, for example, School District of the City of NY or Philadelphia? How many students do you serve?

[xoxoxoBruce]

Nevada, Tasadar24?
Oh, and welcome to the cellar.