Slashdot Mirror
More On Paid Distributed Computing
Nastard writes: "Theres a story over at C-Net News.com about making money with distributed processing. The article talks about several companies that are planning to launch per-per-idle projects this fall. Apparantly someone has finally caught on that there is money to be made in this. No surprise that one of the companies is headed up by SETI@Home founder David Anderson." I've always been a fan of distributed.net -- (Subliminal Message: Sign up for Team Slashdot!), but I do wonder with these pay schemes if the payment will actually be enough to cover the cost of electricity. Hurm.
[timothy butts in ...] Also, you may want to check this out. A semi-anonymous reader writes: "Distributed.net President David McNett recently did an interview here with the guys over at Geeknik.net. In the interview, he discusses his role with Distributed.net, future projects they are going to work on, and how he views competition between the various distributed computing organizations. Great read."
Google sells search engine services to other companies to place on their web sites such as Redhat, Yahoo, Latino.com, the Washington Post, eToys, eGroups, WebVan, etc. These search engines include a good amount of customization and Google gets a monthly fee out of it of upwards of $2,000.
The world is neither black nor white nor good nor evil, only many shades of CowboyNeal.
I just think it's a shame that it can only be done once. By that I mean, there's a huge amount of untapped CPU time out there, but once a system is in place to use it that's it, you can't use it twice. It's like compression -- one go and it doesn't recompress. Where compression "doubles" the storage capacity of a system, distributed computing "doubles" the efficiency of a computing system. From there on it's just tiny, incremental upgrades...
People allready offer banner clients.
Some make money some banners trade for services (free ISPs)
The problems with them are the timeout.. forcing you to click on a banner every once in a while.
If instead you were asked to trade cycles for bandwith..
You know who to call if they have a back door...
Same rules that keep open source develupers from sleaking back doors into software will keep closed source clients from doing anything more than using your procesing cycles...
If you discover otherwise... class action lawsute...
Then you make some sereous money...
This is good. Admitedly for closed source.. it might hurt open source a tad..
But it burns Microsoft...
Basicly why buy Netscape or Opra when you have IE for free?
Thats what keeps Netscape free... but... Netscape with a distributed client... can generate money and be free..
It could also mean we'll see more Linux software and games ported to every Unix platform including Solarus.
You are sharing processing cycles in trade for software... Linux, BSD and Solarus boxes tend to be left on and running while Windows boxes tend to be turnned off..
This is less a stability issue and more a geek factor... Macs get in the 50/50 range as they have a strong Geek, Busness and newbe attraction..
Avrage people turn things off.. lights.. TV... radio etc...
Windows users turn computers off when they are done like turnning off a TV or turnning off the lights...
*nix users tend to leave computers running as if it were vital hardware...
Any computer can run as an alarm clock.. *nix and Mac systems are however more likely than Windows to run as such...
Simply becouse of the kinds of users who run Windows...
Anyway so this translates into prefering systems that stay on all the time normally over systems that get turned off...
Why support 11,000 users who only provide 1 to 2 hours of idle cycles... vs 1,000 users who provid 24 hours of idle cycles a day... in some cases 48 hours of idle cycles a day.. (Two computers... and the occasional time bender)...
It costs money to support users... and while *nix users support themselfs you gota convence the avrage CEO of this..
So basicly paying to support 1% of the users who represnt 80% of available idle procesor time is a really good deal..
and once they realise we don't accually need support... it'll be even better...
I don't actually exist.
I doubt any company can sneak a distributed computing client onto a PC; it takes one person to notice and publish their findings. The backlash against the company would be huge.
I occasionally run a load monitor, and I suspect many others do.
On a device like a WebTV client, there is no way to run a load monitor. The software is completely controlled by the manufacturer, and probably embedded into the device.
In addition, most of the people accessing the internet through such a device are unlikely to have the computer literacy to notice, or even care. So it makes sense to exploit the idle time of a device like that.
Nix absolutably seriousness.
but I do wonder with these pay schemes if the payment will actually be enough to cover the cost of electricity. Hurm.
I quote the distributed.net FAQ:
- Michael Cohn
-----
Go ahead, blame me... I voted for Nader!
What is thought of running these types of pay-for-idle processor time setups at work. I know that a lot of people set up distributed.net or SETI clients at work (machines not owned by themselves) I wonder if companies will mandate that all computers run a clients that pay you back. The money could significant if you are at a site with several hundred computers. And what are the implications of making money off your employers spare computer cycles?
What the Slashdot story and the underlying CNet article don't mention is that nobody is actually getting paid yet. I got all excited when I read these stories, and I proceeded to visit every company site named in the CNet story. None of them are paying out yet, and none of them even have pay rates figured out. Save your time.
Going further, I did a search on Yahoo, and hit just about every company listed under Distributed Computing. None of them are paying out - they're just taking in money from investors.
The Slashdot story is only partly right - there is indeed money to be made from this idle-cycle scheme, but it isn't going to be made by folks like you and me. It's being made by the companies who are suckering investors into this. Of all the sites I went through, I counted a rough total of over sixty million bucks in funding that the companies had gotten from investors. And not one dime has been paid out yet....
What's your damage, Heather?
There was a thread about the problems with open source and hacked clients in an earlier discussion. I'm not sure if it's much help.
===
If you look at Andover.net's financial disclosure reports, you'll find that Slashdot, which earns revenue through solely ads, is profitable. Therefore the model, if properly executed, can be profitable.
Interested in seeing it sooner? mojonaton is an opensource project.
The most difficult part (IMHO) of writing a payment for cpu cycles type of system is sandboxing the code so that it can't intentionally or unintentionally do anything evil (ie: it needs to be MUCH more secure than Microsoft Outlook ;).
Also, it didn't really get much into the fact that not all problems are particularly well suited to distributed computing. In fact, I'd say the majority of all computation problems AREN'T easily suited for distributed computing. Either because you'd lose the computational benefits due to the overhead of transmitting really large blocks of data or the algorithm is too dependent upon being feed serial data.
thick if you think your (landlord, dorm, etc) isn't going to pass the cost to you.
It's averaged out into one rent figure that everyone in the apartment/residence hall pays. Think "local phone service in the US" as opposed to "long distance phone service."
<O
( \
XGNOME vs. KDE: the game!
Will I retire or break 10K?
I'd let them have my CPU cycles for free if they'd hook me up with free DSL!!! That would be something I would hop onto. My computer wouldn't be much use to them anyway on the current 56K modem that basically gets connected when my wife checks her e-mail. Give me the DSL, I get fast internet access, you get my free clock cycles, everyone wins :-)
Do not teach Confucius to write Characters
Today's Sesame Street was brought to you by the number e.
DCTI releases the source to its client apps' computing cores; hotshot assembly coders can get their names in lights by submitting a patch against the public-source core-only clients. The official client binaries, OTOH, are considered "trusted binaries" and may in the near future be digitally signed.
IMHO this is a nice compromise between ESR's open-source ideals and obs^H^H^Hsecurity issues. Why doesn't SETI do this?<O
( \
XGNOME vs. KDE: the game!
Will I retire or break 10K?
Also, don't count on too much information being released on the clients. Any company who is a client of one of these CPU-for-dollars service will want to believe their data is accurate and that the people running individual clients can't steal the majority of their precious intellectual property.
Most of the free 'for fame & glory' distributed projects wont even release source code because they are afraid it will cause people to send in garbage data results in an attempt to cheat to get to the top of the list. And putting in some sort of fancy security layer on top of the client to make sure data is valid would likely invalidate most of the gains of a distributed project, because you'd burn tons of cycles on the server trying to do re-validation...
It's a bit harder to reverse-engineer binaries when they're digitally signed and 1024-bit elliptic-curve PK encrypted.
<O
( \
XGNOME vs. KDE: the game!
Will I retire or break 10K?
DCTI does release the source for d.net test clients (so that all the hot-shot asm coders can improve the cores), but test clients don't connect to the d.net servers. Only trusted binaries connect to the servers.
<O
( \
XGNOME vs. KDE: the game!
Will I retire or break 10K?
This brings up something I hadn't thought of. If paying for distributed CPU cycles becomes the norm, and every appliance you ever buy eventually has a powerful computer+internet connection, you could feasibly get those appliances for free.
:)
Imagine walking into your average Best Buy and EVERY applince there is free. Buy(for free) whatever you want, the only condition is that you hook it up and use it. The internet connection will be free and you will never notice the use of the appliance as the computer inside it is horribly wasted anyways.
No down payment, no monthly fees, all electricity, phone bills, and internet bills paid for. Just make sure its hooked up.
This sure would keep the economists shaking their heads in agony.
-Restil
Play with my webcams and lights here
...but not in a FREE SOFTWARE® way. For example, distributed.net releases the source to its computing cores; asm wizards have optimized the cores to run on AltiVec, 3DNOW!, etc. However, only d.net's trusted binaries connect to the official servers.
<O
( \
XGNOME vs. KDE: the game!
Will I retire or break 10K?
Does it worry anybody that most of these kind of projects coming down the pipe will be run by corporations that most likely won't release the source to the client software?
How will we know with confidence that we're not signing up to be a part of Echelon or Carnivore or something similar?
I guess if the cash incentive is good enough, it won't matter for most people...
Anyone who has ever listened to an AM radio placed within 5 feet of their computer can tell you that no matter how fast you work, you are rarely if ever going to be maxxing out your machine's performance. Wether busy or idle, most computers spend the majority of their lives waiting for human interaction -- Not crunching numbers.
While the idea for large-area distributed computing is cool, of what practical purpose is it? Its not going to help me do the things I do any faster than I would normally. Its not going to help me browse the web any faster, read the daily news any faster, dial out any faster, sell CDs any faster.. The only sort of applications (it seems, perhaps i'm being short-sighted here) that could benefit from this sort of thing are 3D modelling/rendering applications, graphics apps, and scientific applications, all of which require tremendous amounts of computing effort.
To me, it would probably take more time for my machine to "distribute" a task and reap the result than it would be to perform the task singularly with my own box. Sure, if I had some enourmous image I was doing work on, it might be a good idea.. But for most things, the overhead incurred by distributing the workload over X number of remote machines via the net would be more time-consuming than just doing it locally on a single box with multiple CPUs.
Mind you, I have nothing against the idea. I think it rules. But i'm at a loss to find any particular mainstream usage for such a technology. Most of us are leaves on a tree, not branches. We're seated at the end of the line when it comes to the life of a particular piece of data.
By the way, the 1998-2000 PROPAGANDA Image Archive CD is now available. Just click the "Enjoy!" link below for more info.
Bowie J. Poag
Bowie J. Poag
I would /consider/ signing up for a program like this, provided:
1) I have full, immediate disclosure about *exactly* what sort of computations and calculations *my* computer is being used for (this is my largest gripe with ProcessTree, and why I *refuse* to sign on with them).
2) Full source code (for security auditing). I'm not sure yet how this could be used with a system to prevent user fraud (and I understand the desire to distribute only binaries). Perhaps an encryption verification scheme, using PGP or such?... This may be a problem - but I simply do not trust closed binaries running strange unknown processes for mysterious third parties on *my* systems.
Without these conditions, it is simply **too much** to ask me to compromise for a few extra dollars here and there.
My question is, who will be the first sysadmin to get fired for pushing distributed clients on all the corporate workstations without the bosses knowing?
On the legitimate side of this, the company that is most likely to succeed is the one that markets itself best to corporations. In my limited experience, most corporations just run lame screne savers. If you go to them and explain that they are losing money running screen savers, they will be happy to oblige.
On the sinister side of this, the distributed processing company that integrates an employee monitoring system into the client and markets that to large corporations will probably do pretty well also.
You don't have to cover the total cost of the electricity anyway. You just have to cover the difference between the cost of power consumed at idle now, and power consumed running the client. The question then hinges on how well the "power saving idle mode" works on most PCs. Anybody got the numbers?
Assuming that the numbers make it beneficial for corporations to run the client, the question may then be one of whether or not there is sufficient demand for distributed services. Right now, render farms and weather forcasts are the two obvious commercial applications. I think we're going to have to think of a few more if distributed computing is to become a significant industry.
For all intensive purposes, "whom" is no longer a word. That begs the question, "who cares"?
I do wonder with these pay schemes if the payment will actually be enough to cover the cost of electricity. Hurm
Nope
I've been at this for a while *chuckles*... and I'm working on a longer whitepaper, but I'll give slashdot a quick preview.
The real (hidden) costs:
Why no company would ever touch it:
So you have huge hidden costs, total loss of any capitalist advantage, and huge headaches for the admins. Keep dreaming folks.
- Adam L. Beberg - The Cosm Project - http://www.mithral.com/
Can open, worms everywhere:
So maybe the toaster can tailor the ads to match your food choices? If you put a lot of bread in the toaster, will you be shown ads for jelly? Perhaps something like the supermarket checkout thingies which give you coupons for competing brands. If you're always microwaving Celeste Pizza for One, maybe your microwave will force you to watch an ad for Stouffer's French Bread Pizza?
And I don't even want to think about what the privacy advocates are going to say about this.
This is on both the d.net and SETI FAQ lists. Such clients transfer small (<300 KB) chunks of data during (say) your wife's e-mail check.
<O
( \
XGNOME vs. KDE: the game!
Will I retire or break 10K?
At least 10% of the blocks in distributed.net's CSC contest were verification duplicates. Given that each user normally gets at least 100 blocks, 10 invalid results might be enough evidence to detect a tampering problem.
<O
( \
XGNOME vs. KDE: the game!
Will I retire or break 10K?
Who wouldn't go see a film they had helped render on their PC!
People who are boycotting MPAA until it drops the DeCSS suit.
<O
( \
XGNOME vs. KDE: the game!
Will I retire or break 10K?
Something like the human genome-mapping seemed very appropriate, wonder why they never started something like SETI@home for that.
Getting all the pieces of the human genome in the right order obviously requires the pieces. There are gigabytes of pieces, and many users have slow (5 KB per second modem) connections. The final data set is estimated to be 1 GB (3 billion base pairs, three base pairs per code word) in size. They're looking into compressibility of certain sequences.
<O
( \
XGNOME vs. KDE: the game!
Will I retire or break 10K?
Of course, you'd run the risk of having your toaster or microwave make you watch an ad before you can open it and remove your food. Perhaps the "premium" appliances would be the ones that allowed you to dismiss these ads, or gave you more control over which tasks ran in the idle processes. The "free" appliances could be the equivalent of today's NadaPC, while the ones that cost money would be more useful and less idiot-proof.
I'm not sure I like the idea of my refrigerator being a neuron on the Borg collective, though. What if the government were to use the spare cycles to simulate fusion reactions for bomb research? I might have an ethical problem with that...
Was that out loud?
The question then hinges on how well the "power saving idle mode" works on most PCs. Anybody got the numbers?
Cuts power consumption by about 10% considering the RAM, the network card, the video card, the audio card, the rest of the chipset, and peripherals such as the monitor, the speakers, the department's printers, etc.
<O
( \
XGNOME vs. KDE: the game!
Will I retire or break 10K?
Needless to say there are some serious security issues here that no doubt won't be properly assessed.
icqqm [ICQ:11952102]
It's cheaper for companies to outsource some projects to another company and have them do the computing however they can than to rent time on a mainframe.
:)
If you design a cool screensaver to go with it, and make it run on Windows, I'm sure you'll get the support of college students everywhere, even if it beams information back to an evil corporation, does tests on nuclear missle aging, DNA analysis, or hacking your friend's box... They won't know the difference. Of course, it'd be nice to check these things for trojans too.
Yes, this is an application for Beowulf clusters as well; for massively parallel problems, it might be worth setting up shop in that business instead. I bet IBM does just that, for one, but I'm sure they'd rather sell you a mainframe, where possible...
---
pb Reply or e-mail; don't vaguely moderate.
pb Reply or e-mail; don't vaguely moderate.
The first distributed computing company to strike up a deal with Tivo, WebTV, and all the other consumer appliances wins.
Just ship the devices already signed up for the distributed client... how would the user know that the process is running in the background? The app can pick-up/send the work packets when the device calls in to sync (or for webtv when they logon).
Now that I think about it... how do we know that they aren't already doing this???
Also in the case of the medical research they can make it a selling point, "Buy Tivo and help cure AIDS."
(1) It seems that most posters fail to appreciate that this form of distributing computing is limited to problems which are essentially "trivially parallelizable". Anyone who has done parallel computations knows that MOST interesting parallel algorithms MUST exchange information with other processors while doing useful computation. Both the bandwidth and the latency over the internet will NEVER come close to matching what can be found inside a high-end parallel machine where the processors are all mounted in a single unit. The performance of any "coupled" problem, be it a traffic simulation or a climate model, will be absolutely dog poor on these widely distributed computer networks.
All said, this is still a very cool concept for SOME projects, like distributed rendering for films, and analysis of vast quantities of data (ala SETI@home). One shouldn't underestimate the marketing value in a distributed rendering project for a film, either! (Who wouldn't go see a film they had helped render on their PC! Especially if one could "preview" the result as it was being rendered.) But I think that this form of distributed computing will tend to be a niche, rander than a general solution for scientists and technologists with parallel computing needs.
Science, like Nature, must also be tamed, with a view turned towards its preservation.
considering the electricity problems that states like california have been having this summer, would giving people an incentive to leave their computer on all the time actually be detrimental to our safety?
imagine if all of the AOL newbies out there caught wind of this and decided that they'd all leave their crappy emachines and imacs online all of the time in 30 million household's worldwide. that's alot of energy that is being used up on our already taxed power grids.
there was actually a great article in the Industry Standard last month that talked about this energy problem. While processing and computerized applicances keep on growing in popularity, the whole internet infrastructure is built upon the assumption that we have a virtually unlimited source of electricity. Truth is that this summer California hit 95% total capacity. They were 2% away from having rolling brownouts. Some electric companies have even started PAYING their large corporate customers to take days off and to shut down their power supplies so that these electric companies can keep electricity flowing to the general public without causing safety risks.
While distributed computing might be a very kewl idea and concept, there comes a point when we need to fix the base limitation of the computerized world and that's the power grid.
The thing is, any revenue from idle time beats what you get if you just let the cpu burn cycles. Hopefully, if this gets to be a successful business model, the price will get bid up; but a box generating a buck a month off idle time is still (a little bit) more profitable than one that isn't.
When all you have is a hammer, everything looks like a skull.
As soon as one company starts giving away money to anyone with spare CPU power, many others companies with similar claims will spring up overnight. Soon there will be so many that people will begin mass mailing viruses or spying programs disguised as programs that pay you. Most people would say that if someone is dumb enough to open attachments in mail from someone thay do not know, they deserve to be spied on and have their important files or even their entire hard drive erased, but this might end up killing the good programs.