Distributed/Grid Computing for Profit?

[DrMrLordX]

Hello there. I hope this question isn't terrible out-of-line considering that most, if not all, of you seem to be donating your CPU time for a good cause rather than profiteering. However, I noticed a post in the CPU forum or General Hardware (I can't remember which) in which one poster, upon observing the many computers owned by another poster, observed that said other poster should sell CPU time in bulk to a distributed computing effort. I didn't think much of it at the time considering that I only have this one machine plus a laptop, but I've mulled the idea over recently and come to the conclusion that it might be a good idea to try selling CPU time in bulk if anyone actually pays for it reliably.

I've looked around for obvious buyers and found virtually nothing. Gomez seems to want to pay dialup users $45/month which, while amusing, is hardly worth the effort for a broadband user like me. Then there's Ubero which claims to pay users dividends for work completed on paying projects; sadly, Ubero seems to have been working on the same project (a genetics project for UC Irvine) since 2001 that is a non-profit project. I've attempted contacting them at numerous email addresses without success. Something tell's me that Ubero met a similar fate to Popular Power and other would-be brokers of CPU time that sputtered out years ago, the only difference being that Ubero is somehow still running.

The question is, who, if anyone, would actually pay for a substantial amount of computing power? If they pay, how consistantly do they do it, and how much do they pay?

Sorry if this is redundant or not acceptable within the forum's rules. I browsed the sticky and other posts looking for any sign that this would be unacceptable and found none.

 

[amdxborg]

Hi DrMrLordX, first don't worry, no harm in asking!

Well to be honest, none that I know of. I've got a pc running Ubero for a long time now, but just for the stats and I've also had a look at Gomez a while back, but got no response from them. Maybe somebody else here know's of one, but I wouldn't get my hopes up..

Then again, if you, like most of us like competition and searching for answers and cures, you could always join for the fun of it!

 

[petrusbroder]

I do not see any problem selling my CPU-time - if ...

... greed does not make the fun go away,
... if I can be sure that my CPU-time is not used for calculating something which is against my values (e.g. I do not want to crunch anything for the weapons industry, code cracking, policing personal freedom, analysing bugged phonecalls and such)
... there is some reasonably secure way of eliminating cheaters,
... there is some equtable way of calculating the work done,
... if there is some other profit than the money, e.g. it is of value for humanity (medical projects, Seti and such)

This requires the project to be open, controlable, with checks and balances, with some kind of organisation which represents the crunchers. We crunchers are spread all over the world and it takes some doing to get groups of these highly individualistic persons to get united if something goes wrong ...
For me it is far simpler to crunch for fun, I can quit when I want to, have no obligation for getting a certain amount of work done ...
Just some thoughts - my two cents.

Edit: some of the pesky spelling errors

 

[Silverthorne]

I would love to be able to get paid for all the cpu power I have in my house, it costs me about $150 a month.

 

[imported_Questar]

Here's who sells it. $1.00 per CPU hour.

http://www.network.com/

 

[networkman]

IBM also has some "On Demand" computing solutions as well, but that's really geared toward high-end customers that have a short-term need for intensive computing and don't want to invest in the necessary hardware(and associated depreciation and maintenance.)

On a much smaller scale, I've heard of some college students that have rented time on their faster PCs for rendering projects and such for other students without access to such computer power.

 

[Freewolf]

http://www.gomez.com/
http://peer.gomez.com/

 

[DrMrLordX]

Yeah, I agree that most of the incentive is the fun of figuring out how to crunch the most numbers with the least expensive hardware. Part of the reason I thought of this was that it would be a hell of a lot of fun to try and figure out how to crank out a huge amount of computing power with dedicated machines, something which my pathetic budget just does not allow.

If I got paid enough to offset the cost of housing and running PCs, however, it'd be an entirely different ball game and would take number crunching to an entirely different level.

Competing on an open market with the likes of IBM or Sun would be rather difficult, granted . . . without clientele, there's no money to keep the machines crunching.

 

[The Borg]

I suppose that is why these forums are so good - different ideas. I have been thinking for a while that being payed for my CPU time (and all except the laptop are MY machines - bought for the sole purpose to crunch BOINC). However when I read Petrusbroders list, all I could think about was the problems caused on the SETI site when crunch3r left and caused a 'bit' of a stir. Because those items will (shall I dare to say) never be sorted out to everyones satisfaction, payment will never work.

IMO, you will only get the best when it is a work of passion. Who in their right minds spends the amounts of money we do 'for the numbers and science' if it was not a passion?

 

[DrMrLordX]

I agree with you Borg. It seems clear to me now that simply deploying a bunch of hardware and expecting to be paid to run a bunch of BOINC clients isn't necessarily a viable solution. It hasn't worked yet. From a technical standpoint, it's still intriguing that you could have a bunch of mismatched hardware deployed all over the place and use a distributed client model to get them all working on the same project without necessarily losing any CPU time to overhead traditionally associated with parallelized computing. However, when all that hardware is owned by a bunch of people who have different ideas about how things should be done, have variable usage cycles, and may even be running unstable overclocks, your distributed computing model becomes less economically viable. Paying customers wouldn't really want to entrust their data to a computing grid with highly variable performance levels and no real security. Obviously it's worked very well for not-for-profit research efforts.

However, there are people out there selling CPU cycles as a commodity (IBM and Sun being two), so a market clearly exists. The thought of competing against two computing giants and winning on such a market is a potentially exciting (and economically suicidal) challenge. Everyone has their own passion I suppose.

For now, I'll just crunch some numbers for UC Irvine via Ubero and try to figure out who's been buying all this CPU time from Sun and IBM. Thanks for all of your responses, btw.

 

[George Powell]

The main problem with Gomez is that even if you are selected it only pays out a maximum of $0.72 per day per CPU. That is significantly less than the actual cost of running the program.

So as you don't gain anything I would choose a project that interests you or has some significance for you.