Alternatives to mining

[mintymark]

After having a run of luck on my miner and calculating that my hardware would be paid off in a month, and that if I brought 10 machines I would have a comfortable income, and then realising none of that was true, I began to wonder what the future holds for "miners" and what income could be made from running a GPU farm in the future. Its true, so far as I can see that so far as income is concerned, bitcoin is in fact the only show in town right now. Seti and protein folding at home are both worthy efforts, but I wonder if for the good of bit coin (as well as for our own interest) if there isnt a gap in the market for an alternative, and yes, of course it would use bitcoins for transactions, thats the point and exactly the type of thing bitcoin is really good for.

So this would be an application that I could run on my hardware. It would be very like a bitcoin miner, it would cause my fans to whirr, and my GPUS to run hot, and once so much calculation had been done, it would credit my bitcoin wallet. But the coins would come from the person who had commissioned the calculation, using bitcoins that already exist, not newly mined ones.

To a "commissioner" they would write code, perhaps using an interface similar to OpenCl, but a special im;ementation, not to run on a GPU but to put the job up for offer, depending on certain parameters, the commisioner woulduse a website to pay for the job, for it to be done fast on many peaples GPU's or slow on just a few, and would get the answer back.  What are the applications: I am sure there are many, but some that I can see are :

a) Cryptographic cracking.
b) Animation
c) Protein and drug-related calculations.

A bit about a) - its clear with a GPU on every desktop and a GPU farm in every high street, that the world will need to up the game cryptographically speaking, yesterdays safe cipers may not be secure tomorrow. But such has always been the case from enigma on.

But generally there will be many applications and there is bound to be a demand for this.

Apart from the normal OpenCl interface, the piece that controls metering, payment and job control needs to be added, but is not necessarily a small deal. We also need to ensure safety. What resources are allowed to be used on the target machine? Some file space may be needed, but file access, no.

Given that one of the key mechanisms for inflating the bitcoin economy is mining, I see this as another, and the value of GPU's and GPU time will directly impact the value of bitcoins as miners make a trade between doing this and mining.

How would we make this happen? Write a spec and commission it? Hope someone just writes it?

Any further thoughts?

[1713929872]

1713929872

[1713929872]

1713929872

[1713929872]

1713929872

[1713929872]

1713929872

[1713929872]

1713929872

[TenthReality]

Not for nothing, but it seems you're pretty much described what BOINC does.  There are pleanty of both public and private boinc projects doing exactly what you describe here.

The biggest issue I've seen as a "for profit" model on this, is many of the companies that would be interested in the service just are unwilling to let their data go out to "anonymous" users, added to the fact that "crypto cracking" can be used for both good and evil purposes and depending on your geographic location, local laws can make that a HUGE no-no for you.

Animation wise, if you go look at the BURP render project you'll see the obvious issue, the only way to partition out the rendering of an animation requires you to send every single model/texture the animation requires.  No company is going to be willing to hand out those assets to "anonymous" people.  It also leads to everyone needs to grab HUGE packages, so then bandwidth starts becoming a giant issue.

[mintymark]

All of these are fair comments.

I need to take a look at BOINC.

So far as letting the data go out to "Mr anonymous" I agree that is going to be a bigge, and the only way round this is to ensure that the code is encrypted in such a way that the owner of the client does not have acces to the data. Is that even possible, I do not know. Certainly if it were me, I would not even want to know what was being computed. The same is true of animation, nobody wants their newly animated film copied at source, and encryption could be the key here. Encryption or dispersion (where the data is so distributed that nobody gets enough of it to be useful) or both.

I believe that use of these assets will be increasingly valuable, and that they would be more valuable again if you could by them by the gallon, on demand, ready packaged and risk free, and ready to go. In fact I'd go further: it seems to me that in the future there will be no other cost effective way to do a super computer task.

You have highlighted some of the risks. The technical challenge here is to make that risk disappear by the use of technology. 

[TenthReality]

Go over to BURP(http://burp.renderfarming.net/) and start a dialog with Janus, he might be able to help a good bit here.  He made a huge attempt to take BURP as a pay to play project and just couldn't make it happen. 

Even with super encryption, the code/engine on the local users machine would need some means to decrypt it to process the work.  The minute that is in place, you're running the risk of someone opening it up.  Some of your examples could have the data-subset which would help a good bit, but others will require the "whole picture" to do any work at all.  Be it a memory scrape, HDD IO watching or various other means, it would not be hard at all to gain access to unencrypted versions of these.

Also, 90% of users who are running a system like this actually want to know what they are doing.  They want the little screensavers showing them proteins folding, or looking at stars etc.  The users also care about HUGE metrics on the various tasks they are doing.  Go checkout some of the BOINC projects like Milkyway@Home, Einstein@Home, Folding@Home. in BURP's example, you can actually see realtime images in the gallery of what you're currently rendering.

I'd love for you to find a means to make it happen, it just seems fairly technically impossible given what you're attempting to do.

I wish you the best of luck here.

Shameless plug here, but those of you with spare CPU cycles, go join Enigma@Home and help read Hitlers mail from WW2....

[Meni Rosenfeld]

This is something I've been thinking about for years, before the advent of GPGPU (or my awareness of it). In fact one of the things that attracted me to Bitcoin is that it allows individuals to sell computation for money (the other is that I've also been thinking about cryptocash for a while). My vision is more or less similar to yours, and I'd love to see it happen one day.

The closest thing I know of is Parabon, but it's fairly primitive - it only does CPU, its software is outdated, it only gives payouts to sufficiently large institutions, etc.

TenthReality's points are all valid but I hope there will be a sufficient market for computation which is either not confidential, or can be sufficiently obfuscated. The ideal is tractable fully homomorphic encryption which will allow the client to send encrypted input, have the server work on it while still encrypted, receive an encrypted output and then decrypt it.

In any case, pulling something like this off will be a huge project which will require resources and people with lots of relevant experience.

[zoro]

i am totally font of this idea to rent your GPU raw power in universities companies or other institutions in order to help in calculation
of their projects!
i believe that this may be the "job" of the future

[Fiyasko]

Nah, The workload will goto the lowest bidder with the faster speed, Creating an unstable income