search

MeshachBlue / compushare

A program to be used by GPU bitcoin miners "workers" that will either hash or run "pay-per-iteration" Monte Carlo simulations depending on which pays higher at the time.
MIT License
7 stars 2 forks source link

Question #2

Open githubrho opened 11 years ago

githubrho commented 11 years ago

I'm interested in learning more about this - possibly contributing. My understanding of this so far - the mining pool is the cluster which will run monte carlo calculations - bitcoins are earned in proportion to the amount of computing power contributed to a particular computation, right?

githubrho commented 11 years ago

also - where do the bitcoins which pay the miners for calculating Monte Carlo come from? The miners are sacrificing GPU power that would normally be devoted to mining coins to running MCs for a payout in BTC less than what they would earn from dedicated BTC mining. What is the incentive for doing that? Am I confused?

MeshachBlue commented 11 years ago

Essentially, once this is set up, I would begin spreading the word amongst researchers and people in the medical physics field (in which I will be working). They would greatly benefit from accessing this Monte Carlo computing power.

They would pay in bitcoins in an auction style. Say "willing to pay 0.0000001BC per iteration for a total of 10 000 000 iterations." Users would test a set of iterations to see if the payout is greater than hashing per computation time. If the payout is greater, they would run the MC code until the iterations are complete. And then return to hashing. On 22/05/2013 11:46 AM, "githubrho" notifications@github.com wrote:

also - where do the bitcoins which pay the miners for calculating Monte Carlo come from? The miners are sacrificing GPU power that would normally be devoted to mining coins to running MCs for a payout in BTC less than what they would earn from mining. What is the incentive for doing that? Am I confused?

— Reply to this email directly or view it on GitHubhttps://github.com/MeshachBlue/compushare/issues/2#issuecomment-18252863 .

MeshachBlue commented 11 years ago

Essentially, once this is set up, I would begin spreading the word amongst researchers and people in the medical physics field (in which I will be working). They would greatly benefit from accessing this Monte Carlo computing power.

They would pay in bitcoins in an auction style. Say "willing to pay 0.0000001BC per iteration for a total of 10 000 000 iterations." Users would test a set of iterations to see if the payout is greater than hashing per computation time. If the payout is greater, they would run the MC code until the iterations are complete.

If these are being run on GPUs, with say 1000 parallel processes, at say a rate of 100 iterations per second. That would be 0.01 BC per second for a couple of minutes depending on how many others take up the code.

Depending on supply and demand the price users would be willing to accept would vary. I imagine it would eventually hover around electricity prices. On 22/05/2013 11:46 AM, "githubrho" notifications@github.com wrote:

also - where do the bitcoins which pay the miners for calculating Monte Carlo come from? The miners are sacrificing GPU power that would normally be devoted to mining coins to running MCs for a payout in BTC less than what they would earn from mining. What is the incentive for doing that? Am I confused?

— Reply to this email directly or view it on GitHubhttps://github.com/MeshachBlue/compushare/issues/2#issuecomment-18252863 .

githubrho commented 11 years ago

OK I get it now (I think). So I guess the next question is - will federal funding agencies (NSF, NASA, DOE, NIH, DOD, DARPA, ARPA-E, etc...) allow researchers to covert some of their funds to bitcoins for this purpose?

In any case the bitcoins would have to be escrowed somehow (my understanding is that you can set this up using the bitcoin protocol scripting language) to guarantee that payments are maid only when computing is complete.

You could specify in the bitcoin transaction that the prospective MC miner will not be able to receive/spend the coins unless they generate a key that proves that have completed the MC calculation.

githubrho commented 11 years ago

In my case, my GPUs currently produce 0.1 BTC per day. If a researcher paid me such that I could make more than that in a day, I might accept.

MeshachBlue commented 11 years ago

Yes, the code would need to escrow the payments.

About funding problems in the USA I am unsure. In Australia I do not foresee any issues.

The main problem and limitation would be the open nature required for the code submission. Many researchers would not be willing to make their code open source in order for it to run.

I would need to talk with someone about how to make code submitted to the network by the owner a declaration that it is being placed into the creative commons. On 22/05/2013 1:06 PM, "githubrho" notifications@github.com wrote:

OK I get it now (I think). So I guess the next question is - will federal funding agencies (NSF, NASA, DOE, NIH, DOD, DARPA, ARPA-E, etc...) allow researchers to covert some of their funds to bitcoins for this purpose?

In any case the bitcoins would have to be escrowed somehow (my understanding is that you can set this up using the bitcoin protocol scripting language) to guarantee that payments are maid only when computing is complete.

— Reply to this email directly or view it on GitHubhttps://github.com/MeshachBlue/compushare/issues/2#issuecomment-18254948 .

MeshachBlue commented 11 years ago

Currently though, there are a few areas that I have been alerted to that may be so technically difficult that it may not be reasonable. The decentralised nature of it was one part.

But in my opinion the decentralisation is one of the biggest factors for likelihood of the bitcoin GPU miner community picking it up.

In my case, my GPUs currently produce 0.1 BTC per day. If a researcher paid me such that I could make more than that in a day, I might accept.

— Reply to this email directly or view it on GitHubhttps://github.com/MeshachBlue/compushare/issues/2#issuecomment-18255097 .

githubrho commented 11 years ago

Decentralization issues of this sort have been solved by Open Science Grid (https://www.opensciencegrid.org/). Could probably use a method similar. Or check CoinLabs plan for something like this from 2012 - http://forum.wurmonline.com/index.php?/topic/61990-coinlab-faq-how-to-earn-silver-coins-with-your-graphics-card/

MeshachBlue commented 11 years ago

Awesome. Will check that out. On 22/05/2013 2:09 PM, "githubrho" notifications@github.com wrote:

Decentralization issues of this sort have been solved by Open Science Grid (https://www.opensciencegrid.org/). Could probably use a method similar.

— Reply to this email directly or view it on GitHubhttps://github.com/MeshachBlue/compushare/issues/2#issuecomment-18256476 .

MeshachBlue commented 11 years ago

It looks like a big chunk of the distribution of load part of the network would benefit from their code. But I imagine they still have a single computer run the network. Essentially they would have a centrally run distributive network.

In order to decentralise it we need the nodes themselves to run the system.

Which is difficult.

githubrho commented 11 years ago

funding converted to bitcoin not really a problem.as long as there is a mechanism to pay for computing service using university purchase order - funds can be converted to bitcoin inside the service. However, universities dont pay up-front; only AFTER the product or service is received.

MeshachBlue commented 11 years ago

Universities would pay a very small fee 0.1% upfront for administrative costs. And then the administrative nodes will collect the results off the workers and send the results to the clients once the client directly pays the worker by bitcoin. All automated of course. On 23/05/2013 9:45 AM, "githubrho" notifications@github.com wrote:

funding converted to bitcoin not really a problem.as long as there is a mechanism to pay for computing service using university purchase order - funds can be converted to bitcoin inside the service. However, universities dont pay up-front; only AFTER the product or service is received.

— Reply to this email directly or view it on GitHubhttps://github.com/MeshachBlue/compushare/issues/2#issuecomment-18315385 .

MeshachBlue commented 11 years ago

And you are correct, bitcoin purchasing companies could have an add-on of sorts that allows clients to pay them in fiat and have it transferred to bitcoin into the system. On 23/05/2013 9:45 AM, "githubrho" notifications@github.com wrote:

funding converted to bitcoin not really a problem.as long as there is a mechanism to pay for computing service using university purchase order - funds can be converted to bitcoin inside the service. However, universities dont pay up-front; only AFTER the product or service is received.

— Reply to this email directly or view it on GitHubhttps://github.com/MeshachBlue/compushare/issues/2#issuecomment-18315385 .

githubrho commented 11 years ago

Having looked at the CoinLab approach how is your proposed system substantially different? http://coinlab.com/computing

MeshachBlue commented 11 years ago

The payment is "pay-per-iteration" not "pay for access for an hour".

The network is managed by the community keeping management costs competitive. Community involvement due to the decentralised nature of management is much more likely.

Management fees are distributes to the nodes accordingly. On 23/05/2013 10:28 AM, "githubrho" notifications@github.com wrote:

Having looked at the CoinLab approach how is your proposed system substantially different? http://coinlab.com/computing

— Reply to this email directly or view it on GitHubhttps://github.com/MeshachBlue/compushare/issues/2#issuecomment-18316805 .