New coin idea to incorporate cosmology into the blockchain

[sednax]

Maybe you have heard of the millennium simulation once, the biggest simulation of the universe ever made. It is of great importance to do large scale simulations in modern cosmology, as new theoretical models in physics can be tested that way. It's here: http://www.mpa-garching.mpg.de/galform/virgo/millennium/

I was thinking about how quickly a massive computational grid can be built using the bitcoin protocol and how to do science with it. You donate computer time and get coins in return. The computational part has to be done by the CPU, and the verification of the block chain can be Scrypt based. ASICs would not exist (if they did, modern cosmology would be really, really happy ).

A block should contain:

- The hash of the last block of course
- All transactions of the block, like other coins
- The result of one computation 'unit'. The blockchain would deal with time ordering of events, but one timeframe in the millennium simulation is already in the 100's of GB. So some kind of parrallellization should be made

Each full set blocks (so all units) would represent one unit of time. The millennium simulation only has 64 (representing almost 14 billion years!). We can do that better!

Anyone else think this is a good idea? I am a decent programmer, but would like to get this idea a bit more concrete before I jump into the C

[OnkelPaul]

Some things to think of:
1 - it must be hard to generate one block, but cheap to verify. I doubt that the simulations in this context can be easily verified.
2 - difficulty must be adjustable to keep block rate mostly constant. Of course, you could adjust the simulated time for each unit, don't know how well that would work.
3 - each block must depend on the previous block, i.e. it must not be possible to perform all or part of the hard work before the previous block is known.

I have a gut feeling that most grid computing efforts don't fit these constraints well.

Onkel Paul

[jerrybusey]

Some things to think of:
1 - it must be hard to generate one block, but cheap to verify. I doubt that the simulations in this context can be easily verified.
2 - difficulty must be adjustable to keep block rate mostly constant. Of course, you could adjust the simulated time for each unit, don't know how well that would work.
3 - each block must depend on the previous block, i.e. it must not be possible to perform all or part of the hard work before the previous block is known.

I have a gut feeling that most grid computing efforts don't fit these constraints well.

Onkel Paul

2 and 3 may be manageable through centralization but then you're out of the bitcoin "family" and into another mess of issues. The need for one way functions- point 1- is where you would really need to work some algorithmic magic to make many types of problems workable if they could be made to work at all.