I recently came across a thread here on maidsafe.org that got me thinking : Is it possible to have Distributed Computing on the SAFE Network?
The concept is almost exactly similar to BOINC : http://boinc.berkeley.edu/. For example would it be possible for users provide their CPU or GPU’s computational power on the network for others to use or rent out?
This has virtually limitless uses, but here’s a quick example I thought about. It focuses on providing computational power for gamers who have hardware that is not powerful enough to run their desired games at optimal settings.
Example :
The player of the game could then rent out X amount of computing power for Y amount of Safecoins and play the game at its highest settings all via the SAFE Network.
The computing power will be sourced from users on the SAFE Network providing their CPU & GPU’s for use, which make the most impact on games. Each frame buffer for the game that needs to be rendered can be sent via the SAFE Network to the users providing the computing power, rendered, and then sent back via the SAFE Network to the player of the game who has rented out the computing power.
The concept is like a faster, more powerful, decentralized version of this : https://www.gaikai.com/
This process would virtually go like this everytime, for most if not all uses :
Buy X amount of computational power with Safecoins
The renter of the computational power sends data to the users providing their computational power via the SAFE Network for processing .
The users providing their computational power resend the processed data back to the renter, who can now make use of it.
Would something similar to that concept, where users can buy computational power from other users on the SAFE Network with Safecoins, (not the actual game itself) be possible to implement on the SAFE Network?
My opinion is that farmers should do the computing and that the SAFE network should automatically distribute that extra computing power on demand. Actually, I think the same should be done for disk space. I don’t like having my personal computer bogged down with additional burdens. The farmers would get safecoins and I would be able to use disk space and even distributed compute power for free. Of course there has to be a limit above which both disk space and CPU power cost safecoins to run. And the network can automatically and dynamically calculate the level for the free quota.
Think about it like this: end users contribute a lot of value simply by using the system. So end users shouldn’t be forced to have to pay with extra disk space, CPU or bandwidth.
But with this distributed computing concept you’re essentially converting disk space into processing power. When you farm you use processing power and disk space to get your safecoins so why can’t you convert those safecoins back into either disk space or processing power?
The current “reward” model does not account for other resources but storage.
Also there are no mechanisms that allow you to do those things (dispatch work and collect it).
Of course those things can be added, but it requires planning and development resources.
I think this is the kind of thing that @dirvine and the rest of the devs should look at as distributed computing has been mentioned a couple of times and it makes sense given the nature and scale of the SAFE network.
Also wasn’t it mentioned in one of the videos that some nodes would handle processing and others verifying. So it’s not all about space. Would it be such a jump to handle distributed computing?
Don’t remember that part (and not sure if I saw that video), but probably someone from the project can clarify.
AFAIK that’s should still be about IO (processing of I/O requests and verifying of the correctness of data).
Project SAFE already includes plans for use as a compute platform. I don’t know if it is in the docs or not but can check my notes for where I learned this if you can’t find anything. Possibly just in conversation with @dirvine.
But generally, my understanding is the first beta implementation will be focused on storage (even though PoR does already incorporate CPU and bandwidth indirectly) and processing/bandwidth will be solidified in the coming year or two.
At any given time, some nodes will be processing and others verifying however these roles are adaptable and change according to the needs of the network.
