For several years now, kindly volunteers have been contributing their spare computing power to vast, distributed supercomputers, all in the name of worthy causes.
Spare computing cycles have been used for everything from scanning cosmic radio signals for signs of extraterrestrial life through to calculating climate change scenarios. Now, a new project hopes to let people charge for their spare computing cycles.
Traditionally, in a community grid project, one central resource (such as SETI@Home) co-ordinates thousands of volunteer computers in a cloud, relying on people giving up their computing cycles for free. After all, it has been practically impossible for people to charge for them – until now.
Zennet is a distributed supercomputing project that will use blockchain technology to remove the central administrators from the problem. Instead, it will be a completely decentralised system, according to its founder, software engineer Ohad Asor.
Asor calls it a market for computational power, in which many different jobs would be distributed among many different computers, with no middleman taking commission or co-ordinating things.
How Zennet will work
Publishers provide jobs ripe for large-scale distribution, which would typically be the kinds of workloads used in Hadoop, the open source project used for the distributed processing of large data sets (you’ll often see Hadoop and ‘big data’ used in the same context).
Publishers distribute information about those jobs on the network. Software operated by providers – owners of computers with spare computing power – will contact the publishers, and automatically negotiate a price for their part in the job.
Anyone can be a provider, as long as their computer is capable of running the Zennet client. Every computer has spare computational cycles, in between each rest break, cup of coffee and keystroke, which could be used by a provider.
The key concept behind Zennet is that publishers’ jobs are designed to be distributed among hundreds or thousands of providers, each of whom take on a tiny proportion of the job.
“More computers is better, because large equals fast. If you have twice as many computers, you will fold your protein twice as quickly.”
Need for speed
Managing thousands of agreements with as many different providers is no mean feat. That’s where blockchain technology will come in, Asor said.
Publishers will announce their jobs on the blockchain and providers will search the public ledger for those jobs. They negotiate the contract directly, then the publisher commits a specific amount in funds to the blockchain.
The provider starts work, and the publisher controls the work on the provider’s computer. Once the work is done, the funds on the blockchain can be released.
Asor would have liked to have used bitcoin’s blockchain for the system, but it is too slow, he said.
“I have been thinking about the coin algorithm for a long time. It has to be a secure coin, and fast,” he explained. “If you did it over bitcoin, when you ordered a computational job it might take half an hour until it begins. That is unacceptable.”
The potential slowness is because of a payment method that Xennet needs to function effectively as the jobs completed are likely to be small, Providers would be paid off-chain, using the micropayments protocol already outlined in the bitcoin wiki. Setting up that micropayments channel takes a confirmation on the blockchain.
“To initiate a micropayment channel, it is necessary to deposit money in a multisig address, and the other party has to wait for confirmations of this deposit,” says Zennet’s explanatory literature.
Instead, Asor proposes an alternative blockchain and cryptocurrency, called XenCoin, which would be used to pay for computational work. He suggests a fixed amount of coin at the start of the Zennet project, rather than ongoing mining.
As a blockchain-based token, XenCoin still needs a consensus mechanism to ensure that the system stays reliable, and every client believes the same version of the blockchain. Asor is mulling two different options here: delegated proof of stake (DPOS), and some kind of modified proof of work.
“They are both good solutions, as far as I can tell, but we take more time to decide on these things because we really want to come with the most secure thing possible,” Asor said.
He’d better get a move-on, as the presale for the Zennet network was supposed to happen in November. Although Asor says that the lawyers are still working it out. “It’s always in a few days,” he said.
In any case, why would people want XenCoins in exchange for their work? If they are used to pay for large-scale computing resources, and the average desktop PC owner doesn’t have a whole bunch of 3D raytracing to do, then it’s difficult to see why they’d bother spending computational power in return for a token like this.
“The formal answer for legal reasons is, I don’t care how you redeem XenCoins,” said Asor. But what about if an exchange offered to trade them for fiat currency? “That’s the informal answer,” he joked.
There’s still a lot of fundamental work to be done in areas such as consensus algorithms before Zennet goes live. However, Asor says that around a quarter of the coding has been done.
If the concept can be made to fly, then it’ll be an interesting example of how blockchain technology opens up hitherto unexploitable markets.
Quite how the folks relying on volunteers to search the skies or predict future climate chaos will feel about all of that is another thing entirely.
Decentralisation image via Shutterstock.