Hashgraph — replacement or supplement?
Hashgraph definitely is about to revolutionize the world of decentralized networking.
At the same time, this revolution is still home-brewed, meaning for now there are no hashgraph main-net running as public, so all the breakthroughs are based on some form of private network. Nethertheless, achieved results, assuming they’ll be implemented (even partially) in public network can dramatically change the world as we know it.
As a technology, hashgraph is great. But when, or if, it goes public other factors will come in action. On the technical side, main potential vulnerability of the technology is the assumption of Byzantine setting — meaning that no more than one third of all masternodes are malicious. For now, hashgraph networks run on private masternodes, all of them trustworthy. And there are doubts that the technology will keep it’s performance if masternodes will be made public.
And on the social side — one can find amazing new blockchains better and faster than Bitcoin, but they far-cry from it in terms of capitalization and public trust. Just because the Bitcoin is first and foremost.
Сompares?
Fundamental difference between blockchain and hashgraph lies in the way nodes achieve the consensus about the state of the network.
Blockchain model is non-deterministic, meaning it can’t achieve 100% agreement of all nodes on the state, this agreement only tends to 100% finalization with growth of main chain, and length of this chain grows linear with production of new blocks (i.e. once ~15 seconds for Ethereum network). Hashgraph is deterministic, meaning each new frame (when next witnesses are formed on all masternodes) will be finalized with 100% warranty, the thing is that hypothetically it may take an eternity. But once it’s done — state is 100% determined. So, blockchain doesn’t guarantee finalization, but goes steadily. And hashgraph finalize its state but without any guaranty of time it’ll take. No worry, if all is fine in terms of Byzantine settings — finalization will be achieved with nearly the speed of light.
The variety of blockchain consensuses is quite instructive. The hashgraph has one, but it is the consensus indeed (again, it is working so far as more than ⅔ of all masternodes are trustworthy). It is named ABFT for Asynchronous Byzantine Fault-Tolerant. The cost of being asynchronous was mentioned above — hypothetically, the finalization of a frame can take infinite time, quite unlikely though. The method of reaching the consensus is gossip-about-gossip, meaning exactly what it is — nodes gossip to each other what they see, once a node see ⅔ of previous witness events, it recursively creates new witness event. Masternodes continuously communicate, ‘gossip’ with each other. Once new witness is created at each of masternodes, witnesses carry out voting on the state they observe, and once the consensus is reached the previous witnesses events become ‘decided frame’. Beautiful mathematics behind it proves the process being dramatically fast and secure — all one needs is proper infrastructure and compliance with Byzantine settings:
And then what?
No doubt, that hashgraph will open a new era in networking. Applications will be freed from transaction latency limitations and seems that decentralized networks catch up with centralized ones in terms of performance, and will actually substitute the Internet in many of it’s today applications. Especially in those where decentralization is the game-changer.
We go in step with this crazy new stream — here at Sfxdx, one of our groups is deeply involved in Fantom project development, which is another implementation of aBFT consensus, and our concern is to build much more efficient, decentralized, open-source and permission-less DAG-based technology.
More about our cooperation — click here
