University of Sydney scientists have created an optimised version of blockchain they say is safer and substantially faster than current versions of the distributed ledger technology.
The concurrent systems research group's 'Red Belly Blockchain' aims to be highly scalable.
Tests on Amazon Web Services' Elastic Cloud Compute platform indicated Red Belly Blockchain could scale to over 440,000 transactions per second, using 100 compute instances, they said.
In comparison, the Bitcoin blockchain is limited to seven transactions per second, while Visa's does not scale beyond 56,000.
The RBB system also has low, three-second latency, compared to one hour for Bitcoin's blockchain.
Dr Vincent Gramoli, head of the CSRG, told iTnews the RBB scales up to 90 nodes by making sure the next block is unique.
"There is no limit to the number of clients transferring money by sending and receiving transactions," Dr Gramoli said.
"Typically, the system serves requests coming at a rate of 440,000+ of transactions per second from potentially as many clients as possible.
"For example, if one billion clients were each issuing a transaction every hour then our system could handle it as the generated workload would represent 278,000 transactions per second."
The researchers have also devised a consensus protocol for participants in the blockchain with performance improvements and safety features.
A binary Byzantine consensus algorithm is used for the Red Belly Blockchain. Unlike current blockchains for crypto-currencies Bitcoin and Etherium which could fork, or split, into older and newer versions due to software bugs, the RBB doesn't allow for this to happen.
This means it's not possible to double-spend the money in the blockchain, or steal the assets its contains.
Verification of RBB address balances is possible without downloading the full blockchain, meaning participants don't need large amounts of storage.
"[The new multivalued Byzantine consensus algorithm] is asymptotically time optimal and resilience optimal and does not rely on a leader, randomisation or sig- natures," the researchers wrote [pdf].
"It combines a reduction from multivalued to binary consensus that applies a bitmask to an array of proposals and a binary consensus to build this bitmask.
"By spawning binary consensus instances in parallel, our reduction is as far as we know the first to decide a non-predefined value in a sequence of O(1) binary consensus instances."
The researchers say the blockchain bridges the gap between private and public blockchains by allowing the list of participants to be adjusted on the fly.
As a next step, the researchers will develop recommendation systems for automated selection of participants in RBB consensus instances, and wallets for the blockchain for end user transactions and balance checking.
