Chinese Academy of Sciences Proposes A New Type of BFT Consensus Algorithm
With the eyes of the most of the crypto industry pointed at Ethereum’s upcoming second iteration, Ethereum 2.0, and the staking functionality it will unlock, few have focused on the underlying mechanisms that actually make Ethereum possible. With over 6,000 different cryptocurrencies currently on the market, most of them based on several prominent blockchains, the science that actually makes cryptocurrencies possible is often overlooked.
This, however, isn’t the case in China.
The country’s top computer scientists and mathematicians have long been working on researching and improving blockchain algorithms—partly to use as a basis for China’s upcoming digital currency electronic payment (DCEP) system, and partly to produce a superior algorithm that could potentially be further commercialized.
According to the latest report coming from the Xinhua News Agency, Chinese scientists have actually achieved a major breakthrough in their research of consensus algorithms. The breakthrough was made in the Institute of Software Research, an integral part of the Chinese Academy of Sciences, by software researcher Zhang Zhenfeng and his team of collaborators—Guo Bingyong, Qiang Tang, Xu Jing, and Lu Zhenliang.
The “creative breakthrough,” as it was described by Xinhua, was made during the research of the Byzantine Fault Tolerance (BFT) consensus, the core technology underlying most of the market’s dominant blockchain platforms. And while the results of their extensive research were presented at the 27th International Conference on Computer and Communications Security, which took place in November 2020, it wasn’t until the second week of February that the public actually became aware of the research.
Zhang and his ream proposed the Dumbo Byzantine Fault Tolerance (DumboBFT) algorithm, which improves on the HoneyBadger Byzantine Fault Tolerance (HoneyBadgerBFT). DumboBFT essentially solves the theoretical problem of asynchronous consensus algorithm design and improves the performance of algorithm currently adopted by the industry.
Byzantine fault tolerance (BFT) is a type of protocol that enables a set of untrusted peers to reach consensus. The main categorization of BFT protocols is based on the timing assumptions—while synchronous BFT protocols assume that all values sent by honest peers will be delivered within a certain period of time, asynchronous protocol don’t require any time bounds to exist.
Zhang and his team designed two new asynchronous protocols, both of which improve the running time of asynchronous common subset protocol (ACS) asymptomatically and practically—Dumbo1 and Dumbo2. According to the official proposal paper, Dumbo1 only runs a small number of instances of asynchronous binary agreement protocols, while Dumbo2 further reduces the number of these instances to a constant.
The team did heavy research on the protocols, deploying both of them on 100 different Amazon AWS EC2 instanced dispersed in four continents. Every test showed that the Dumbo protocols outperformed HoneyBaderBFT by several times. However, the team did note that there are still ways to further improve the protocols and added that they may further reduce the communication complexity.