We present Turing, a reinvented Proof-of-Stake decentralized blockchain system that offers high throughput, low latency, and scalability to large networks. Experimental results on large wide-area and geo-replicated networks show that Turing achieves a high throughput of more than 4,000 transactions per second and a low latency of around 10 seconds on a 9,600-node network, which is comparable to some existing centralized payment systems.
Three major technical innovations in Turing are introduced in this work. 1) Turing uses a workload sharding scheme to provide throughput that scales out with the network size, while avoiding double-spend attacks. 2) Turing applies Polychain, a directed-acyclic-graph data structure, to organize blocks and ensure a global total order of transactions for supporting higher-level applications, such as smart contracts and DApps. 3) Turing executes the Tempo BFT algorithm to achieve fast consensus. Specifically, in Tempo, we introduce a novel message exchange protocol to reach consensus with reduced communication cost. In addition, a highly safe leader committee is used to expedite the process of BFT consensus and avoid the expensive view changes. While the Tempo protocol is conceptually simple and practically efficient and secure, we provide rigorous mathematical proofs of its safety and liveness properties.
Dr. Yuan Zhou will join University of Illinois at Urbana-Champaign as an Assistant Professor. He is currently an Assistant Professor at Indiana University at Bloomington, an Adjunct Assistant Professor at University of Illinois at Urbana-Champaign, and a Visiting Assistant Professor at Shanghai University of Finance and Economics. Prior to Indiana University, Dr. Zhou obtained his Ph.D. in Computer Science at Carnegie Mellon University and was an Applied Mathematics Instructor at Massachusetts Institute of Technology.