High-Performance Asynchronous Byzantine Fault Tolerance Consensus Protocol

Abstract

In response to new and innovating blockchain-based systems with Internet of Things (IoT), there is a need for consensus mechanisms that can provide high transaction throughput and security, despite varying network quality. Honeybadger was the first practical, asynchronous Byzantine Fault Tolerance (BFT) consensus protocol, achieving high scalability and robustness without making any timing assumptions regarding the network. To improve the current asynchronous consensus protocols, we designed Asynchronous Byzantine Fault Tolerance (ABFT) consensus protocol through integrating threshold Elliptic Curve Digital Signature Algorithm (ECDSA) signatures and optimization of erasure coding parameters, as well as additional implementation-level optimizations. We implement a prototype of ABFT, and evaluate its performance at scale in a global WAN network and a network affected by asymmetric network degradation. Our results show that ABFT provides considerably higher performance, significantly lower computational overhead, and greater scalability than its predecessors. ABFT can reach up to 38.700 transactions per second in throughput. Furthermore, we empirically show that ABFT is unaffected by asymmetric network degradation within the fault threshold.