Scaling Byzantine Fault-Tolerant Consensus With Optimized Shading Scheme

This article introduces a novel scalable multishard Byzantine fault tolerance (SharBFT) consensus protocol combined with a blockchain sharding optimization scheme. SharBFT builds upon the classic BFT state-machine replication approach and extends it into a hierarchical multishard prototype to enable scalable and concurrent Byzantine consensus. This prototype enhances scalability and bolsters the security of global consistency in comparison to existing protocols. Moreover, SharBFT employs a novel consensus voting mechanism based on the threshold signature scheme, resulting in linear message communication complexity and optimized consensus operations. In additional, SharBFT integrates a sharding optimization model (SOM) to enhance consensus efficiency in dynamic system environments. The proposed SOM aims to minimize the average consensus latency while ensuring security and scalability. This article presents experimental results conducted in a real-world cloud environment, illustrating significantly improved performance.

I. Introduction

The adoption of the practical Byzantine fault tolerance (PBFT) protocol in blockchain systems, as a viable alternative to the traditional proof of work, has primarily been motivated by the need to enhance transaction throughput and minimize energy consumption [1]. Nevertheless, PBFT encounters a notable performance limitation stemming from its quadratic communication complexity and resource demands [2]. To address this issue, recent PBFT protocol variants have been devised. Notable among them is HotStuff [3], which achieves a linear message complexity of by leveraging threshold signatures [4]. However, it is crucial to note that the performance of these variants experiences a linear decline as the number of replicas increases, especially in the presence of failures [5]. This can be attributed to linear consensus operations that rely on the entire replica set, particularly when encountering a growing number of replicas. Consequently, a significant challenge lies in the limited scalability of recent PBFT variants due to their architecture based on single full set consensus (C1).