Performance Improvement Techniques in BFT Consensus Blockchain

Abstract

Recently, many studies have been conducted on Byzantine fault tolerance (BFT) consensus protocol to handle random node failures in the blockchain system. However, most BFT consensus protocols have limited performance scalability because of the complexity of the protocol. Furthermore, in a highly decentralized network (e.g., increasing the number of nodes or distributing nodes globally), the performance limitations of the BFT consensus protocol become more severe. This problem is called blockchain trilemma, that it’s hard for blockchains to achieve optimal levels of all three properties - security, decentralization, and scalability - simultaneously. Some BFT consensus protocols dealing with performance scalability have improved throughput, but latency has not been fully considered. In this paper, we propose two performance improvement techniques - Pipelining and Overlapping techniques - for the BFT consensus protocol that offers a high throughput and low latency while at the same time maintaining high decentralization. The proposed techniques can be applied to multiple BFT consensus protocols. For evaluation, we apply the proposed techniques to AuditChain, a blockchain platform that uses a simple BFT consensus algorithm and is deployed in AWS cloud environment. We analyze and evaluate the effectiveness of each technique through AWS experiments. In a highly decentralized network, the optimization technique increases throughput by 978% and reduces latency by 99%. This experimental evaluation demonstrates that the proposed techniques maintain both throughput and latency even if we strengthen decentralization with more consensus nodes or global scale deployment.