Accelerating BFT Database with Transaction Reconstruction

Data stores utilized in modern data-intensive applications are expected to demonstrate rapid read and write capabilities and robust fault tolerance. Byzantine fault-tolerant database (BFT database) can execute transactions concurrently and tolerate arbitrary faults (Byzantine fault). We consider cryptographic and communication processing as performance bottlenecks in the transaction processing of BFT databases. This paper presents a transaction reconstruction method, re-constructing a single transaction from multiple transactions to streamline cryptographic and communication processes. We evaluated the proposed method with Basil (state-of-the-art BFT database) in experiments. In an environment where nodes are geographically centralized, the proposed method demonstrates up to approximately 2.5 times higher throughput and reduces latency by up to about 30% than vanilla Basil. In an environment where nodes are geographically distributed, the proposed method demonstrates up to approximately 50 times higher throughput and reduces latency by up to about 75% than vanilla Basil.