RoarChain: A Robust Sharding Blockchain System for Enterprise Consortium

Enterprise consortium blockchains are increasingly adopted in domains such as supply chains and finance, where secure, high-throughput data storage and verifiable, highly available query services are essential. While existing solutions like ScorpioBase leverage off-chain databases and fine-grained signatures to improve data throughput, they suffer from a critical limitation: the lack of data replication makes them vulnerable to Byzantine failures, leading to permanent service unavailability when individual storage nodes fail. To address this challenge, we propose RoarChain, a robust and storage-efficient sharding system designed for permissioned blockchain environments. RoarChain introduces a novel replica placement strategy based on Balanced Incomplete Block Designs (BIBDs) to generate structured, permutation-based replica sets, achieving three key objectives: (1) high query availability under anonymous node failures, (2) minimized storage overhead for each participating organization, and (3) integrity-preserving data recovery. Unlike traditional data center techniques that rely on full replication or trusted hardware, RoarChain provides strong fault tolerance without assuming trusted infrastructure or imposing excessive redundancy. We formally prove the correctness of RoarChain's replica scheme using combinatorial theory and analyze its efficiency and robustness. Our system is implemented on a large-scale cluster, and extensive experiments demonstrate that RoarChain significantly improves service availability, reduces storage load, and achieves rapid recovery from node failures, outperforming baseline approaches.