ParallelC-Store: A committee structure-based reliable parallel storage mechanism for permissioned blockchain sharding

Abstract

The storage performance of blockchain suffers from serious limitations due to its employed full-replication strategy, especially in large-scale network services such as Jointcloud computing and big data processing. To address this challenge, some storage partitioning mechanisms integrating Erasure Coding with Byzantine Fault Tolerant (BFT) consensus protocol are developed, like BFT-Store and PartitionChain. Whilst promising, there still exist three major issues impacting system effectiveness, scalability and stability. Firstly, the high computational complexity of coding consumes substantial computing time. Secondly, the signature schemes for verifying the integrity and correctness of encoded data lead to massive transmitted data over the network. Thirdly, each process necessitates the participation of all nodes, causing extended time overhead and interruption of system operation. To optimize the above three aspects, this paper presents a parallel storage partitioning mechanism called ParallelC-Store, where the nodes are divided into $g$ Storage Committees (SCs) based on the existing BFT sharding protocol. Firstly, the $g$ SCs engage in parallel implementation of data encoding and decoding of $g$ distinct original blocks in a synchronous manner. Hence, the computational complexity/throughput per block of encoding and decoding can be decreased/increased by about $g$/$g^2$ and $g^2$/$g^3$ times respectively. Secondly, Merkle Tree and Bloom Filter are employed to generate the verification proof of encoded data, which avoids heavy communication burdens. Thirdly, all processes for different scenarios can be implemented exclusively within a specific SC when a node joins/quits the system or a single crashed node needs repair. The experimental results demonstrate that the proposed mechanism generally outperforms the comparison mechanisms in terms of storage consumption, coding efficiency and system scalability.