Service-Aware Dynamic Sharding Approach for Scalable Blockchain

We propose a service-aware dynamic sharding approach for enhancing the scalability of blockchain platforms. The objective of our proposed approach is to decrease the service latency for user transactions and increase the transaction throughput of conventional blockchain platforms. Sharding is a well-known solution to the problem of blockchain scalability. In sharded blockchains, transactions can be processed in parallel on different shards. However, compared to intra-transactions, cross-shard transactions, which are inevitably generated in sharded blockchains, are confirmed after a relatively long time. This attribute may result in the overall degradation of performance. To mitigate the effects of cross-shard transactions on the performance of blockchain systems, our proposed approach dynamically relocates the data accessed through cross-shard transactions to the same shard. Further, two or more highly related shards are adaptively merged into a single shard, or a single shard is split into smaller disjointed shards to fully utilize shards in parallel. To verify the efficacy of our proposed approach, we implemented a sharded blockchain using the Hyperledger-Fabric platform based on the design concepts of existing sharded blockchains. The experimental results show that the proposed approach can improve the scalability of existing blockchains while maintaining the consistency of ledgers during dynamic state changes.