A Novel Blockchain Transaction Sharding Scheme for Low Cross-Shard and Adaptive Load Balance

Sharding technology is an effective solution to improve the throughput of blockchain by enabling parallel processing in multiple shards. However, the excessive cross-shard transactions and the imbalanced transaction volume load caused by sharding degrade the growth of transaction throughput. To address these issues, we propose a static sharding algorithm based on account-weighted graph from the perspective of historical transaction data migration. The complex account relationship is abstracted into an account-weighted graph, the community discovery algorithm is used to cluster high-frequency senders and receivers in the same shard, and finally the account-based shards are formed. To further address hot shards caused by imbalanced sharding, we propose a dynamic sharding algorithm based on adaptive load balance to assign real-time transactions to shards. The transactions are normally pulled according to the account-based shards, and the transactions with new accounts are allocated to the smaller transaction shards based on the shards’ random dynamic weights. Through multiple rounds of transactions on chain, the transaction volume load among shards tends to balance adaptively. Finally, we conduct extensive experiments to evaluate our algorithm. The results show that our algorithms outperform the classical mainstream sharding projects in terms of cross-shard transaction proportion, transaction load and transaction throughput.