ACSarF: a DRL-based adaptive consortium blockchain sharding framework for supply chain finance

Abstract

Blockchain technologies have been used to facilitate Web 3.0 and FinTech applications. However, conventional blockchain technologies suffer from long transaction delays and low transaction success rates in some Web 3.0 and FinTech applications such as Supply Chain Finance (SCF). Blockchain sharding has been proposed to improve blockchain performance. However, the existing sharding methods either use a static sharding strategy, which lacks the adaptability for the dynamic SCF environment, or are designed for public chains, which are not applicable to consortium blockchain-based SCF. To address these issues, we propose an adaptive consortium blockchain sharding framework named ACSarF, which is based on the deep reinforcement learning algorithm. The proposed framework can improve consortium blockchain sharding to effectively reduce transaction delay and adaptively adjust the sharding and blockout strategies to increase the transaction success rate in a dynamic SCF environment. Furthermore, we propose to use a consistent hash algorithm in the ACSarF framework to ensure transaction load balancing in the adaptive sharding system to further improve the performance of blockchain sharding in dynamic SCF scenarios. To evaluate the proposed framework, we conducted extensive experiments in a typical SCF scenario. The obtained experimental results show that the ACSarF framework achieves a more than 60% improvement in user experience compared to other state-of-the-art blockchain systems.