Asynchronous Consensus Algorithm Integrating Dynamic Weight Sharding Strategy

Abstract

Blockchain technology has broad application prospects in many fields due to its unique characteristics such as decentralization, traceability, and tamper-proofing, and has become a research hotspot in recent years. As a key component of blockchain technology, the consensus algorithm is one of the important factors affecting blockchain performance. Many consensus algorithms currently used in consortium chains are based on timing assumptions and lack horizontal scalability. That is to say, these algorithms cannot reach an agreement in an asynchronous network where the receiving time of network packets is uncertain, and their efficiency decreases as the number of nodes increases, which hinders the large-scale application of the consortium chain. In order to solve the above problems, this paper proposes the dynamic sharding Dumbo (DS-Dumbo) algorithm, an asynchronous consensus algorithm that integrates dynamic sharding strategies, based on the currently excellent Dumbo Byzantine fault tolerance (DumboBFT) asynchronous consensus algorithm. The main work of this paper revolves around how to fragment and optimize the consensus process. This paper designs a node asynchronous sharding model based on multi-dimensional weights, so that the re-sharding work of each blockchain node can be executed concurrently with the asynchronous consensus algorithm, reducing the conflict between blockchain sharding and asynchronous consensus algorithms. We also designed an intelligent transaction placement strategy, which calculates the relevance score of each transaction for all shards to determine which shard the transaction is processed in order to reduce the number of complex cross-shard transactions. We optimized the execution process of the DumboBFT algorithm, converted its internal synchronous working mode to an asynchronous working mode, and reduced the consumption of consensus work to a certain extent. The experimental evaluation shows that the DS-Dumbo algorithm has higher throughput and lower delay than the DumboBFT algorithm, can increase the throughput with the increase of nodes, and has the ability of horizontal expansion.