BDLedger: A Scalable Distributed Ledger for Large-Scale Data Recording

Gang Huang, Kaidong Wu, Chaoran Luo, Su Zhang, Huaqian Cai, Xiang Jing, Yun Ma

Abstract: Nowadays, blockchain becomes a hot topic in academics and industry. Traditionally, due to the “Impossible Triangle”, most existing blockchain systems prefer decentralization and security to scalability, which causes their limited throughput. However, it is undeniable that there exist scenarios that require extremely high scalability of blockchain, such as the on-chain record of vehicle driving data. In this paper, we propose a new blockchain-like distributed ledger called BDLedger that can realize an approximately linear increase of throughput to solve the above challenge. In BDLedger, we design a novel consensus mechanism called Random Witness Consensus and organize blocks as a DAG structure. To achieve the linear increase of throughput, we make sacrifices in two aspects: 1) abandoning the consensus on the order of transactions and keep only the consensus on the content of transactions, which means that there are no conflict transactions so that all transactions can be processed in parallel. 2) The content of each block only exists in a constant number of random nodes, which ensures that the overall storage and bandwidth costs will not increase rapidly as the number of nodes increase. We discuss the rationality of the above approach and introduce the design of BDLedger in detail. Experimental results show that BDLedger can achieve an approximately linear increase of throughput, from about 20000 TPS for 10 nodes to about 160000 TPS for 100 nodes.