Performance Benchmarking of DAG-based Blockchain Applications

Abstract

Performance has always been the bottleneck of blockchain technology, making it difficult to meet the efficiency and scalability requirements of real-world applications. Many efforts are being made to improve the performance of blockchain technology. Directed Acyclic Graph (DAG) is one of the principal research interests in this field and has been proven to be able to improve the performance effectively. This new distributed ledger technology is getting more attention and is considered to be the next-generation blockchain. There have already been several well-known DAG-based consensus protocols, such as IOTA, Nano, Byteball, and Hashgraph. To better compare the performance of different blockchain protocols, some performance analysis frameworks have been developed, such as Blockbench, Hyperledger, and DAGBENCH. Among them, DAGBENCH is specially designed for the DAG-based platform and currently supports IOTA, Nano, and Byteball but does not support Hashgraph which is also considered to be one of the most promising DAG projects. To evaluate the performance of Hashgraph, and compare Hashgraph with other well-known DAG projects, in this paper, I extended DAGBENCH, a performance benchmarking tool for DAG protocols, to support Hashgraph protocol. Besides, to enrich the diversity of benchmark metrics, I also extended some new workloads to DAGBENCH. Then I tested some performance metrics of Hashgraph on Amazon EC2 using DAGBENCH and analyzed the test result. Our work will lay the groundwork for other developers when extending other DAG protocols in the future. Our experimental results can help developers evaluate different performance metrics to identify bottlenecks and improve performance.