Ladder: A Convergence-based Structured DAG Blockchain for High Throughput and Low Latency

Recent literature proposes the use of Directed Acyclic Graphs (DAG) to enhance blockchain performance. However, current block-DAG designs face three important limitations when fully utilizing parallel block processing: high computational overhead due to costly block sorting, complex transaction confirmation process, and vulnerability to balance attacks when determining the pivot chain. To this end, we propose Ladder, a structured twin-chain DAG blockchain with a convergence mechanism that efficiently optimizes parallel block processing strategy and enhances overall performance and security. In each round, a designated convergence node generates a lower-chain block, sorting the forked blocks from the upper-chain, reducing computational overhead and simplifying transaction confirmation. To counter potential adversarial disruptions, a dynamic committee is selected to generate special blocks when faulty blocks are detected. We implemented and evaluated Ladder in a distributed network environment against several state-of-the-art methods. Our results show that Ladder achieves a 59.6% increase in throughput and a 20.9% reduction in latency.