PuffChain: A Dynamic Scaling Blockchain System with Optimal Effective Throughput

Despite its promising potential, scalability issues have hindered the widespread adoption of blockchain applications. Existing blockchain systems incur high communication and computation costs, resulting in poor throughput. Furthermore, they do not provide flexible scalability to match the dynamic changes in user transaction demand. In this paper, we present PuffChain, a dynamic scaling blockchain system that can dynamically scale up the transaction throughput when there is a surge of user demand while automatically scaling down upon the drop of user demand. We decouple the functionalities of blockchain nodes into three roles: packers that pack transactions into blocks, proposers that propose blocks for consensus, and validators that validate blocks. The number of packers is dynamically adjusted according to user transaction demand, achieving dynamic scalability. Proposers and validators use a three-phase consensus protocol with filters to maximize appending valid transactions while discarding invalid ones. By decoupling these functions, PuffChain enables adaptive scaling and optimized consensus efficiency. We also develop a theoretical model to analyze PuffChain's throughput and derive the optimal parameters to achieve maximal effective throughput. Extensive experiments on Amazon EC2 clusters demonstrate that PuffChain achieves a throughput of 6061 transactions per second with 100 nodes, highlighting its potential for real-world applications.