Performance Evaluation of Transaction Verification in Blockchain Using GPU

Blockchain, a distributed ledger consisting of a P2P network with a large number of nodes, is expected to be used in a variety of applications. However, it suffers from the problem of low transaction creation throughput, because all the participating nodes verify the transactions. To solve this problem, various protocol improvements have been proposed, but they have not enhanced the performance of the nodes themselves. Therefore, when protocol improvements increase the transaction throughput, the transaction verification process at each node can become a bottleneck. In this study, we propose an acceleration method using a graphics processing unit (GPU), which exhibits high parallel computation performance, for blockchain transaction verification. In the proposed method, the verification processes are divided into three main parts: searching past transactions, hashing, and signature verification. It is a CPU-GPU heterogeneous method that takes into account the characteristics of each process and assigns the appropriate process to the CPU and the GPU, respectively. As a result, the proposed method achieves a transaction verification throughput that is 14.0 times higher than that obtained with CPU processing even when the batch size is approximately equal to the current Bitcoin block size, and 46.0 times higher than that obtained with CPU processing when the batch size is approximately four times the current size.