Abstract
Permissionless blockchains such as Ethereum offer decentralization and security. However, their performance is low compared to permissioned and traditional centralized systems, which may hinder their adoption. Transaction latency is an important performance metric that impacts the use of blockchain systems. That is, there is a need for a systematic evaluation to analyze the factors that may contribute to transaction latency. In this article, we propose a queueing model and conduct extensive simulation experiments to evaluate transaction latency within the Ethereum blockchain, considering all the contributing factors. Our simulation results suggest that transaction fees, block limit, block interval time, transaction arrival rate, and the behavior of the network nodes can all significantly contribute to the latency figures. We find the behavior of the network nodes can significantly impact the waiting time for transactions, especially since nodes are not enforced to utilize their blocks properly in the current implementation of permissionless blockchains. Finally, we suggest improvements to transaction latency by (1) increasing the block limit, (2) reducing the block interval time, and (3) encouraging nodes to fill their blocks with transactions as well as discussing the relevant security issues.