Abstract
Discrete-event simulation is an important tool for the analysis of permissionless blockchain systems. Simulation allows one to evaluate blockchains for scenarios that cannot be mimicked easily on the life system. To simulate blockchain systems as close to reality as possible a realistic characterization of the probability distributions of various variables is required. In this paper, we obtain frequency distributions for Ethereum smart contract transactions with respect to gas limit, used gas, gas price, and CPU time. The distribution fitting utilizes Gaussian mixture models, correlation analysis, and ensemble regression techniques, and we demonstrate the appropriateness of each of these methods. We use publicly available Ethereum smart contract data, supplemented with experimental data for over 300,000 smart contracts obtained on a test bed. In addition, we extend the BlockSim simulation tool to support the required distributions and use the simulator for a case study in optimizing miner returns through transaction selection strategies in Ethereum under future scenarios. The outcomes not only demonstrate the impact of selection strategies on miner return but also demonstrate the sensitivity of the outcomes to chosen distributions and therefore illustrate the importance of using realistic distributions.