EVM-Perf: High-Precision EVM Performance Analysis

Autor: Stefan Tai, Anselm Busse, Jacob Eberhardt
Rok vydání: 2021
Předmět:
Zdroj: IEEE ICBC
Popis: Smart contract execution engines are a central part of transaction processing in blockchains. One of the most widely used execution engines is the Ethereum Virtual Machine (EVM). EVM performance is a key determinant of the overall blockchain system. Being able to gather detailed insights into EVM performance characteristics is essential not only when implementing EVMs, but also when designing and dimensioning blockchain-based information systems. Today, there is no precise and easy-to-use solution to gather such information. To address this issue, we introduce EVM-Perf, a high-precision EVM evaluation, and analysis framework. Our framework allows detailed performance analysis of arbitrary EVM implementations. Unlike previous work, it leverages statical analysis methods to achieve higher levels of accuracy. We provide an open-source implementation of EVM-Perf and use it to conduct extensive performance analysis for different physical machines. It comprises both machines with x86_64 as well as ARMv8 instruction set architecture. We discuss how EVM-Perf supports system architects with machine selection as well as configurations of Ethereum deployments, e.g., determining Gas limits or block intervals.
Databáze: OpenAIRE