Introduction

The purpose of this guide is to provide a methodology for running Spark benchmarks on EMR. By following this guide, you will be able to identify the lowest price-performance option for running Spark workloads, considering various variables such as engine type (EMR, OSS), deployment models (EC2, EKS, Serverless), or hardware options (M, C, R, family).

The focus of this guide is on price-performance. Other considerations, such as features, user experience, or compatibility with other services, are out of scope. However, it's essential to evaluate these aspects based on your customers' use cases and needs.