Representative, reproducible, and practical benchmarking of file and storage systems

Benchmarks are crucial to assessing performance of file and storage systems; by providing a common measuring stick among differing systems, comparisons can be made, and new techniques deemed better or worse than existing ones. Unfortunately, file and storage systems are currently difficult to benchmark; there is little consensus regarding the workloads that matter and insufficient infrastructure to make it easy to run interesting workloads. This dissertation attempts to simplify the task of file and storage system benchmarking by focusing on three of its important principles---developing an understanding of and creating solutions for representative, reproducible and practical benchmarking state and benchmark workloads.;We first develop an understanding of file-system metadata that comprises much of the file-system state by performing the first large-scale longitudinal study of file system snapshots. We then develop means to recreate representative and reproducible file-system state for benchmarking by building Impressions , a framework to generate statistically accurate file-system images with realistic metadata and content. We develop a system Compressions , that makes it practical to run large, complex benchmarks on storage systems with modest capacities, while also being faster in total runtime if desired. We also develop an understanding towards creating representative, reproducible and practical synthetic benchmark workloads, and describe our first steps in creating "realistic synthetic" benchmarks by building a tool called CodeMRI.