Storage System Trace Characterization, Compression, and Synthesis using Machine Learning – An Extended Abstract

Abstract

This study addresses the knowledge gap in request-level storage trace analysis by incorporating workload characterization, compression, and synthesis. The aim is to better understand workload behavior and provide unique workloads for storage system testing under different scenarios. Machine learning techniques like K-means clustering and PCA analysis are employed to understand trace properties and reduce manual workload selection. By generating synthetic workloads, the proposed method facilitates simulation and modeling-based studies of storage systems, especially for emerging technologies like Storage Class Memory (SCM) with limited workload availability.