Explore groundbreaking research on disk failure prediction in this 27-minute conference talk from USENIX FAST '20. Delve into a comprehensive study covering 380,000 hard drives across 64 data center sites over a two-month period. Learn about advanced machine learning models that achieve impressive 0.95 F-measure and 0.95 Matthews correlation coefficient for 10-day failure prediction horizons. Gain insights into the challenges of accurate disk failure forecasting and discover how this research contributes to improving data center reliability. Understand the significance of this work in addressing one of the most commonly replaced hardware components in modern computing infrastructure.
Overview
Syllabus
FAST '20 - Making Disk Failure Predictions SMARTer!
Taught by
USENIX