How Predictive Algorithms Are Making Data Center Disk Scrubbing Smarter

This article introduces a predictive framework that optimizes data-center disk scrubbing. Instead of treating drives as simply “healthy” or “failing,” a Mondrian Conformal Prediction model assigns each disk a health confidence score to guide targeted maintenance. Combined with a workload predictor using a Probabilistically Weighted Fuzzy Time Series (PWFTS) algorithm, it determines the best time to perform scrubbing when system load is low. The result: reduced downtime, improved efficiency, and lower carbon emissions in large-scale storage systems.