Storage Disk Failure Forecasting via ML Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current solutions for disk drive failure forecasting are not highly reliable, despite significant efforts in the industry and academia.
Innovation Solution
A method and system that utilize a select-gapless dataset to initialize a classification learning model, apply incremental learning for disk failure forecasting, and perform proactive responses based on the forecasts, incorporating machine learning classification and prediction reliability scoring.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If current disk failure forecasting methods are used, then forecasting capability is provided, but reliability of forecasting is not high
Solution Approach 1:
The patent transforms continuous disk health parameters into discrete reliability scores through classification learning. The system divides the continuous health degradation process into discrete levels (e.g., healthy, degraded, critical) with corresponding reliability scores, enabling more reliable forecasting by mapping complex continuous data to interpretable discrete states that better reflect actual disk condition thresholds.
Solution Approach 2:
The patent replaces traditional mechanical threshold-based monitoring with machine learning classification models. Instead of using fixed thresholds to detect failures, the system employs trained classification algorithms that learn optimal decision boundaries from historical data, substituting rigid mechanical rules with adaptive intelligent systems that improve forecasting reliability.
2Ease of operation
If traditional monitoring methods are used, then simple implementation is maintained, but proactive response capability is insufficient
Solution Approach 1:
The system enables proactive maintenance by automatically generating reliability forecasts and triggering maintenance workflows without human intervention. The classification model continuously monitors disk health, predicts failures before they occur, and initiates automated responses such as data replication or maintenance scheduling, allowing the system to serve itself in detecting and responding to potential failures.
Solution Approach 2:
The patent implements preliminary action by forecasting disk failures before they actually occur. The classification model analyzes current disk states and predicts future failure probabilities, enabling maintenance actions to be taken in advance. This preliminary forecasting capability allows the system to prepare replacement disks, replicate data, or schedule maintenance before the actual failure happens.
Data Source
AI summary
A method and system for reliably forecasting storage disk failure. Specifically, the method and system disclosed herein entail predicting whether one or more storage disks may fail within a future time period. Further, the storage disk failure forecasts may rely on machine learning classification coupled with prediction reliability scoring.