Storage Failure Prediction via Duty Cycle Normalization

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for predicting storage device failures are inadequate, relying on internal metrics like SMART data and being typically inaccurate.

Innovation Solution

A method that involves obtaining and normalizing input and output data from storage devices to determine their duty cycle, and using this data along with pre-established failure data based on hardware characteristics to predict the likelihood of failure, associating a high, medium, or low failure indicator with the devices.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If existing methods use internal metrics like SMART data for failure prediction, then the prediction process is simple, but the accuracy of failure prediction is low

Engineering Contradiction:
Improvefailure prediction accuracyVSAvoidprediction system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent combines multiple data sources including internal SMART metrics with external workload characteristics and hardware configuration data to create a comprehensive failure prediction model. This merging of diverse data types enables more accurate predictions while justifying the increased system complexity through improved reliability

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The prediction system is designed to handle multiple types of storage devices and workloads universally. It processes various hardware characteristics, RAID configurations, and workload patterns through a unified analysis framework, making the system adaptable across different storage scenarios while maintaining prediction accuracy

Inventive Principle:
Principle #6Universality (Multi-functionality)

2Measurement precision

If the prediction method considers workload and device characteristics, then the failure prediction accuracy improves, but the data processing complexity increases

Engineering Contradiction:
Improvefailure prediction accuracyVSAvoiddata processing difficulty
Core Design Contradiction:
Measurement precisionVSDifficulty of detecting and measuring

Solution Approach 1:

The patent segments the failure prediction analysis into distinct components: workload characterization, hardware configuration analysis, SMART metric evaluation, and synthesis of failure likelihood. This segmentation makes the complex data processing manageable by breaking it into smaller, more tractable analytical tasks that can be performed systematically

Inventive Principle:
Principle #1Segmentation

3Measurement precision

If pre-established failure data based on hardware characteristics is obtained and used, then the prediction accuracy improves, but the time and resources required for data collection increase

Engineering Contradiction:
Improvefailure prediction accuracyVSAvoiddata collection time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent implements preliminary action by pre-collecting and storing hardware characteristic data, RAID configuration information, and historical workload patterns before they are needed for prediction. This advance preparation of reference data reduces the time required during actual failure prediction events, as the foundational data is already available for immediate analysis

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentUS10459815B2Method and system for predicting storage device failures
Publication Date: 2019.10.29 DELL PROD LP
  • US10459815B2 patent drawing
  • US10459815B2 patent drawing
  • US10459815B2 patent drawing

AI summary

The invention pertains to a method and information handling system (IHS) for predicting storage device failures. An embodiment of the method of the present invention begins by obtaining data regarding input and output of the at least one storage device. Next, the obtained data is normalized to determine a duty cycle of the at least one storage device. Finally, using the normalized data, a likelihood of failure of the at least one storage device is determined. In another example embodiment, a failure indicator is associated with the at least one storage device based upon the determined likelihood of failure.