Group Reliability Rating for Storage Device Failure Prediction

Solution Overview

Problem

Existing methods for predicting disk drive failures focus on individual drives and fail to account for group reliability, leading to unexpected failures in large-scale storage systems due to the lack of predictive signs for majority of failures.

Innovation Solution

A method for assigning and dynamically updating reliability ratings to groups of storage devices based on shared operational characteristics, monitoring failure statistics, and reassigning devices to maintain peer groups with similar reliability, allowing for proactive management of storage configurations and I/O operations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If individual disk drive failure prediction methods are used, then analysis of specific drive behavior is possible, but majority of failures occur without warning signs and cannot be predicted

Engineering Contradiction:

Improvefailure prediction accuracyVSAvoiddata loss prevention

Core Design Contradiction:

Measurement precisionVSReliability

Solution Approach 1:

The patent combines multiple disk drives into peer groups based on similar failure characteristics, merging individual drive data into collective group profiles. This allows the system to predict failures for individual drives by referencing the statistical behavior of their peer group, thereby improving prediction accuracy even when individual drives show no warning signs.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The system performs preliminary classification of disk drives into peer groups based on failure characteristics before failures occur. By pre-establishing these groups and calculating historical failure rates, the system prepares prediction models in advance that can immediately assess risk when a drive shows early signs of degradation or when preventive replacement is needed.

Inventive Principle:

Principle #10Preliminary action

2Quantity of substance

If larger capacity storage devices are used to store more data, then storage capacity increases, but data loss and recovery become more expensive when failures occur

Engineering Contradiction:

Improvestorage capacityVSAvoiddata loss cost

Core Design Contradiction:

Quantity of substanceVSReliability

Solution Approach 1:

The system performs preliminary classification of disk drives into peer groups based on failure characteristics before failures occur. By pre-establishing these groups and calculating historical failure rates, the system prepares prediction models in advance that can immediately assess risk when a drive shows early signs of degradation or when preventive replacement is needed.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system continuously monitors drive performance metrics and failure characteristics, feeding this information back into the peer group classification and failure rate calculations. This feedback loop allows the system to refine its predictions over time, improving accuracy as more data is collected from the storage system's operational history.

Inventive Principle:

Principle #23Feedback

3Quantity of substance

If the number of storage devices is increased to expand storage capacity, then more data can be stored, but the probability of multiple device failures increases

Engineering Contradiction:

Improvenumber of storage devicesVSAvoidsystem failure probability

Core Design Contradiction:

Quantity of substanceVSReliability

Solution Approach 1:

The patent combines multiple disk drives into peer groups based on similar failure characteristics, merging individual drive data into collective group profiles. This allows the system to predict failures for individual drives by referencing the statistical behavior of their peer group, thereby improving prediction accuracy even when individual drives show no warning signs.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The system segments the storage system into multiple peer groups based on failure characteristics, allowing differentiated monitoring and prediction for each group. This segmentation enables targeted preventive maintenance strategies for high-risk groups while avoiding unnecessary interventions in low-risk groups, optimizing resource allocation across the entire storage system.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS7580956B1System and method for rating reliability of storage devices

Publication Date: 2009.08.25 ARCTERA US LLC

AI summary

A method for rating reliability of storage devices is disclosed. A reliability rating for a group of storage devices is assigned to a first rating. The first rating indicates an expected reliability that is the same for each individual one of the storage devices in the group. The expected reliability provides an indication of how reliable the storage devices in the group are expected to be. Information indicating one or more operational characteristics for one or more of the storage devices in the group may be periodically received and analyzed to determine whether the reliability rating for the group of storage devices should be changed. If so then the reliability rating for the group is changed to a different rating, e.g., to indicate either a decrease or an increase in the expected reliability of the storage devices in the group.