Flash Storage Qualification Using ML Telemetry Comparison

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional storage systems rely on manual engineering for qualifying types of storage systems, which is tedious, inefficient, and time-consuming, and may lead to incorrect identification of storage media characteristics.

Innovation Solution

Utilize machine learning to automate and streamline the telemetry process for qualifying storage media for use in the storage systems.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If manual engineering is used to qualify storage media, then the process can be performed with simple tools, but it is tedious, inefficient, and time-consuming

Engineering Contradiction:

Improvetelemetry process efficiencyVSAvoidtime consumption for qualification

Core Design Contradiction:

ProductivityVSLoss of time

Solution Approach 1:

The patent replaces manual engineering processes with machine learning models that automatically analyze storage media characteristics. The ML model processes telemetry data and identifies deterministic characteristics without human intervention, substituting the mechanical/manual qualification process with an automated computational system.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The storage media qualification process becomes self-service through the ML model, which autonomously evaluates storage devices, identifies their characteristics, and determines their suitability for the storage system without requiring external manual engineering input.

Inventive Principle:

Principle #25Self-service

2Measurement precision

If manual engineering is used to qualify storage media, then the process can be performed with existing expertise, but it may lead to incorrect identification of storage media characteristics

Engineering Contradiction:

Improveidentification accuracyVSAvoidconsistency of characteristic identification

Core Design Contradiction:

Measurement precisionVSReliability

Solution Approach 1:

The ML model incorporates feedback loops where telemetry data from storage media is continuously analyzed, and the model's predictions are refined based on actual performance data. This feedback mechanism improves the accuracy and reliability of storage media characteristic identification over time.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent transforms the qualification process by changing the parameters from subjective manual assessment to objective ML-based analysis of multiple telemetry parameters simultaneously, enabling more precise and consistent identification of storage media characteristics.

Inventive Principle:

Principle #35Parameter changes

3Productivity

If machine learning is used to automate the telemetry process, then efficiency and accuracy improve, but the device complexity increases

Engineering Contradiction:

Improvetelemetry process automationVSAvoidsystem complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The ML model serves multiple functions: it qualifies storage media, identifies characteristics, analyzes telemetry data, and makes suitability determinations. This multi-functionality consolidates what would otherwise require multiple separate systems into a single automated platform, managing complexity through consolidation.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS20250348473A1Qualifying flash storage devices through machine learning

Publication Date: 2025.11.13 PURE STORAGE INC

AI summary

Data associated with a first set of managed flash storage devices of a cloud-based storage system is provided as an input to a machine learning model executed by a processing device that identifies one or more characteristics of the first set of managed flash storage devices from the data. A type of change associated with a second set of managed flash storage devices is determined by the machine learning model based on a comparison of the one or more characteristics of the first set of managed flash storage devices and one or more characteristics of the second set of managed flash storage devices. The type of change associated with the second set of managed flash storage devices is provided to a cloud services provider of the cloud-based storage system.