AI-Based Virtual Testing of Data Storage Workloads and Components
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Ongoing Reliability Testing (ORT) for Data Storage Devices (DSDs) is time-consuming and may not accurately reflect actual field workloads, leading to delayed detection of issues and increased costs, and identifying the cause of issues in complex DSDs is challenging due to numerous components and interactions.
Innovation Solution
Implementing virtual testing models using Artificial Intelligence (AI) and Machine Learning (ML) to simulate DSD workloads, creating embeddings that represent DSDs or their components, and using these embeddings in virtual testing models to predict performance and identify potential issues before shipment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If Ongoing Reliability Testing (ORT) is performed on a sample group of DSDs, then reliability issues can be detected, but the detection time is delayed and many DSDs may already be shipped before issues are identified
Solution Approach 1:
The patent applies preliminary action by performing virtual testing of DSDs before they are manufactured or shipped. Virtual testing models simulate real-world workloads and environmental conditions in advance, allowing reliability issues to be identified prior to physical production. This enables proactive detection of potential failures rather than reactive detection after issues manifest in the field.
Solution Approach 2:
The patent uses virtual testing models that create digital copies or representations of physical DSDs. These virtual models replicate the behavior, components, and operational characteristics of actual DSDs, allowing reliability testing without physical hardware. The virtual copies can be tested extensively under various conditions without consuming physical resources, enabling rapid and comprehensive reliability assessment.
2Reliability
If physical Ongoing Reliability Testing is conducted over several months, then reliability can be assessed, but the testing process is time-consuming and costly
Solution Approach 1:
The patent replaces physical mechanical testing systems with virtual simulation models. Instead of physically operating DSDs under various conditions for extended periods, the system uses software-based virtual testing models that simulate workloads, environmental stressors, and operational scenarios. This substitution eliminates the need for physical hardware consumption and extended real-time testing, dramatically increasing testing throughput.
Solution Approach 2:
The patent changes the testing parameters from physical time-based measurement to simulated computational evaluation. Rather than measuring reliability through extended physical operation time, the system uses virtual models to evaluate multiple DSDs simultaneously under varied parameters such as workload intensity, environmental conditions, and operational patterns. This allows parallel testing of numerous DSDs without proportionally increasing time requirements.
3Quantity of substance
If a small sample group of DSDs is tested, then testing resources are conserved, but the testing may not accurately reflect actual workloads experienced by DSD operators
Solution Approach 1:
The patent creates virtual testing models that serve multiple functions simultaneously. A single virtual model can simulate various workload types, environmental conditions, and operational scenarios that would require testing multiple physical DSDs. The virtual models are designed to be universally applicable across different DSD configurations and operational contexts, allowing comprehensive workload representation without proportionally increasing the number of physical units needed.
Solution Approach 2:
The patent transitions from one-dimensional physical testing (testing a small number of actual DSDs) to multi-dimensional virtual testing. Virtual models enable testing across multiple dimensions including different workloads, environmental conditions, and operational patterns simultaneously. This dimensional expansion allows comprehensive reliability assessment of numerous DSD variants without physically testing each one, effectively increasing measurement precision while maintaining resource efficiency.
4Reliability
If complex electronic devices with many components are tested, then comprehensive reliability can be assessed, but identifying the cause of problems becomes difficult and time-consuming
Solution Approach 1:
The patent segments the complex DSD system into distinct virtual components and their interactions within the testing model. By dividing the complex electronic device into separable virtual representations of individual components, the system can isolate and analyze specific elements that may be causing reliability issues. This segmentation allows systematic examination of component interactions without overwhelming complexity, making root cause analysis more manageable and time-efficient.
Data Source
AI summary
A Data Storage Device (DSD) is represented by an embedding for virtual testing. A plurality of metrics is obtained for components of the DSD and is encoded into the embedding using a first Artificial Intelligence (AI) model. The embedding is provided as an input to a second AI model configured for virtual testing based on one or more simulated DSD workloads. In one aspect, virtual testing is performed for at least one DSD component by obtaining at least one metric for the at least one component for encoding into an embedding representing a virtual DSD including the at least one component. In another aspect, the embedding is provided to a DSD operator for virtual testing. In yet another aspect, the second AI model simulates at least one workload of the DSD operator or for one or more quality assurance tests of a DSD manufacturer.