Non-Volatile Memory Block Failure Prediction via Erase Time Monitoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Flash memory systems fail unevenly due to variations in program/erase cycle life across blocks, leading to premature system failure as wear leveling methods assume uniform block lifespan, resulting in wasted capacity and reduced storage device life.
Innovation Solution
A method and system where a controller monitors and tracks the time necessary to erase each block, identifying when it exceeds a threshold, and uses this data to predict block failure, shifting wear burden to blocks with greater remaining life and reducing wear on those with less, thereby optimizing block management and extending storage device life.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If standard wear leveling is applied to distribute program/erase cycles evenly across blocks, then block usage is balanced, but storage system fails prematurely due to weakest blocks and wastes blocks with more remaining life
Solution Approach 1:
The patent changes the parameter used for wear leveling from simple cycle count to a composite metric that includes erase time duration. By monitoring actual erase time and detecting when it exceeds thresholds, the system identifies blocks with varying durability characteristics and adjusts wear leveling decisions accordingly, preventing premature failure and reducing waste of usable capacity
Solution Approach 2:
The system implements feedback by continuously monitoring erase time for each block and using this information to dynamically adjust block selection for future operations. The controller detects when erase time exceeds thresholds and uses this feedback to modify wear leveling behavior, creating a closed-loop system that adapts to actual block conditions rather than relying on uniform assumptions
2Device complexity
If blocks are managed assuming uniform lifespan, then wear leveling is simple to implement, but blocks with different actual lifespans cause premature system failure
Solution Approach 1:
The patent replaces the simple mechanical counter-based wear leveling with a monitoring system that measures actual erase time. This substitution allows the system to detect variations in block durability that simple counting cannot capture, enabling more reliable block management while maintaining reasonable complexity through threshold-based detection
3Reliability
If spare blocks are consumed to replace failed blocks, then failed blocks are replaced, but system fails when spares are exhausted even though remaining blocks have considerable life
Solution Approach 1:
The system performs preliminary action by monitoring erase time and identifying blocks that are likely to fail before they actually fail. By detecting when erase time exceeds thresholds, the system can proactively manage these blocks through enhanced wear leveling or early retirement, preventing the scenario where spares are exhausted while usable blocks remain
Data Source
AI summary
A method and system are disclosed for improved block erase cycle life prediction and block management in a non-volatile memory. The method includes the storage device tracking information relating to a first erase cycle count at which the block erase time exceeded a predetermined threshold relative to a first erase cycle at which this occurred in other blocks. Blocks having a later relative erase cycle at which the erase time threshold is exceeded are assumed to have a greater erase cycle life than those that need to exceed the erase time threshold at an earlier erase cycle. This information is used to adjust wear leveling in the form of free block selection, garbage collection block selection and other block management processes. Alternatively or in combination, the predicted erase cycle life information is used to adjust program and/or erase parameters such as erase voltage and time.