Data Storage Device Read Scrub Using Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data stored in memory devices is susceptible to corruption due to read disturb effects and data retention failures, leading to potential data loss beyond the capability of error correction codes, and frequent read scrub operations degrade performance and lifespan.
Innovation Solution
A data storage device controller tracks age and read count of memory blocks, uses machine learning to predict error-prone blocks, and performs targeted read scrub operations on these blocks to minimize read amplification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If frequent read scrub operations are performed to prevent data loss, then data reliability is improved, but performance and memory lifespan deteriorate
Solution Approach 1:
The patent applies local quality by performing read scrub operations selectively on specific memory blocks rather than uniformly across all blocks. The controller identifies and targets only those blocks that exhibit error-prone characteristics, applying the read scrub operation locally to these specific areas while leaving other blocks untouched, thus resolving the contradiction between maintaining high data reliability and preserving system performance
Solution Approach 2:
The patent implements partial action by performing read scrub operations on only a subset of memory blocks that are predicted to be error-prone, rather than performing exhaustive read scrub operations on all blocks. This partial approach maintains adequate data protection while reducing the overall performance impact and extending memory lifespan
2Reliability
If frequent read scrub operations are performed to correct errors, then data reliability is improved, but memory lifespan deteriorates
Solution Approach 1:
The patent extends memory lifespan by applying read scrub operations locally to only those memory blocks that are predicted to be error-prone based on tracked characteristics. This selective local approach reduces the cumulative read amplification across the entire memory array, thereby reducing wear and extending the operational lifespan of the memory device while still maintaining data reliability for the targeted blocks
Solution Approach 2:
The patent implements partial action by performing read scrub operations on only a subset of memory blocks rather than all blocks, reducing the total read amplification and wear on the memory device. This partial approach maintains adequate data protection for critical blocks while extending overall memory lifespan
3Productivity
If targeted read scrub is performed on selected blocks, then read amplification is reduced, but the complexity of block selection increases
Solution Approach 1:
The patent applies self-service by enabling memory blocks to effectively identify themselves as error-prone through tracked characteristics such as age and read count. The controller uses these self-reported characteristics to automatically select blocks for read scrub operations without requiring complex external analysis, thus reducing read amplification while keeping the selection mechanism relatively simple
Solution Approach 2:
The patent implements feedback by continuously tracking characteristics of memory blocks (such as age and read count) and using this feedback information to predict which blocks are error-prone and should undergo read scrub operations. This feedback mechanism enables intelligent block selection that reduces read amplification while maintaining manageable system complexity
Data Source
AI summary
A data storage device tracks information about the ages and read counts of blocks of memory, which are indicative of data retention and read disturb issues, respectively. The data storage device uses a machine-learning model to predict which blocks in the memory are likely to contain errors based on the tracked information, and a read scrub operation is performed only on those blocks. Limiting the number of blocks that are read scrubbed reduces read amplification and reduces the time required for the read scrub process.