Adaptive Read Disturb Algorithm for 3D NAND Storage
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for managing data corruption due to read disturbance in NAND storage devices often result in frequent garbage collection, leading to a high error rate and reduced quality of service, as they rely on aggressive thresholds that are not adaptive or proactive.
Innovation Solution
An adaptive read counter threshold system that tracks read commands and performs test reads and garbage collection proactively, reducing the number of evaluations and test reads, and focusing on the least significant bits affected by read disturb, thereby minimizing errors and improving SSD performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If an aggressive threshold is used for garbage collection, then the frequency of garbage collection increases, but the error rate in blocks increases and quality of service deteriorates
Solution Approach 1:
The patent implements dynamic threshold adjustment where the read disturb threshold is not fixed but adapts based on the actual read disturb count measured from test reads. The threshold is adjusted upward or downward depending on the observed error rates, allowing the system to optimize between garbage collection frequency and error rate based on real-time conditions
Solution Approach 2:
The system performs test reads to measure actual read disturb effects and uses this feedback to adjust the garbage collection threshold. The error rate from test reads feeds back into the threshold adjustment mechanism, creating a closed-loop control system that continuously optimizes the balance between garbage collection frequency and data integrity
2Adaptability or versatility
If a coarse threshold is used to track read disturb, then the threshold can accommodate large error rates, but the superblock is garbage collected too frequently
Solution Approach 1:
The system performs preliminary test reads before making garbage collection decisions to accurately measure the read disturb effect. This preliminary action provides accurate data about the actual error rates, allowing the system to set appropriate thresholds before executing garbage collection, thereby avoiding both premature and delayed collection
Solution Approach 2:
The patent changes the threshold parameter dynamically based on measured read disturb characteristics. Instead of using a fixed or coarse threshold, the system adjusts the threshold value according to the observed error rates from test reads, enabling fine-grained control over garbage collection timing while adapting to different read patterns and storage conditions
3Reliability
If the number of test reads and evaluations is increased to reduce errors, then the quality of service improves, but the system performance and lifespan are reduced due to excessive operations
Solution Approach 1:
The system performs test reads selectively rather than continuously or excessively. Test reads are conducted at strategically chosen intervals based on the read disturb threshold and actual usage patterns, performing just enough verification to ensure data integrity without over-testing. This partial action approach reduces wear from unnecessary read operations while still catching read disturb errors
Solution Approach 2:
The system uses the normal read operations themselves to gather information about read disturb effects through periodic test reads, rather than requiring separate extensive evaluation procedures. The read disturb counting mechanism leverages the existing read infrastructure to monitor degradation, reducing the need for additional performance-intensive operations
Data Source
AI summary
A storage device includes 3D NAND including layers of multi-level cells. Test reads are performed by reading only LSB pages and reading layers in a repeating pattern of reading two and skipping two. A test read of a block is performed when its read count reaches a threshold. The counter threshold is updated according to errors detected during the test read such that the frequency of test reads increases with increase in errors detected. Counter thresholds according to errors may be specified in a table. The table may be selected as corresponding to a range of PEC values including the current PEC count of the 3D NAND. Each table further specifies a number of errors that will result in garbage collection being performed.