Probabilistic Memory Integrity Scans Under Read-Disturb Stress
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional data integrity scans in memory devices are triggered based on the worst-case scenario of read disturb stress, leading to unnecessary scans that cause host collisions and degrade system quality of service (QoS) due to varying sensitivities of memory blocks to gate voltage stress.
Innovation Solution
Implementing a probabilistic data integrity scheme that manages scan frequency using risk factor estimation, considering initial read window budget and program erase cycles for each memory block, thereby accounting for manufacturing variabilities and reducing unnecessary scans.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data integrity scans are triggered based on worst-case read disturb stress, then data reliability is improved, but host collisions increase and system quality of service deteriorates
Solution Approach 1:
The patent applies local quality by transitioning from a uniform worst-case scan trigger policy applied to all memory blocks to individualized risk factor-based triggering for each memory block. Each block receives a customized scan frequency based on its specific sensitivity characteristics, manufacturing variations, and operational history, thereby optimizing the balance between reliability and productivity locally for each block rather than applying a one-size-fits-all approach.
Solution Approach 2:
The patent implements dynamics by making the scan trigger decision adaptive rather than static. The risk factor for each memory block is dynamically updated based on cumulative read disturb stress, manufacturing variations, and operational conditions. This allows the system to adjust scan frequency in real-time based on actual block state, improving both reliability and productivity compared to fixed worst-case thresholds.
2Reliability
If frequent data integrity scans are performed to ensure data reliability, then reliability is improved, but unnecessary scans increase causing host collisions
Solution Approach 1:
The patent applies partial action by performing data integrity scans only when necessary based on calculated risk factors, rather than uniformly scanning all memory blocks regardless of actual need. Memory blocks with low risk factors skip unnecessary scans, while high-risk blocks receive targeted scanning. This reduces overall scan frequency and host collisions while maintaining adequate integrity checking where actually required.
Solution Approach 2:
The patent implements feedback mechanisms by continuously monitoring read disturb stress, updating risk factor estimates for each memory block, and using this information to dynamically adjust scan triggering decisions. This closed-loop approach ensures scans are performed based on actual conditions rather than predetermined fixed schedules, reducing unnecessary scans and host collisions while maintaining reliability.
3Ease of operation
If uniform scan thresholds are applied to all memory blocks, then ease of operation is improved, but manufacturing variabilities are not accounted for reducing precision
Solution Approach 1:
The patent applies parameter changes by transitioning from a single uniform scan threshold parameter to multiple block-specific risk factor parameters. Each memory block is characterized by its own risk factor that incorporates manufacturing variations, sensitivity measurements, and operational history. This allows the system to account for manufacturing precision differences while maintaining automated management through the standardized risk factor calculation and update process.
Data Source
AI summary
Methods, systems, and apparatuses include determining a read counter for a portion of memory of a memory device satisfies a read threshold. A weighted subportion identifier for the portion of memory is selected in response to the read counter satisfying the threshold. The weighted subportion identifier is selected probabilistically, a probability of selection based on defectivity information for subportions of memory of the portion of memory. A subportion of memory is determined using the weighted subportion identifier. A data validity scan is performed on the subportion of memory.