Memory Device Health Monitoring With Read-Wear Metrics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing memory sub-systems inaccurately determine remaining device life due to neglecting the wear caused by read operations, particularly in workloads with a high ratio of reads to writes, leading to potential data loss.
Innovation Solution
A memory sub-system calculates remaining device life by incorporating read and write counts, along with bad memory unit counts, using normalized metrics and a scaling factor to account for the wear caused by read operations, providing a more accurate assessment of the device's operational lifespan.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If remaining device life is determined using only write count metrics, then the calculation is simple, but the accuracy deteriorates due to neglecting read operation wear
Solution Approach 1:
The patent segments the device life metrics into distinct components: read count metric, write count metric, and bad memory unit count metric. Each metric is collected and normalized separately, then combined to determine remaining device life. This segmentation allows comprehensive measurement while maintaining manageable complexity through modular metric collection and processing.
2Reliability
If read operations are included in device life calculation, then the accuracy of remaining device life assessment improves, but the device complexity increases due to additional metrics
Solution Approach 1:
The patent merges multiple metrics (read count, write count, bad memory unit count) into a unified remaining device life determination. Each metric is normalized to a common scale and then combined, allowing the system to leverage multiple data sources for improved reliability while presenting a single integrated assessment result to the host system.
Solution Approach 2:
The patent transforms different types of metrics (read operations, write operations, bad units) into normalized parameter values that can be directly compared and combined. By changing the parameters to a common normalized scale, the system can accurately assess device life from multiple metric types without being constrained by their different units and scales.
3Measurement precision
If normalized metrics with scaling factors are used, then the comparison between different metrics becomes accurate, but the calculation complexity increases
Solution Approach 1:
The patent applies parameter changes by normalizing each metric type to a common scale using scaling factors. Read count, write count, and bad memory unit count are each transformed into normalized values that represent their contribution to device wear on a comparable scale. This enables accurate metric comparison and combination while the normalization process systematically handles the complexity of different metric scales.
Data Source
AI summary
A plurality of memory device life metrics are determined, where one of the plurality of memory device life metrics comprises a read count metric that specifies a number of read operations performed on the memory device. A plurality of normalized metric values are calculated, where each of the normalized metric values is based on a ratio of a respective memory device life metric to a respective lifetime target value associated with the respective memory device life metric. A normalized metric value that satisfies a selection criterion is identified from the plurality of normalized metric values. The identified normalized metric value corresponds to an amount of used device life of the memory device. An amount of remaining device life of the memory device is determined based on the identified normalized metric value. An indication of the amount of remaining device life is provided to a host system.