Nonvolatile Memory Wear Management via Retention-Aware Block Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current memory systems, such as solid state drives (SSDs) with NAND flash memory, face challenges in improving reliability due to wear and tear on memory blocks, leading to varying data retention terms and inefficient allocation of blocks for different data retention requirements.
Innovation Solution
A memory system with a controller that manages wear by classifying blocks based on their data retention terms, estimated from erase count and erasing time, and selects appropriate blocks for writing data, using a wear reduction program mode to extend the life of memory cells by reducing electrical stress.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If blocks are allocated without considering wear levels, then write operations can be performed quickly, but data retention reliability deteriorates due to varying wear levels across blocks
Solution Approach 1:
The patent segments blocks into different wear level groups (first wear level group with lower wear, second wear level group with higher wear) and allocates data to appropriate groups based on retention requirements. This segmentation enables reliable data retention by matching data importance with block wear levels while maintaining manageable allocation complexity through systematic grouping.
Solution Approach 2:
The patent applies local quality by assigning different allocation strategies to different wear level groups. High-retention data is allocated to blocks in the first wear level group with lower wear, while less critical data goes to the second wear level group with higher wear. This localized quality differentiation optimizes overall system reliability without requiring complex individual block evaluation.
2Duration of action of stationary object
If wear reduction program mode is used, then memory cell lifespan is extended, but write operation time increases
Solution Approach 1:
The patent applies wear reduction program mode selectively rather than universally. Normal program mode is used for blocks in the second wear level group (higher wear) where wear reduction is less critical, while wear reduction mode is applied to blocks in the first wear level group (lower wear) where extending lifespan provides greater benefit. This partial application balances write speed and memory cell lifespan extension.
Solution Approach 2:
The patent changes the program operation parameters dynamically based on block wear level. For blocks with lower wear, program parameters are adjusted to reduce stress and extend lifespan. For blocks with higher wear, standard program parameters are used to maintain write speed. This parameter adaptation resolves the contradiction between write speed and lifespan extension.
3Quantity of substance
If blocks with higher wear are used for data writing, then available block pool increases, but data retention term decreases
Solution Approach 1:
The patent segments the block pool into wear level groups and implements differentiated allocation strategies. Blocks in the first wear level group (lower wear) are reserved for high-retention data, while blocks in the second wear level group (higher wear) are used for less critical data. This segmentation ensures that the available block pool is maximized while maintaining appropriate data retention terms for different data types.
Solution Approach 2:
The patent applies different quality standards to different block groups. High-quality blocks (lower wear) are allocated to high-retention data requirements, while lower-quality blocks (higher wear) are allocated to data with less stringent retention requirements. This local quality matching increases the effective usable block pool while maintaining data retention integrity.
Data Source
AI summary
According to one embodiment, a memory system manages wear of each of a plurality of blocks in a nonvolatile memory. The memory system receives, from a host, a write request including a parameter specifying a data retention term required for first data to be written. The memory system selects, from the blocks, a first block in which a data retention term estimated from the wear of the first block is longer than or equal to the specified data retention term. The memory system writes the first data to the first block.