Cyclic Block Selection for Non-Volatile Memory Cluster Failures
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing non-volatile memory systems face issues with cluster failures, where multiple adjacent blocks become defective, leading to repeated errors during write operations due to the selection of blocks from affected areas, which can result in performance degradation and device failure.
Innovation Solution
The memory system logically divides its array into multi-block zones and maintains a free block list, cyclically selecting blocks from zones without valid data to minimize the risk of cluster failures by distributing write operations across different zones, thereby avoiding consecutive selection from potentially faulty areas.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If blocks are selected from the free block list without considering zone distribution, then write operations can be performed quickly, but cluster failures occur leading to repeated errors and performance degradation
Solution Approach 1:
The memory array is divided into multiple zones, and the free block list is organized to reflect this segmentation. When selecting blocks for write operations, the system cyclically chooses from different zones rather than selecting consecutively from the same area, thereby distributing write operations across the memory array and avoiding concentration of stress in single regions.
Solution Approach 2:
The system proactively organizes the free block list into zone-based groups before write operations occur. By pre-distributing available blocks across different zones and implementing a cyclic selection algorithm, the system prevents cluster failures before they can develop, rather than reacting to errors after they occur.
2Ease of operation
If write operations are concentrated in specific block areas, then block selection and management is simplified, but adjacent blocks are more likely to fail together causing cluster failures
Solution Approach 1:
The memory array is divided into multiple zones, and the free block list is organized to reflect this segmentation. When selecting blocks for write operations, the system cyclically chooses from different zones rather than selecting consecutively from the same area, thereby distributing write operations across the memory array and avoiding concentration of stress in single regions.
Solution Approach 2:
Different zones of the memory array are treated as distinct regions with independent management. The cyclic selection algorithm ensures that each zone receives appropriate attention and that wear or failures in one zone do not propagate to adjacent zones, giving each region its own operational characteristics.
3Device complexity
If the memory system does not distribute writes across zones, then block management is simpler, but defective blocks in clusters cause repeated errors and reduce memory yield
Solution Approach 1:
The memory array is divided into multiple zones, and the free block list is organized to reflect this segmentation. When selecting blocks for write operations, the system cyclically chooses from different zones rather than selecting consecutively from the same area, thereby distributing write operations across the memory array and avoiding concentration of stress in single regions.
Solution Approach 2:
The system implements a cyclic selection algorithm that periodically moves through different zones when selecting blocks for write operations. This periodic distribution pattern ensures that no single zone is overused and that wear is evenly distributed, improving overall device reliability and yield.
Data Source
AI summary
In a non-volatile memory system, such as flash memory, when selecting a block for write operation, the system selects blocks from a free block list (FBL). The memory circuits of non-volatile systems often experience cluster failures, where multiple blocks of a physical region are bad. If the free block list is loaded with blocks from a region having a cluster failure, this can result in multiple back to back write errors. To help avoid this situation, the blocks of a memory array are divided into physical zones and, when selecting blocks to replenish the free block list, blocks are chosen cyclically from the zones.