Managed NAND Validity Table Segmentation for Latency Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing NAND memory management techniques face latency issues during garbage collection operations due to the fine granularity of validity tables, which require checking each physical address for data validity, slowing down memory management processes.
Innovation Solution
Implementing a high-level validity table that indicates data validity at a coarser granularity, allowing for the transfer of data without checking each physical address, thereby reducing latency by using a combination of high-level and low-level validity tables to manage data validity during operations like garbage collection, unmap, and rewrite.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a fine-granularity validity table is used to track data validity at each physical address, then data validity tracking precision is improved, but memory management operation speed deteriorates due to requiring individual address checks
Solution Approach 1:
The validity table is segmented into two hierarchical levels: a high-level validity table that tracks validity at a coarse granularity (e.g., per block or per page group) and a fine-grained validity table that tracks validity at the individual physical address level. This segmentation allows the system to use the high-level table for quick initial assessments and only access the fine-grained table when necessary, thus maintaining precision while improving operation speed.
Solution Approach 2:
The patent introduces a hierarchical dimension to the validity table structure, organizing validity information in multiple levels of abstraction. The high-level validity table provides a macro view of data validity across larger memory regions, while the fine-grained table provides micro-level detail. This dimensional organization enables faster validation by allowing operations to be resolved at the high level without always requiring detailed individual address checks.
2Reliability
If individual physical address checks are performed during garbage collection, then data validity accuracy is improved, but operation latency increases
Solution Approach 1:
The high-level validity table is updated and maintained in advance to reflect the validity status of larger memory regions. During garbage collection operations, the system first consults this pre-computed high-level table to identify regions that are likely to contain valid data, thereby avoiding the need to check every individual physical address and reducing operation latency while maintaining accuracy.
Solution Approach 2:
The high-level validity table acts as an intermediary between the garbage collection algorithm and the fine-grained validity information. Instead of directly querying individual address validity, the garbage collection process uses the high-level table as a mediator to quickly identify candidate regions, reducing the number of individual address checks needed and thus lowering latency while preserving reliability.
3Speed
If a high-level validity table with coarser granularity is used, then memory management operation speed is improved by avoiding individual address checks, but data validity tracking precision deteriorates
Solution Approach 1:
The validity tracking system is segmented into two complementary tables: the high-level validity table for fast operations and the fine-grained validity table for precise tracking. The segmentation allows each table to serve its specific purpose optimally, with the high-level table providing speed and the fine-grained table providing precision, and both working together to resolve the contradiction.
Solution Approach 2:
Different parts of the validity tracking system have different qualities optimized for different needs. The high-level validity table has the quality of fast access and low overhead, while the fine-grained validity table has the quality of high precision. The system applies local quality by using each table's strengths in appropriate contexts, thereby achieving both speed and precision overall.
Data Source
AI summary
Methods, systems, and devices for techniques for accessing managed not-AND (NAND) memory are described. An indicator of a first type that indicates whether each physical address in a group of physical addresses stores valid data may be accessed. Indicators of a second type may be used to indicate whether respective physical addresses of the group of physical addresses store valid data. Data stored at the group of physical addresses may be transferred to a different group of physical addresses based on the indicator of the first type. Also, another indicator of the first type that indicates whether each physical address in the different group of physical addresses stores valid data may be updated.