Semiconductor Memory Data Storage Method for Sequential Read Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Semiconductor memory devices face inefficiencies in random read and write operations due to the Flash Translation Layer, which hinders sequential access, impacting reading speed and efficiency, especially when similar data is scattered across memory pages.
Innovation Solution
A data storage method where similar data is stored in adjacent physical memory pages within the same block, utilizing a control circuit to generate input data vectors and determine similarity with stored data, allowing for sequential writing and reading of highly similar data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If data is stored using traditional random write methods in SSDs, then data can be written to any logical address, but sequential read efficiency is lost and reading speed decreases
Solution Approach 1:
The system performs preliminary analysis of incoming data to determine similarity with existing data blocks before writing. By pre-calculating data similarity metrics and predicting future read patterns, the system proactively places data in optimal physical locations that will enable sequential reads, rather than simply writing to the next available block.
Solution Approach 2:
The patent introduces a new dimension of data organization by creating multiple data blocks with different similarity characteristics. Instead of linear sequential storage, data is organized in a multi-dimensional space where blocks are arranged based on data similarity metrics, allowing the system to transition between random and sequential access modes by selecting appropriate blocks.
2Productivity
If similar data is scattered across different memory blocks, then data can be stored continuously, but reading efficiency decreases due to random access requirements
Solution Approach 1:
The system merges similar data blocks together by identifying data with similar characteristics and consolidating them into contiguous physical memory locations. This combining of similar data types enables sequential read operations while maintaining efficient storage utilization, as the merger is based on data similarity rather than arbitrary allocation.
Solution Approach 2:
Different memory blocks are assigned different qualities based on their data similarity characteristics. Blocks containing highly similar data are optimized for sequential reads, while blocks with diverse data maintain random access efficiency. This local optimization allows each block to be tuned for its specific data type's access patterns.
3Ease of operation
If the Flash Translation Layer maps logical addresses to physical addresses randomly, then data management is simplified, but sequential access patterns are lost
Solution Approach 1:
The Flash Translation Layer is made dynamic by allowing it to adapt its mapping strategy based on data characteristics and access patterns. Rather than using a fixed random mapping algorithm, the FTL dynamically adjusts physical address allocation to create sequential access patterns when data similarity indicates that sequential reads are likely, while maintaining random mapping when access patterns suggest otherwise.
Data Source
AI summary
The application discloses a semiconductor memory device and a data storage method. When determining that an input data conforms to a target format, an input data vector is generated based on the input data. When determining that the input data is similar to a stored data in a target block of the memory array, the input data is written to a blank target memory page of the target block of the memory array.