Log-Structured File System Management Area for Flash Memory Wear
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional log-structured file systems for flash memory require lengthy mount processes to identify valid data and uneven erase operations, leading to reduced usable periods due to excessive wear on certain blocks.
Innovation Solution
A file management method that divides the storage medium into a data area for logs and a management area for storing sequence numbers, allowing for sequential recording and reducing the need to check the entire storage area at startup, thereby shortening startup time and prolonging the usable period by minimizing erase operations in the management information area.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all data in flash memory are read and valid data range is identified in the mount process, then the front log can be found, but the process is lengthy and takes much time at the start of operation
Solution Approach 1:
The patent applies preliminary action by pre-recording management information (sequence numbers) in the data structure before mounting. This allows the system to identify valid data ranges without reading and checking all data during the mount process, thus reducing startup time while maintaining accurate identification of valid logs.
Solution Approach 2:
The patent extracts management information (sequence numbers) from the data structure and stores it separately in the superblock. This extraction allows the mount process to quickly access only the necessary management information rather than scanning all data, significantly reducing the time required to identify valid data ranges.
2Adaptability or versatility
If the position of the index file changes in the storage medium, then the file management structure can be updated, but the contents of the superblock are updated frequently causing uneven erase operations
Solution Approach 1:
The patent segments the file management structure by separating the index file from the superblock. The superblock contains only essential management information (sequence numbers) that changes infrequently, while the index file can change position freely. This segmentation reduces frequent superblock updates and promotes even erase operations across flash memory blocks.
Solution Approach 2:
The patent inverts the conventional approach by storing minimal, stable management information in the superblock rather than comprehensive file management data. This inversion reduces the frequency of superblock updates, thereby extending the usable period of flash memory by distributing erase operations more evenly across all blocks.
3Reliability
If data is continuously added and stored without being overwritten in log-structured method, then wear leveling is achieved, but the mount process requires checking all data to identify valid range
Solution Approach 1:
The patent applies preliminary action by pre-recording sequence numbers in the data structure before mounting. This allows the mount process to quickly identify valid data ranges without checking all data, reducing mount process complexity while maintaining the wear leveling benefits of the log-structured method.
Solution Approach 2:
The patent introduces sequence numbers as an intermediary element between the log data and the mount process. These sequence numbers serve as a mediator that enables quick identification of valid data ranges without requiring the mount process to examine all data, thereby simplifying the mounting operation while preserving wear leveling.
Data Source
AI summary
There is disclosed a file management method in a log-structured file system for storing accesses to files that are sequentially performed by sequentially adding logs. A storage area of a storage medium is previously divided into a data area for storing a log including data of the file and a management area that can store management information related to the log stored in said data area for a plurality of entries. Then, the management information including a first number indicating the order of writing the entry is added and stored in the management area as a new entry.