Storage Device Segmentation for Log-Structured File System Wandering Tree
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Log structured file systems face inefficiencies due to the 'wandering tree problem', where correcting data requires additional write operations and metadata updates, leading to increased latency and operational complexity.
Innovation Solution
A data management method that divides a storage device into two areas: one for storing node address tables and another for data and nodes, using sequential access for data storage and random access for metadata updates, minimizing the need for additional write operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If data correction is performed by adding corrected data to the last portion of the log, then sequential write operations are maintained, but a large amount of metadata must be corrected causing the wandering tree problem
Solution Approach 1:
The patent segments the storage device into two distinct areas: a first area for storing node address tables and a second area for storing data and nodes. This segmentation allows independent management of metadata (node address table) and data, enabling efficient updates without affecting the entire file system structure.
Solution Approach 2:
The node address table acts as an intermediary between the file system and the actual data storage. When data needs correction, only the node address table entry needs to be updated rather than correcting the entire metadata tree, thus resolving the wandering tree problem while maintaining sequential write operations.
2Loss of time
If the whole disk is constructed as one log with sequential write operations, then seek latency and rotational latency are minimized, but additional write operations are necessary for metadata corrections
Solution Approach 1:
By dividing the storage device into separate areas for node address tables and data/nodes, the patent enables efficient random access to metadata while maintaining sequential access for data. This resolves the contradiction by allowing metadata updates without requiring full sequential traversal.
Solution Approach 2:
The patent extracts the node address table from the traditional unified log structure and places it in a separate first area. This extraction allows metadata to be updated independently through direct access, eliminating the need for additional write operations across the entire log while preserving sequential write benefits for data.
3Stability of the object's composition
If corrected data is added to the last portion of the log instead of correcting at the original position, then the original data structure is preserved, but a large amount of related metadata must be updated
Solution Approach 1:
The node address table serves as an intermediary that decouples the data structure from metadata. When data is added to the log, only the corresponding node address table entry needs to be updated, not the entire metadata tree. This reduces the quantity of metadata that must be corrected while preserving data structure stability.
Solution Approach 2:
By extracting metadata management into a separate node address table in the first area, the patent reduces the amount of metadata that needs to be updated. Instead of correcting a large metadata tree, only specific node address table entries corresponding to changed data need modification.
Data Source
AI summary
A data management method of a computing system includes dividing a storage device into a first area and a second area, storing data and nodes related to the data in the second area, and storing a node address table in the first area. The node address table includes node identifiers corresponding to the nodes and physical addresses corresponding to the node identifiers.