Disk Fault Tolerance via Dynamic Node Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current disk fault tolerance methods, such as RAID 1 and RAID 5, suffer from low disk utilization efficiency, poor writing performance, high implementation complexity, and inability to handle simultaneous system failures or power outages, limiting data availability and read-write performance in large-scale distributed storage systems.
Innovation Solution
A disk fault tolerance method and system that employs a pairing storage technology with a dual machine duplexing mechanism, where multiple independent machines are backed up from the node level to the disk level, ensuring data redundancy and reliability by maintaining identical data across corresponding disks in different nodes, allowing for seamless failover and concurrent access operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If RAID 1 uses two disks for backup with master-slave configuration, then data availability is improved, but disk utilization efficiency deteriorates to 50%
Solution Approach 1:
The patent implements dynamic node selection where the system can switch between different nodes based on real-time availability. Instead of a static master-slave configuration, any node in the cluster can serve as the active node, and the system dynamically selects which node to access based on current operational status, thereby improving disk utilization while maintaining data availability.
Solution Approach 2:
Each node in the cluster is designed to be universal and multi-functional, capable of serving both as primary storage and backup storage. Any node can assume the role of the active node, eliminating the need for dedicated master-slave pairs. This multi-functionality allows all disks to be actively utilized for data storage while maintaining redundancy, thus improving overall disk utilization efficiency.
2Productivity
If RAID 5 distributes data and parity chunks across multiple disks, then disk space efficiency is improved, but writing performance deteriorates due to sequential access requirements
Solution Approach 1:
The patent segments data into multiple chunks and distributes them across different nodes in the cluster. Each node stores specific data chunks independently, allowing parallel access during read and write operations. This segmentation eliminates the sequential access bottleneck of traditional RAID 5 while maintaining disk space efficiency through distributed storage.
Solution Approach 2:
The system enables continuous parallel access to multiple disks simultaneously for both reading and writing operations. By distributing data chunks across multiple independent nodes, the system can perform multiple I/O operations concurrently without the sequential dependency that plagues traditional RAID 5, thereby maintaining high writing performance while achieving efficient disk space utilization.
3Reliability
If traditional RAID systems are implemented, then fault tolerance is improved, but implementation complexity deteriorates due to RAID controller and metadata management
Solution Approach 1:
The patent extracts the complex RAID controller and metadata management functions from the storage system, replacing them with a simplified distributed node architecture. Each node operates independently with its own storage management, eliminating the need for centralized RAID controllers and complex metadata structures. This extraction reduces implementation complexity while maintaining fault tolerance through distributed redundancy.
4Ease of operation
If RAID 1 uses master-slave disk configuration, then ease of operation is improved, but productivity deteriorates due to single-node service limitation
Solution Approach 1:
The system implements dynamic node selection where any node can be accessed for read-write operations based on real-time availability. Instead of being limited to a single master node, the system can dynamically switch between multiple nodes, allowing concurrent access and operations across the cluster. This dynamic approach maintains operational simplicity while significantly improving read-write performance through parallel processing.
Data Source
AI summary
A method for data access is disclosed. The method includes: receiving a file access request; acquiring one or more node lists corresponding to the file, wherein each node list comprises at least two nodes, and the mutually corresponding disks between the respective nodes store the same contents; accessing data chunks included in the file from the respective nodes selected from each of the node lists; and when an accessing from a certain node fails, selecting another node from the node list that comprises the certain node, and accessing a respective data chunk included in the file from the selected node according to the identifier of the file. A system and a device with tolerance of disk fault are also disclosed.