Lost Write Detection in RAID Storage Redundancy Layers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current storage systems are inadequate in detecting and correcting 'lost writes' in RAID layers, as existing error detection techniques fail to identify corrupted data due to lost writes, leading to potential data corruption and performance degradation during parity checks.
Innovation Solution
A method is introduced where a generation count is assigned to on-disk data blocks during writes and maintained in a bitmap within the parity block, allowing for detection of lost writes by verifying the generation count during reads, enabling the recovery of lost data from parity and remaining blocks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If block-appended checksum is used for error detection, then data corruption due to bit flips and partial writes can be detected, but lost writes cannot be detected because all metadata appears valid
Solution Approach 1:
The patent applies preliminary action by pre-computing and storing checksums of metadata fields (including generation count) before writes occur. This allows subsequent detection of lost writes by comparing stored checksums with current data, enabling identification of writes that were not properly committed to disk.
Solution Approach 2:
The patent implements feedback by incorporating checksum verification into the read operation. When data is read, the system computes checksums of current metadata and compares them with stored checksums. A mismatch indicates a lost write, providing feedback about write failures without requiring additional monitoring infrastructure.
2Reliability
If parity checking is performed to detect corrupted blocks, then data integrity can be verified, but performance degradation occurs due to computational expense
Solution Approach 1:
The patent applies preliminary action by pre-computing and storing checksums of metadata fields before writes occur. This allows subsequent detection of lost writes by comparing stored checksums with current data, enabling identification of writes that were not properly committed to disk.
Solution Approach 2:
The patent implements feedback by incorporating checksum verification into the read operation. When data is read, the system computes checksums of current metadata and compares them with stored checksums. A mismatch indicates a lost write, providing feedback about write failures without requiring additional monitoring infrastructure.
3Reliability
If generation count is maintained in bitmap within parity block, then lost writes can be detected on all reads, but additional storage space is required
Solution Approach 1:
The patent merges the generation count tracking functionality into the existing parity block structure. By combining the bitmap data structure with the parity block, the system leverages existing storage infrastructure to track generation counts, avoiding the need for separate dedicated storage structures.
Solution Approach 2:
The patent makes the parity block serve multiple functions: traditional parity calculation, generation count storage via bitmap, and lost write detection. This multi-functionality reduces the need for additional dedicated storage structures, as the parity block becomes a universal component for both error correction and detection.
Data Source
AI summary
A storage system includes a set of mass storage media and a plurality of functional layers that provide a data storage path between a file system and the set of mass storage media. The storage system can detect that a write issued by the file system was lost by one of the functional layers that provide the data storage path between the file system and the mass storage media.