Point-in-time Snap Copy Management in Deduplication Storage
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing point-in-time snap copy methods in distributed computing systems require redundant read/write copying operations, leading to performance overhead and inefficient cache utilization, especially in deduplication environments where duplicate data is eliminated.
Innovation Solution
A copy-on-write operation is performed by mapping a block to a physical block containing preserved data, eliminating the need for read/write copying by establishing a point-in-time snap copy relationship and using the existing reference pointer to map the new logical block to the original physical block, thereby bypassing redundant data copying and recalculation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional read/write copying is performed for point-in-time snap copy, then data preservation is achieved, but performance overhead increases and cache utilization becomes inefficient
Solution Approach 1:
The patent uses logical copying instead of physical copying. When a snap copy relationship is established, the target logical block is mapped to the same physical block as the source through the deduplication engine's reference pointer mechanism. This allows data preservation without actual data transfer, eliminating performance overhead while maintaining reliability.
Solution Approach 2:
The deduplication engine acts as an intermediary between the storage system and the snap copy operation. It intercepts write operations to the source logical block and automatically updates the target logical block's mapping through reference pointer manipulation, eliminating the need for traditional read/write copy operations while ensuring data consistency.
2Reliability
If read/write copy operation is performed in background mode, then data is preserved at point-in-time, but cache utilization is inefficient due to redundant data transfer
Solution Approach 1:
The patent eliminates physical data copying by using logical block mapping. The target logical block shares the same physical block mapping as the source until a write operation occurs, at which point the deduplication engine handles the update. This approach preserves point-in-time data while completely avoiding redundant cache operations and improving cache utilization efficiency.
Solution Approach 2:
The deduplication engine automatically manages the snap copy relationship and reference pointer updates without requiring background read/write copy processes. When the source data changes, the system self-updates the target mapping through the existing deduplication infrastructure, eliminating wasteful cache transfers and improving overall system efficiency.
3Quantity of substance
If duplicate data is eliminated by deduplication engine, then storage requirements are reduced, but snap copy operations become more complex
Solution Approach 1:
The deduplication engine is enhanced to perform both its original deduplication function and snap copy management. By integrating reference pointer manipulation and logical block mapping capabilities into the existing deduplication infrastructure, the system achieves snap copy functionality without adding separate complex mechanisms, thus reducing overall system complexity despite the multi-functional requirements.
4Quantity of substance
If reference pointer is maintained for deleted copies, then storage space is freed, but access to preserved data requires additional mapping operations
Solution Approach 1:
The deduplication engine serves as an intermediary that manages reference pointers for both deduplication and snap copy relationships. When a snap copy is created, the engine establishes a reference from the target logical block to the source physical block, maintaining storage efficiency while providing direct access paths. This integrated approach avoids additional mapping complexity by using the existing reference pointer infrastructure for dual purposes.
Data Source
AI summary
In accordance with one embodiment of the present description, a logical file unit containing a set of data stored in cache or other memory and mapped by a reference pointer to a physical file unit in a storage, may be overwritten with new data in the memory without first read/write copying the data in memory to a new logical file unit in the file system. Instead, the original physical file unit of data in the storage is preserved and the original reference pointer for the original physical file unit of data may be used to map a new logical file unit in the file system to the original physical file unit storing the original set of data. Other aspects are described.