Deduplicated Cache Write-Evict Unit for SSD Lifespan Extension
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional storage architectures face challenges in balancing cost, performance, and capacity, particularly in building tiered infrastructure for Protection Integrated Primary (PIP) storage systems, where SSDs offer high performance but high cost and limited write cycles, and disk storage provides lower performance but lower cost, necessitating an efficient deduplication and compression solution.
Innovation Solution
A deduplicated and compressed storage device system that utilizes a fingerprint index and file index to identify and store cached data efficiently, employing a write-evict unit (WEU) to extend SSD lifespan and maintain a small fingerprint index, with mechanisms for efficient insertion and eviction to optimize cache capacity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If SSD is used as cache medium, then performance (IOPS) is improved, but cost increases and write cycle lifespan decreases
Solution Approach 1:
The patent segments the cache write operations into two distinct paths: deduplicated data (which is identified and excluded from writing to SSD) and unique data (which is written to SSD). This segmentation reduces the write volume on SSD, extending its lifespan while maintaining performance benefits.
Solution Approach 2:
The system performs deduplication identification before writing data to the SSD cache. By calculating fingerprints and checking against the fingerprint index in advance, the system identifies duplicate data blocks that should not be written to SSD, preventing unnecessary write cycles and extending SSD lifespan.
2Speed
If SSD is used as cache medium, then performance (IOPS) is improved, but cost increases
Solution Approach 1:
The patent segments cache data into duplicate and unique portions, writing only unique data to the expensive SSD medium while storing duplicate data references in the fingerprint index. This reduces the amount of data requiring expensive SSD storage, lowering overall system cost while maintaining performance.
Solution Approach 2:
Instead of storing actual duplicate data blocks in the SSD cache, the system stores fingerprint references (copies of identification data) in the fingerprint index. This allows rapid access to duplicate data without consuming expensive SSD capacity, reducing cost while maintaining fast access performance.
3Measurement precision
If fingerprint index size is increased to improve deduplication accuracy, then deduplication effectiveness is improved, but cache capacity is reduced
Solution Approach 1:
The patent implements a two-tiered storage structure where the fingerprint index (requiring high precision and small size) is stored in volatile memory with fast access characteristics, while the actual cache data is stored in the SSD cache. This local optimization allows the fingerprint index to be small and precise without significantly reducing overall cache capacity.
Solution Approach 2:
The system transitions from storing all cache data in a single dimension (SSD) to a two-dimensional structure: a small fingerprint index in volatile memory and the main cache data in SSD. This dimensional separation allows the fingerprint index to be optimized for precision without compromising overall cache capacity.
Data Source
AI summary
A computer-implemented method for caching content in a cache memory device is disclosed. The method starts with receiving, at a cache manager, one or more data chunks to be cached in a cache memory device, where the one or more data chunks are retrieved from a persistent storage disk of a storage system in response to a read request of a region of a file. Then the one or more data chunks of a file extent is compressed using a predetermined compression algorithm, and the file extent is packed into a write-evict unit (WEU) maintained in a random-access memory (RAM) that has been open to store a plurality of file extents. In response to determining that the WEU is full, the cache manager writes the WEU from the RAM into the cache memory device.