SSD Storage Management via Data Pool Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Solid State Drives (SSDs) experience performance reduction and shortened lifespan due to write amplification caused by garbage collection, where valid data pages need to be copied before erasure, leading to excessive data writing and wear.
Innovation Solution
Classifying data into categories (hot, warm, cold) and storing similar data in the same physical blocks, reducing the number of valid pages to be migrated during garbage collection, thereby minimizing write amplification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If garbage collection is performed by copying valid pages to idle locations before erasure, then the SSD can maintain write capability, but write amplification increases and performance degrades
Solution Approach 1:
The patent segments the SSD storage space into multiple pools based on data access patterns (hot data pool, warm data pool, cold data pool). Each pool is further divided into physical blocks that are managed independently. This segmentation allows garbage collection to be performed on a per-pool basis, reducing the scope of valid page copying required and thereby reducing write amplification while maintaining write capability.
Solution Approach 2:
The patent applies different management strategies to different data pools based on their access characteristics. Hot data pools use dynamic block management with smaller block sizes and more frequent garbage collection, while cold data pools use larger block sizes and less frequent garbage collection. This localized quality approach optimizes write performance for each data type while ensuring overall system reliability.
2Productivity
If physical blocks with large quantities of invalid data are selected for erasure, then garbage collection efficiency improves, but the risk of erasing blocks with many valid pages increases
Solution Approach 1:
The patent implements a feedback mechanism where the garbage collection controller continuously monitors the validity ratio of each physical block and adjusts block selection strategy accordingly. When a block's invalid data ratio exceeds a threshold, it is added to a candidate list for erasure. The system feedbacks on actual erasure results to refine future selections, ensuring high garbage collection efficiency while preserving valid data through validation checks before and during erasure operations.
3Adaptability or versatility
If data is written randomly to the SSD without categorization, then storage flexibility is maintained, but write amplification increases due to uneven valid page distribution
Solution Approach 1:
The patent segments storage space into multiple pools (hot, warm, cold) with distinct characteristics. Each pool is further divided into physical blocks that are managed independently. This segmentation enables targeted garbage collection and reduces write amplification by ensuring that valid pages are concentrated in specific pools rather than randomly distributed, while still maintaining overall storage flexibility through dynamic block allocation across pools.
Solution Approach 2:
The patent changes the parameter of data organization from random distribution to categorized distribution based on access patterns. By adjusting the organization parameter (data categorization into hot/warm/cold pools), the system reduces write amplification while maintaining adaptability through dynamic adjustment of pool characteristics and block allocation strategies.
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
The present invention provides a storage system, a storage management apparatus, a storage, a hybrid storage apparatus, and a storage management method. The storage system includes a storage and a manager. The manager receives a data write request, determines a data category of a to-be-written data page, designates different logical blocks for to-be-written data pages of different data categories, and sends a data write instruction to the storage, where the data write instruction carries to-be-written data pages corresponding to at least one designated logical block. The storage is configured to receive the data write instruction, and write, according to the data write instruction, the to-be-written data page corresponding to the at least one designated logical block to a physical block corresponding to the at least one designated logical block, where to-be-written data pages written to each physical block are corresponding to a same logical block. By using the present invention, write amplification resulting when a physical block of an SSD is erased can be effectively reduced.