Hybrid-Page Storage Array for SSD Write Amplification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Solid-state disks (SSDs) in all-flash arrays experience high write amplification due to the need to copy valid data to new locations for garbage collection, especially when dealing with small, frequently accessed data, which increases the workload and reduces storage efficiency.
Innovation Solution
Implementing a hybrid-page logical block structure within the storage array, where small-page SSDs are used for frequently accessed data and large-page SSDs for less frequently accessed data, utilizing conversion tables to manage data migration and access requests, thereby reducing write amplification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If an SSD with a large page size is used to store frequently accessed small data, then the storage capacity is improved, but write amplification increases significantly
Solution Approach 1:
The storage system is segmented into two distinct SSD types: large-page SSDs for infrequently accessed data and small-page SSDs for frequently accessed small data. This segmentation allows each SSD type to operate in its optimal performance range, preventing write amplification in large-page SSDs while maintaining high storage capacity.
Solution Approach 2:
Different regions of the storage system are assigned different page sizes based on access patterns. Frequently accessed small data is localized to small-page SSDs, while infrequently accessed data resides in large-page SSDs. This local quality optimization ensures that write operations occur only where necessary, reducing overall write amplification.
2Productivity
If garbage collection is performed on large-page SSDs storing small frequently accessed data, then storage management is improved, but additional data writing is added
Solution Approach 1:
The problem of write amplification during garbage collection is solved by extracting frequently accessed small data from large-page SSDs and placing it in small-page SSDs. This extraction eliminates the need for costly copy operations during garbage collection of large-page SSDs, as small-page SSDs can be garbage collected more efficiently due to their smaller page size matching the data granularity.
3Speed
If data is migrated between large-page and small-page SSDs, then access efficiency is improved, but system complexity increases
Solution Approach 1:
A conversion table acts as an intermediary data structure that tracks the mapping between logical addresses and physical locations across large-page and small-page SSDs. This intermediary simplifies the migration process by providing a centralized lookup mechanism, reducing the complexity of managing data across heterogeneous SSD types while maintaining fast access through direct address translation.
Data Source
AI summary
A disk array includes SSD of different page sizes and a virtual disk is constructed based on the SSDs of different page sizes. The constructed virtual disk formed by a hybrid-page chunk which includes a large-page chunk, a small-page chunk, and the small-page chunk, is not provided for the host for use, and is only used to migrate data corresponding to the large-page chunk. Based on the virtual disk formed by the hybrid-page chunk, the data with a large granularity can be stored in the SSD with a large page size, and data with a small granularity is migrated to the SSD with a small page size.