Aggregated Queues for SSD Data Distribution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing SSD writing algorithms result in unsatisfactory data distribution in NAND-flash memory, leading to increased read time due to the sequential aggregation of data blocks in a flushing queue, necessitating multiple read commands and prolonged data retrieval.
Innovation Solution
A method involving the aggregation of data into multiple designated queues based on a strategic aggregation strategy, followed by data interception and writing, optimizes data distribution and reduces read time by aligning data blocks with multi-plane page alignment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If data blocks are sequentially aggregated in a single flushing queue regardless of data block length, then the writing process is simple, but the data distribution written into the memory becomes unsatisfactory, requiring multiple read commands and prolonging read time
Solution Approach 1:
The patent divides a single flushing queue into multiple aggregated queues (first aggregated queue, second aggregated queue, etc.), each handling data blocks of specific length ranges. This segmentation allows data to be distributed more effectively across memory planes, reducing the number of read commands needed and thereby reducing read time.
Solution Approach 2:
The patent introduces a new dimension to data aggregation by organizing queues based on data block length characteristics rather than simple sequential order. This dimensional change in organization enables better alignment with multi-plane page structures in NAND-flash memory, improving read efficiency.
2Productivity
If data blocks are aggregated without considering data block length, then the aggregation process is simple, but multiple read commands are required when reading data distributed over multi-plane pages
Solution Approach 1:
The patent applies different aggregation strategies to different aggregated queues based on their specific characteristics. Each queue is optimized for specific data block lengths, creating local quality variations that match the underlying memory structure. This allows data to be written in a manner that minimizes cross-plane reads, improving read operation efficiency.
Solution Approach 2:
The patent changes the aggregation parameter from simple sequential order to data block length-based categorization. By organizing queues according to data block length parameters, the system achieves better alignment with multi-plane page structures, reducing the number of read commands required and improving overall productivity.
Data Source
AI summary
The present disclosure relates to a method for data writing, a device and a storage medium, wherein the method includes acquiring a first data to be written and saving a plurality of sub-blocks of the first data in one or more designated aggregated queue of multiple aggregated queues according to an aggregation strategy; performing, in each designated aggregated queue, data interception on a plurality of sub-blocks in a current queue to obtain a second data to be written; and writing the second data in a storage device. The data distribution written in the storage device becomes more ideal by a multi-queue aggregation, and thus the method for data writing according to the embodiments of the present disclosure can reduce the time consumed in reading the storage device effectively.